CN1499435A - Figured pattern identifier - Google Patents
Figured pattern identifier Download PDFInfo
- Publication number
- CN1499435A CN1499435A CNA2003101038188A CN200310103818A CN1499435A CN 1499435 A CN1499435 A CN 1499435A CN A2003101038188 A CNA2003101038188 A CN A2003101038188A CN 200310103818 A CN200310103818 A CN 200310103818A CN 1499435 A CN1499435 A CN 1499435A
- Authority
- CN
- China
- Prior art keywords
- coin
- mentioned
- floral designs
- image
- recognition device
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Testing Of Coins (AREA)
- Inspection Of Paper Currency And Valuable Securities (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
To enhance identification performance in a pattern identifying device for identifying a pattern of an identifying object such as a coin, based on an image of the identifying object acquired by an image pick-up means. The device is provided with: an illuminating means for irradiating with light the identifying object passing through a passage; and an image pick-up means for photoreceiving reflected or transmitted light from the identifying object, and identifies the pattern of the identifying object based on the image imaged by the image pick-up means. The image of the identifying object is divided into a plurality of areas 143, and discriminated based on a degree of consistency in every of the areas with a reference image to reduce influences by contamination of the identifying object and a passage 13. The pattern is accurately identified thereby.(C)2004,JPO.
Description
Technical field
The present invention relates to read coin etc. floral designs, judge and the floral designs recognition device of being familiar with the true and false and kind.
Background technology
In recent years, automatic vending machine and changer are popularized, and require use therein Coin identifying apparatus to have high recognition performance.
Existing Coin identifying apparatus.Have the magnetic sensor that the iron core by coil and ferromagnetic material constitutes, material that can be by detecting coin and thickness, external diameter are discerned coin.This existing Coin identifying apparatus, such as, open in Japanese kokai publication hei 8-161574 communique.
But, be the material of coin and thickness, external diameter what utilize that this existing structure discerns, the floral designs that detect coin are just very difficult.Therefore, illegally use the different counterfeit money of floral designs sometimes.For example the illegal use of altered coin that the thickness of material and external diameter similar foreign coin is about equally cut etc. utilizes existing structure just can not prevent.
Summary of the invention
Floral designs recognition device of the present invention has the Lighting Division of using the rayed identifying object and acceptance is reflected by this identifying object or the image pickup part of the light of transmission.So the floral designs recognition device can be discerned the floral designs of identifying object according to the image that image pickup part is obtained.At this moment, utilize the normal reflection light of self-identifying object or the non-incident portion of regular transmission light to discern.
Description of drawings
Fig. 1 is the front elevation of summary that the Coin identifying apparatus of embodiments of the present invention is shown.
Fig. 2 is near the sectional view of image pickup part of the Coin identifying apparatus of embodiments of the present invention.
Fig. 3 is near the front elevation of Lighting Division of the Coin identifying apparatus of embodiments of the present invention.
Fig. 4 is the sectional view of path of the Coin identifying apparatus of embodiments of the present invention.
Fig. 5 is the thickness of the Coin identifying apparatus of embodiments of the present invention, near the sectional view of sensor of material dual-purpose.
Fig. 6, Fig. 7 are the pie graph of control circuit of imageing sensor of the Coin identifying apparatus of embodiments of the present invention.
Fig. 8 is the block diagram of formation of control circuit that the Coin identifying apparatus of embodiments of the present invention is shown.
Fig. 9 is the process flow diagram of the work of the Coin identifying apparatus of explanation embodiments of the present invention.
Figure 10 detects the process flow diagram of the work of convex-concave and floral designs for explanation utilizes the Coin identifying apparatus of embodiments of the present invention.
Figure 11 is the concept map of camera watch region of image pickup part of the Coin identifying apparatus of embodiments of the present invention.
Figure 12 A is near the sectional view the coin when a coin position Lighting Division with outlet side throws light on when the Coin identifying apparatus shooting of embodiments of the present invention.
Figure 12 B be when the shooting of the Coin identifying apparatus of embodiments of the present invention in the coin position mode chart of the view data during only with the Lighting Division illumination of outlet side.
Figure 12 C is near the sectional view the coin when a coin position Lighting Division with the input oral-lateral throws light on when the shooting of the Coin identifying apparatus of embodiments of the present invention.
Figure 12 D be when the shooting of the Coin identifying apparatus of embodiments of the present invention in the coin position mode chart of the view data during only with the Lighting Division illumination that drops into oral-lateral.
Figure 13 is the mode chart that the center external diameter of the Coin identifying apparatus of embodiments of the present invention is detected the 3rd view data that describes.
Figure 14 A proofreaies and correct the mode chart of the 4th view data that describes to the coordinate transform of the Coin identifying apparatus of embodiments of the present invention and multiplying power.
Figure 14 B proofreaies and correct the mode chart of the 5th view data that describes to the coordinate transform of the Coin identifying apparatus of embodiments of the present invention and multiplying power.
The mode chart of the rectangular coordinate inverse transformation of the benchmark image of new 500 Japanese yen coins that Figure 15 describes for the threshold setting to the Coin identifying apparatus of embodiments of the present invention.
Figure 16 A is a mode chart of the floral designs of the Coin identifying apparatus of embodiments of the present invention being compared the 6th view data that describes.
The mode chart of the 6th view data that Figure 16 B describes for the Region Segmentation to the floral designs comparison of the Coin identifying apparatus of embodiments of the present invention.
Embodiment
Fig. 1 is the front elevation of summary that the Coin identifying apparatus of embodiments of the present invention is shown.Be provided with the input port 12 of coin on the top of Coin identifying apparatus main body 11,12 paths 13 that are connected with coin downwards from the input port.The part of the sidewall of path 13 is made of optical transparent body 14, disposes Lighting Division 15 and the image pickup part 16 that detects the coin floral designs at the back of optical transparent body 14.And, at the dirty sensor 17 that detects the thickness material that disposes of path 13.In addition, path 13 is connected with the outlet 18 of the coin of the bottom that is positioned at Coin identifying apparatus main body 11.
Fig. 2 is near the sectional view the image pickup part 16.The part of one side's of formation path 13 sidewall 21 forms the bottom surface 22 of path.The coin 23 that drops into rotates on bottom surface 22.Optical transparent body 14, the part of the opposing party's of formation path 13 sidewall 24.15 pairs of coins 23 of Lighting Division of being made up of a plurality of LED25 and soft wiring substrate (not shown) throw light on.Image pickup part 16 in the configuration of sidewall 24 sides detects the reflected light that is reflected by coin 23.Image pickup part 16 has the sensitive surface 27 on imageing sensor (below be called sensor) 26, the sensor 26 and the reflected light of coin 23 is imaged in the lens 28 of sensitive surface 27.
The bottom surface 22 of path 13 is arranged on the sidewall 21 with Lighting Division 15 and image pickup part 16 contrary sides.Can prevent from thus on bottom surface 22, shadow to occur and reduce recognition performance.In addition, the size of the width 22A of bottom surface 22 should be unable to hinder the edge that shines coin 23 from the illumination light of Lighting Division 15.The formation that is bottom surface 22 should be that the minimum thickness than the coin 23 that will discern can not produce shadow for little on the edge of coin 23.In addition, even its size for the coin 23 of minimum thickness along sidewall 24 by not falling yet.
In addition, the relative vertical direction of path 13 are tilted so that coin 23 along stable the passing through of a certain side of sidewall 21,24.Coin 23 is to pass through along sidewall 21 herein.Might not need with the sidewall 24 of the contrary side that tilts.In the present embodiment, also make the adhere to reduction of dirt to these parts with the contrary side configuration optical transparent body 14, Lighting Division 15 and the image pickup part 16 that tilt.In other words, optical transparent body 14 by formed and divided into path 13 sides and image pickup part 16 sides by glass and plastics etc., prevents that dust and coin 23 etc. are from path 13 sides intrusion image pickup part 16 sides.But also not necessarily optical transparent body 14 must be set.
In addition, image pickup part 16 and Lighting Division 15 preferably are arranged at relative path 13 and are a side of vertical top.Like this, just can reduce dirt to image pickup part 16 and adhering to of Lighting Division 15 and correctly discern floral designs.
The shape of Lighting Division 15 is have long limit and minor face non-circular, such as oval or elliptical shape, makes long limit roughly consistent with the direction of passage of coin 23.Thus, coin 23 can enlarge the permission for the precision of the timing of suitable position, the center of the sensitive surface 27 that detects the sensor 26 that arrives with coin 23.Therefore, even coin 23 to pass through speed be not certain free-falling and rotate, sensor 26 also can be discerned coin 23.In addition, coin position during with shooting, the center by making Lighting Division 15 (below be called the position) 33 center is roughly consistent, and illumination can homogenising.In addition, the center of the sensitive surface 27 on the sensor 26 also is disposed at these center.
In addition, in Fig. 2, normal reflection light 29 expression is from the illumination light of Lighting Division 15 reflected light component by near the planar portions normal reflection the outer peripheral edges portion of the coin 23 of maximum outside diameter to be identified.In other words, normal reflection light 29 is reflected light component that incident angle and reflection angle equate.If normal reflection light 29 incides image pickup part 16, then can on the image of input, produce very bright irregular and hinder correct identification.Therefore, the formation of image pickup part 16 makes the non-incident portion can be used to from the normal reflection light of coin 23 discern.In specific words, the configuration of path 13, Lighting Division 15 and image pickup part 16 makes normal reflection light 29 can not incide image pickup part 16.In other words, in Fig. 3, the configuration by position relation makes normal reflection position 34 not prevent in the position 33 the irregular generation of image in the scope of in-position 33.
In addition, become by the shape that makes Lighting Division 15 and to have the non-circular of long limit and minor face, the shape of normal reflection position 34 is also become have the non-circular of long limit and minor face.So normal reflection position 34 just is difficult in the scope of in-position 33.In addition, roughly consistent by the long limit that makes Lighting Division 15 with the direction of passage of coin 23, can make the long limit of normal reflection position 34 of image pickup part 16 and the direction of passage almost parallel of coin 23.So, can enlarge permission for the precision of the timing of suitable position, the center of detecting the image pickup part 16 that arrives with coin 23, even coin 23 to pass through speed be not certain free-falling and rotate, also can discern.In addition, roughly consistent by the center that makes Lighting Division 15 with the center of the position 33 of taking the photograph Fu portion 16, can make the center of normal reflection position 34 roughly consistent with the center of position 33.In addition, can make the illumination homogenising.Perhaps, also can go up the vacancy portion that LED25 is not disposed in setting at the position 15A (long leg branch) of the Lighting Division 15 of the normal reflection position 34A in the scope with in-position 33.So just vacancy portion can be set on normal reflection position 34.No matter be any structure, all be easy to make path 13, Lighting Division 15 and image pickup part 16 to be configured as and make the normal reflection light 29 that sends from coin 23 incide outside the position 33 of image pickup part 16.In addition, can obtain same effect by the center that makes Lighting Division 15 with the center of image pickup part 16 is roughly consistent.
In embodiments of the present invention, image pickup part 16 and Lighting Division 15 are disposed at the same side of relative path 13, the reflected light that is used to the self-identifying object is discerned.But, also image pickup part 16 and Lighting Division 15 can be disposed at the contrary side of relative path 13, the transmitted light that is used to the self-identifying object is discerned.When utilizing transmitted light to discern, the transmitted light component that incident angle equates with emergence angle (regular transmission light), the normal reflection light 29 when being equivalent to utilize reflected light to discern is effective for the irregular same structure of image.
In the present embodiment,, make incident angle become low angle, under dark field illumination, mint-mark is appeared more the irradiates light of coin 23 in order to detect the mint-mark floral designs of coin 23.In addition, constitute a plurality of LED25 of Lighting Division 15, on structure, be divided into two groups in the straight line 35 of path 13 by central vertical by Lighting Division 15.Be to be divided into the Lighting Division 36 of outlet 18 sides and the Lighting Division 37 of input port 12 sides on the structure, Lighting Division 36 and Lighting Division 37 can be independently luminous.
Fig. 4 is the sectional view of the part of path 13.The opposing party's the sidewall 24 of contrary side that is positioned at the sidewall 21 of the side with bottom surface 22 is provided with discharges assisted parts 41.Drop into deformed coin and by the coin of adhesional wettings such as rainwater or adhere to adhesion coin or the like and when stopping in the path 13, can be with the distance expansion of sidewall 21 and sidewall 24.So being fallen downwards, returns by coin 23 from exporting 18.Discharge assisted parts 41, remain in sidewall 24 sides by top and make the discharge of coin 23 more reliable coin 23.
Fig. 5 is near the sectional view the sensor 17.Constituting E core that ferromagnetic material is installed respectively on the sidewall 21,24 of path 13 relatively (below be called iron core) 51,52.The length of unshakable in one's determination 51,52 directions parallel with path 13 is littler than the external diameter of the coin 23 of the minimum of identification.Unshakable in one's determination 51,52 are disposed at the center of coin 23 by near the position unshakable in one's determination 51,52 the center.Sensor 17 is made of two coils 53,54 of iron core 51 and portion's coiling within it and two coils 55,56 of iron core 52 and portion's coiling within it.The 55 relative path 13 subtends configurations of coil 53 and coil anti-phasely are connected in series that to make mutual inductance be negative.By this structure, the magnetic line of force direction of the AC magnetic field that produces in path 13 shown in the dotted line among Fig. 5, is parallel to the surface of coin, and coil 53,55 can detect the thickness of coin delicately.In addition, the 56 relative path 13 subtends configurations of coil 54 and coil, homophase is connected in series and makes mutual inductance for just.By this structure, the magnetic line of force direction of the AC magnetic field that in path 13, produces, shown in the dot-and-dash line among Fig. 5, perpendicular to the surface of coin, coil 54,56 can detect the material of coin delicately.So, sensor 17 can be also used as thickness transducer and material sensor.This is for miniaturization and reduces cost, but also can be respectively sensor independently.
Fig. 6 is the pie graph of the control circuit of the sensor 26 of formation image pickup part 16.What adopt as sensor 26 in the present embodiment, is MOS type two dimension (face) imageing sensor.In addition, also be not limited thereto, also can adopt the time series data of CCD and one dimension (line) sensor etc.,, can obtain two-dimensional image data at a high speed if adopt two-dimensional image sensor.
As a kind of horizontal scanning circuit 62 and vertical scanning circuit 63 of shift register, difference controlling level sweep trace 64 and vertical scan line 65.So, utilize sweep trace 64,65 can select the pixel of forming by the photo detector that is configured as the ranks shape 61.Horizontal scanning circuit 62 is exported the view data that each pixel 61 detects as output 67 orders.At this moment, utilize 63 sweep traces of horizontal scanning circuit 62 and vertical scanning circuit to control and only to export the view data of specific region 66 part.The output of the view data of this part can realize with the imageing sensor of MOS type.By equipping this image pickup part 16, the output of view data only can be focused on essential part and reach the high speed of reading.
In addition, when each pixel 61 was square shape, the pixel count of the horizontal direction of sensor 26 was more than the pixel count of vertical direction, and the face of detecting is a rectangle.Thus, the long side direction of sensor 26 and the direction of passage of coin 23 are configured to roughly consistently just can obtain rectangular view data.By making the consistent pixel that effectively utilizes sensor 26 of long side direction, can enlarge for detecting and the permission of coin 23 arrival by the precision of the timing of suitable position, the center of the sensitive surface 27 of sensor 26 with direction of passage.
In addition, sensor 26 will not have precedence over the scanning of output 67 horizontal scanning circuits that are connected 62 and carry out with the scanning of the output 67 vertical scanning circuit 63 that are connected.In other words, cycle one side of horizontal scanning is short.By such formation, for shortening the level stand-down (horizontal blanking interval) in the picture output signal that detects the zone of growing crosswise.In other words, can be in shorter time sense data.
In addition, as shown in Figure 7, the formation of sensor 26 also can be for having array and control circuit 68 and the neural network 69 as the pixel 61 of photo detector.Control circuit 68 is supplied with the voltage that is used for sensitivity control to the row of pixel ranks.Neural network 69 is handled the electric current that flows into ground connection from the row of pixel ranks.Because in Japanese kokai publication hei 6-139361 number (embodiment 6 etc.), be documented, omit it and constitute and the explanation of working about control circuit 68 and neural network 69.
Owing to have these parts, in sensor 26, can carry out, such as, the long-pending of 3 * 3 wave filter ranks and image exported with computing.In other words, can realize in sensor 26 that edge detection feature described later and edge emphasize function.Based on the computing of the pixel data around the adjacency, can carry out especially at high speed.In addition,, be not only edge image, also the original image former state can be exported by the setting of change wave filter.Like this, by control circuit 68 and neural network 69 are set, can in sensor 26, carry out the long-pending and computing of wave filter ranks and image in sensor 26.Therefore, do not need to be provided with in addition rim detection portion.So can shorten the processing time of rim detection and make the data processing high speed.In addition, the sensor 26 of built-in this edge detection feature is a kind of functional device that has image pickup part 16 and rim detection portion concurrently.Image pickup part 16 and rim detection portion are provided with also passable respectively.
Fig. 8 is the block diagram of formation that the control circuit of whole Coin identifying apparatus is shown.The thickness transducer 71 that constitutes sensor 17 is connected with oscillatory circuit 72, and detects portion 74 through the rectification circuit 73 that waveform is transformed to the signal of representing oscillation level from sine wave with thickness and be connected.The material sensor 75 that constitutes sensor 17 is connected with oscillatory circuit 76, detects portion 78 through rectification circuit 77 and material again and is connected.In addition, the output of rectification circuit 73,77 also is input to the input that detects coin 23 inputs and detects portion 79.Input detects portion 79, control Lighting Division 15 and image pickup part 16.15 pairs of coins of Lighting Division 23 carry out illumination, and image pickup part 16 detects the reflected light of coin 23 reflections, and output to the A/D converter section 80 that the simulated image data that will take is transformed to multivalue digital signal.
The output of A/D converter section 80 is input to the portion 79 that detects that drops into, adjustment part 81, dewfall respectively and detects portion 82, concavo-convex portion 83, rim detection portion 84 and the center external diameter test section 85 of detecting.The 81 pairs of image pickup parts 16 in adjustment part are controlled adjustment.Dewfall detects portion 82 and detects near the image pickup part 16 path 13 dewfall whether.Concavo-convexly detect the concavo-convex degree that portion 83 detects coin 23.The edge part of rim detection portion 84 inspection image data.The center and the external diameter of the coin part of center external diameter test section 85 inspection image data.In addition, in the above description, be built-in with edge detection feature in the sensor 26, but herein will be image pickup part 16 and separately narration of rim detection portion 84.
Center external diameter test section 85 is connected with rim detection portion 84, the interim detection unit 86 of value of money kind and multiplying power correction portion 87.Value to the external diameter that detected by center external diameter test section 85 in the interim detection unit 86 of value of money kind is similar to, interim two kinds of value of money kinds judging the maximum of considering to be.It is polar coordinate converting section 88 from rectangular coordinates transformation that the output of rim detection portion 84 then is input to view data.The shooting multiplying power that in multiplying power correction portion 87 coin 23 assembly error that passes through position and image pickup part 16 in path 13 etc. is caused is revised and coordinate converting section 88 is controlled.
The output of coordinate converting section 88 then is input to the threshold setting portion 89 of the threshold value of setting binaryzation and view data is transformed to the binaryzation portion 90 of bi-level digital value from many-valued digital value.The output of binaryzation portion 90 is connected with anglec of rotation test section 91.Anglec of rotation test section 91 at the benchmark image data that are stored in the storage part 92, detects the anglec of rotation of the view data of the coin 23 that drops into.So anglec of rotation test section 91 is to floral designs comparison portion 93 outputs of the view data and the benchmark image data of the coin 23 of comparison input.
The outgoing side of the interim detection unit 86 of value of money kind is connected with multiplying power correction portion 87, threshold setting portion 89, anglec of rotation test section 91 and floral designs comparison portion 93.Storage part 92 is connected with anglec of rotation test section 91, floral designs comparison portion 93 and each comparing section 94-97.Input thickness detects portion 74 respectively in comparing section 94-97, material detects portion 78, the concavo-convex output that detects portion 83 and floral designs comparison portion 93.Comparing section 94-97 compares itself and the benchmark that is stored in the storage part 92, and in permissible range the signal of the kind of output expression genuine note just when consistent.In addition, with any reference value when all inconsistent just the output expression be the signal of counterfeit money.
In detection unit 99, so long as represent it is the genuine note of same kind all from the signal of comparing section 94-97, the signal of this genuine note kind of output expression just, otherwise will represent that then the signal of counterfeit money exports as decision signal 100.Each output that dewfall detects portion 82 and switching part 98 also is input in the detection unit 99.Control circuit with these each ones is made of microcomputer etc.
Below with reference to Fig. 9, Figure 10 and Fig. 8 etc. the work of the Coin identifying apparatus that as above constitutes is illustrated.At first, in S11-S17, detect the input of coin 23.The input of so-called herein coin 23 and input detect and also comprise the position that arrives identification coin 23 and the meaning that arrival detects.
Detect the setting (S11) that portion 79 carries out Beat-of-the-night-watches picture portion 16 in order to detect the coin input by image pickup part 16, to drop into.In specific words, be image pickup part 16 to be set at only export first among the image pickup scope 101 in being shown in Figure 11 and drop into the parts of images that detects zone 102.And the shooting of initialization (replacement) etc. of carrying out each pixel 61 of sensor 26 is prepared.After these were set, image pickup part 16 began shooting.
Secondly, drop into the portion 79 that detects, drop into the whole pixels 61 that detect in the zone 102 first and all become the moment that is subjected to light state, make 15 short time of Lighting Division luminous (S12).By making its speed of passing through with relative coin 23 only is the short time luminously can needn't make coin 23 stop to improve accuracy of identification with regard to obtaining near the image of stationary state.In addition, can make with big current drives LED25 the illumination high illuminationization in long lifetime.In addition, be to make Lighting Division 36 and Lighting Division 37 boths luminous in Lighting Division 15.Below unless otherwise indicated, the so-called Lighting Division 15 that makes is luminous, represents to make Lighting Division 36 and Lighting Division 37 boths luminous.
So, image pickup part 16, with as shown in figure 11 first drop into the view data input (S13) detect zone 102.Afterwards, when image pickup part 16 detects, just drop into to detect zone 103 and afterwards the 3rd input is detected zone 104 and import equally second.Nextly constantly return first and drop into and detect zone 102 what detect whether coin drop into.Like this, detect regional 102-104 and switch in proper order to drop into and detect, even when coin 23 and path 13 are subjected to dirt and stain, also can detect the input of coin 23 reliably by three places are dropped into.Three places drop into and detect regional 102-104, and the configuration of coin position 33 when making a video recording is relatively preferably carried out locating against calculating according to the speed of importing to following floral designs with required time and coin 23 till the illumination (S40) of passing through.So, roughly made a video recording the center of the sensitive surface 27 on Lighting Division 15 and sensor 26 with the coin 23 that on average passes through by speed.In addition, each input detects regional 102-104 and preferably does one's utmost to leave distance, so that coin 23 and path 13 are difficult to be subjected to simultaneously the influence of dirt.In addition, constantly can not depart from order to make detecting that the external diameter that utilizes coin 23 produces, from each input detect regional 102-104 from the bottom surface 22 height, be disposed near the center of coin position 105 of minimum coin to be identified and up and down.In addition, Pei Zhi input detects regional 102-104 up and down, detects zone 103 a little near the input port configuration than near the input the center of the coin position 105 that is disposed at minimum coin.
Detect the view data of regional 102-104 based on the input of such input, the dewfall of path 13 that comprises the sensitive surface 27 etc. of optical transparent body 14 and lens 28, sensor 26 detects (S14).The whether judgement of dewfall is to utilize owing to have dew to focus on coin 23 lip-deep Focus Clubs when adhering on optical transparent body 14 and lens 28 etc. to depart from and make the input picture this point that thickens.If drop into the deviation of the view data detect regional 102-104, promptly disperse, three are all continued less than the state of certain value, just are judged to be dewfall.When can obtaining dispersion easily, by judgement disperse I very to detect because the departing from of the focus of the image pickup part 16 that dewfall produces.In addition,, just do not judge dewfall, improve the reliability of dewfall judgement so can prevent misinterpretation as long as have a punishment to loose greatly among the regional 102-104 owing to detecting in input.In the occasion of judging dewfall,, prevent owing to delaying work that abnormal view data produces by omitting the control and treatment of sensor 26.Or, prevent that near the dewfall of the dewfall of the path 13 the image pickup part 16 from preventing portion's (not shown) work.Or, also can communicate by letter with the control part (not shown) of notice automatic vending machine with regard to dewfall generation this point, show and notified to the operator of automatic vending machine.Like this, by using image pickup part 16, just can not have to detect dewfall under the situation of sensor special.In addition, by utilize can the output image sensor 26, can obtain dewfall at short notice and detect the image of usefulness and detect dewfall at a high speed.
Utilize the input of image pickup part 16 to detect (S15), the variation of the view data that detects by image pickup part 16 from the illumination light of Lighting Division 15 irradiations of the coin that is dropped into of utilization 23 reflections.Among detecting view data average of regional 102-104 in input or disperseing,, just be judged to be coin 23 and drop into so long as there is a value to change when surpassing certain value from holding state.By average and disperse among, as long as there is a variation just can be judged to be coin 23 inputs at least, even also can detect input reliably when the opposing party does not change.In addition,, can be judged to be input,, also can detect the input of coin 23 reliably even occasions such as dirt adheres to are arranged at path 13 as long as there is a place to change once the side by obtaining average in many places and disperseing.
When the sensor that detects usefulness in input was provided with in addition, this sensor is set must have living space.If this sensor is an optical sensor, just must prevent illumination interference of light with Lighting Division 15.In addition, when input detects assembling with sensor and image pickup part 16, can produce the range deviation of two sensors and image takes place detect error constantly.At this point, as present embodiment, image pickup part 16 is dropped into detect to be also used as to drop into by the irradiates light that utilizes Lighting Division 15 to produce to detect sensor, detect input in many places easily.And, can cut down the space, correctly detect image under the influence of do not thrown light on interference of light and assembling deviation.In addition, by the sensor 26 that adopts exportable parts of images, can obtain dropping into the image that detects usefulness at short notice and detect input at a high speed.
Even judge that for dewfall dewfall but path 13 are not subjected to that dirt etc. adheres to and the occasion that can not detect the input of coin 23 also can be utilized thickness transducer 71 and material sensor 75 to drop in the present embodiment and detect usefulness (S16, S17) in neutralization.So,, just can judge it is that coin 23 drops into if detect input at any one.So, adhere to etc. and also can detect input reliably even path is subjected to dirt.If 12 coins that drop into 23 are near sensor 17 from the input port, then the impedance of coil 53-56 changes, thereby the oscillation level of oscillatory circuit 72,76 changes.If the oscillation level during from standby changes when surpassing certain value, just be judged to be coin 23 and drop into, carry out thickness and material and detect (S20).
Detect in (S20) at thickness and material, from corresponding to because the variation of the oscillation level of the oscillatory circuit 72,76 of the change of the impedance of the coil 53-56 that coin 23 produces detects thickness and material.As mentioned above, to the oscillation level of the thickness transducer 71 of the thickness sensitivity of coin 23, detect portion 74 through the rectification circuit 73 that waveform is transformed to the signal of expression oscillation level from sine wave with thickness and be connected.So, thickness detect portion 74 detect coin by the time oscillation level maximum variable quantity and output to comparing section 94.Material detect portion 78 detect too to the oscillation level of the material sensor 75 of the material sensitivity of coin 23 coin by the time maximum variable quantity and output to comparing section 95.
In the occasion of utilizing image pickup part 16 to drop into to detect (S15), parallelly carry out concavo-convex and floral designs detect (S21) and thickness and material and detect (S20).
Image detects, and as shown in figure 10, at first begins to carry out from adjusting operation (S22).Because material and dirt, the dirt of path 13 and the degradation temperatures characteristic of Lighting Division 15 etc. of the coin 23 that drops into, the image quality of the view data that detects changes, and brightness and contrast change up and down.To this, whenever coin 23 by the time, before carrying out the input of following concavo-convex input and floral designs, all uses and these identical image pickup part 16 and Lighting Divisions 15 of importing use each time, obtain the parts of images of coin 23 in advance.According to this image of obtaining, the initial conditions of image pickup part 16 is adjusted so that the image quality of the view data that image pickup part 16 detects is certain as far as possible.In specific words, be to adjust sensitivity, promptly be contained in the gain and the bias level of the amplifying circuit in the sensor 26.As shown in figure 11, initial conditions setting regions 106 is set at the coin position 107 that its position, big I are contained in the maximal rate of the coin position 105 of minimum coin of minimum speed of anticipation and anticipation fully.In Figure 11, in fact among the required time section that detects from input till importing the input that imposes a condition, coin 23 is moving always.But it is for the purpose of simplifying the description, identical when the coin position 105 of supposing the minimum coin of minimum speed detects with input.
Adjusting the initial of operation (S22), in order to make high speed readout time, the output image zone of image pickup part 16 only is being set among initial conditions setting regions 106, just beginning make a video recording (S31) after preparing making a video recording.Make 15 short time of Lighting Division luminous (S32), the view data (S33) of input initial conditions setting regions 106.Exactly to carry out reading in advance based on the anticipation position of coin 23 of the position of storage.In the present embodiment, very bright pixel count and the very dark pixel count corresponding to this view data changes the gain and the bias level that are contained in the amplifying circuit in the sensor 26, carries out the initial conditions of image pickup part 16, i.e. the setting of sensitivity (S34).In other words, corresponding to the pixel count below second brightness of pixel count more than first brightness of regulation and regulation, adjust the initial conditions of image pickup part 16.Therefore, the comparison of pixel data and the counting of pixel count can be easy to adjust, moreover, and also can be according to the mean value of view data and deviation etc.In addition, do not change the gain and the bias level that are contained in the amplifying circuit in the sensor 26, but the reference voltage of change A/D converter section 80 also can obtain same effect.Like this, obtain the image of coin 23 in advance, adjust the initial conditions of image pickup part 16 etc. according to the image of obtaining, just can obtain the view data of certain image quality by utilizing Lighting Division 15 and image pickup part 16.In addition, adjustment amount is stored, when each coin 23 drops into, revised adjustment amount, can reduce the influence that parts deterioration etc. changes through the time and keep recognition performance.By at the anticipation position of coin 23 input picture, can obtain the image of coin 23 reliably, image pickup part 16 is carried out suitable adjustment.In addition, by using the sensor 26 of image that can output, just can obtain at short notice to adjust the image of usefulness and image pickup part 16 is carried out high speed adjustment.
In addition, even lighting conditions such as the luminous intensity of change Lighting Division 15 and fluorescent lifetime also can obtain same effect.In addition, in the occasion of adjusting lighting condition, specifically, be driving power (such as the drive current of LED25) and the fluorescent lifetime that changes Lighting Division 15.At this moment, must add power circuit or carry out reading in regularly of complicated control break sensor 26.So the initial conditions of adjusting image pickup part 16 is just very simple in contrast to this.
In addition, with light reflective sensor and image pickup part was arranged in 16 minutes and the technology adjusted corresponding to the light income of this light reflective sensor, such as, open in TOHKEMY 2000-331211 communique.In the present embodiment, obtain the image of adjusting usefulness, just the adjustment sensor special can be set by using the identical image pickup part 16 that uses with concavo-convex input and floral designs input.And, adjust and discern by adopting sensor separately, just can under the situation of the influence that is not detected error, adjust.
The subsequent handling of S22 is a concavo-convex input operation (S23).Genuine note duplicated or the paper of the floral designs printing of genuine note etc. pasted on the metal and the counterfeit money of making (below be called to duplicate paste counterfeit money), can not find in following floral designs comparisons (S51) has too big difference with genuine note and is difficult to get rid of reliably.Therefore, in the present embodiment, duplicate the stickup counterfeit money by detecting concavo-convex degree eliminating.Therefore, as previously mentioned, Lighting Division 15 is divided into the Lighting Division 36 that exports 18 sides and the Lighting Division 37 of input port 12 sides on constituting, Lighting Division 36 and Lighting Division 37 can be independently luminous.Shown in Figure 12 A, if exist concavo-convexly, then in the position shown in Figure 12 B shadow appears in first view data when only utilizing Lighting Division 36 to throw light on.And, shown in Figure 12 C, in the position shown in Figure 12 D shadow appears in second view data when only utilizing Lighting Division 37 to throw light on.Detect for concavo-convex, can utilize the difference of the position that is included in the shadow in these two view data, the difference when promptly throwing light on from different directions.
In the present embodiment, the concavo-convex line 108-110 that detects that is represented by dotted lines in Figure 11 is set at the minimum coin position 107 that its position, length can be contained in the maximal rate of the coin position 105 of minimum coin of minimum speed of anticipation and anticipation fully.In other words, attend school after will storing as the anticipation position of the coin 23 of identifying object and get reflected light.Can obtain the image of coin 23 reliably thus and suitably detect concavo-convex.In Figure 11, in fact among the required time section that detects from input till carrying out the concavo-convex input that detects usefulness, coin 23 is moving always.But for the purpose of simplifying the description, suppose when coin position 105 detects with input identical.
Initial in concavo-convex input operation (S23) is set in these concavo-convex line 108-110 that detects with the output image zone sequence of image pickup part 16, just begins make a video recording (S35) after preparing making a video recording.Then, make 37 pairs one concavo-convex line that detects of Lighting Division 36 and Lighting Division switch short time luminous (S36), input image data (S37) alternately.Repeat this operation, till all concavo-convex first, second view data that detects line 108-110 is imported (S38).
In addition, the interval of input of first image and the input of second image is about the short time of 500 microseconds, and coin therebetween moves on average to be equivalent to the size of 1 pixel by speed calculation.So, as described below, when calculating first view data and second view data poor,, can suppress the influence that coin moves generation by revising departing from of a pixel size.
In the present embodiment, Lighting Division 36 and Lighting Division 37 all are that LED25 by same wavelength characteristic constitutes.Therefore, it is concavo-convex to utilize monochrome image to detect.In addition, concavo-convex Lighting Division 36 and the Lighting Division 37 that detects usefulness, as utilize Fig. 3 illustrated, Lighting Division 15 is divided into two groups of formations.In other words,, floral designs are also used as concavo-convex detecting because detecting the Lighting Division 15 of usefulness, and also dual-purpose of image pickup part 16, so can cutting down cost and space.Adjust and also do not need special-purpose adjustment operation.In addition, luminous in different timings by making Lighting Division 36 with Lighting Division 37, can prevent that the illumination light that Lighting Division 36 and Lighting Division 37 produce from interfering, and can obtain catoptrical image separately reliably.In addition, by adopting the sensor 26 of image that can output, can obtain catoptrical image at short notice and import concavo-convex having or not at a high speed.
In addition, shine with the different wave length characteristic from different directions and judge that the concavo-convex technology of coin is disclosed in, such as, in TOHKEMY 2002-183791 number.Relative therewith, because be to shine with same wavelength characteristic in the present embodiment, it is concavo-convex to need not to use different Lighting Division of emission wavelength and color image sensor and wave filter just can detect.In addition, also can not be subjected to the influence of deviation of the transmissison characteristic of the transducer sensitivity that changes because of wavelength and environment temperature and wave filter to detect concavo-convex degree.
In addition, irradiation and carry out the technology that binaryzation is judged the floral designs of coin respectively from different directions according to each catoptrical intensity, such as, open in Japanese kokai publication hei 7-210720 number.Relative therewith, the concavo-convex calculating in the present embodiment, as described below, be that the difference according to each intensity of reflected light detects.Thus, can be with the many-valued floral designs that detect concavo-convex degree and correctly judge coin 23.
The subsequent handling of S23 is a floral designs input operation (S24).At first, the output image zone of image pickup part 16 is set at the largest coins that its position, big I are contained in the maximal rate of the largest coins of minimum speed of anticipation and anticipation fully.So, after the preparation of making a video recording, just begin shooting (S39).Then, make 15 short time of Lighting Division luminous (S40), import the 3rd view data (S41).
The deviation of the characteristic of package size that the coin position in the view data takes place because of each goods and lens 28 etc. is different.Therefore, must set greatly according to the zone of its big young pathbreaker's floral designs input.To this, in the present embodiment, the position of the bottom surface 22 of view data, promptly the lower end position of coin 23 and shooting multiplying power are stored in variable mode.So, cut down the capacity of the input time of view data and the memory circuit that data storage is used and the processing time of data by calculating the zone that upper end position in the view data of largest coins dwindles the floral designs input.In addition, to the lower end position and the shooting multiplying power of the coin 23 stored, all revise when dropping into genuine note each.Can revise reliably thus, also can prevent performance degradation and keep initial recognition performance through the time variation even take place.
In addition, in the present embodiment, the lower end position of storage coin 23 and shooting multiplying power are also revised, and calculate the center of minimum coin from these parameters.According to any one or the more than one condition in these three conditions, set to drop into and detect regional 102-104, initial conditions setting regions 106, concavo-convexly detect the line that detects that line 108-110, floral designs input area and following center external diameter detect usefulness.Thus, image pickup part 16 can correctly be caught the position of coin 23.So, for path 13 and the assembly error of image pickup part 16 and through the time dimensional characteristic change, and no matter the size of the coin 23 that drops into, promptly the value of money kind can correctly detect respectively.
In addition, the technology of the position of storage coin 23, such as, in Japanese kokai publication hei 10-105765 number, open memory technology in adjustment modes.Relative therewith, in the present embodiment, the lower end position of coin 23 and shooting multiplying power are stored in variable mode, and calculate the center of minimum coin.Therefore, revise, keep recognition performance for following through the time variation at the Shi Douke that inserts coin at every turn.
In follow-up concavo-convex detection operation (S25), calculate concavo-convex degree (S44) from first, second view data of input concavo-convex input operation (S23).Before this, from the 3rd view data inspection center and external diameter (S42), and judge temporarily that according to the external diameter that detects the coin 23 of input is near any value of money kinds (S43).
In S42,, obtain external diameter with center as reference point in order to detect the position of the image suitable with the coin 23 of the 3rd view data.When identifying object is circular coin 23, particularly by obtaining the position that center and external diameter can be easy to detect the coin 23 of view data.Therefore, select a plurality of candidate points suitable, determine interim center with periphery.So obtain the distance of each candidate point and Provisional Center.And obtain the center from being judged to be a plurality of candidate points suitable, detect the position of coin 23 with periphery.Utilize Figure 13 that this is illustrated below.
For the 3rd view data, at first set a plurality of parallel with direction of passage line 111 and vertical lines 112 that detect of detecting.Detect line 111,112, according to the lower end position of coin 23 of storage, as the configured length that must intersect with the periphery of the minimum coin of the maximal rate of the periphery of the minimum coin of the minimum speed of anticipation and anticipation.In the present embodiment, utilize the lower end position correction of the coin 23 of storage to detect line 111,112.And, when dropping into genuine note, all revise detecting line 111,112 at every turn.Like this, carry out detecting of center by anticipation position at coin 23, can be when detecting the position reliably, even also can prevent performance degradation for changing through the time.In addition, detect line 111,112 and respectively be provided with 6, in Figure 13 shown in broken lines for the sake of brevity each 4.
So, carry out smoothing and rim detection for the pixel data that respectively detects on the line 111,112.In specific words, the data of 9 pixels of 8 pixels around the pixel of gazing at for comprising are arranged element accumulation with the array data of table 1 and each, obtain resulting 9 elements and absolute value.And, for the data of above-mentioned 9 pixels, the array data of table 2 and each are arranged element accumulation, obtain resulting 9 elements and absolute value.So, obtain these absolute values and as the data after the smoothing rim detection of the pixel of being gazed at.
(table 1) (table 2)
| ????1 | ????0 | ????-1 |
| ????1 | ????0 | ????-1 |
| ????1 | ????0 | ????-1 |
| ????1 | ????1 | ????1 |
| ????0 | ????0 | ????0 |
| ????-1 | ????-1 | ????-1 |
Afterwards, the data after the smoothing rim detection are divided into two in the central authorities that detect line 111,112, and preceding half-sum is later half respectively peaked as periphery candidate point 113 to No. three get.At this moment, on the proportional modified value of distance of interpolation and range line central authorities, can obtain until No. three maximal value.So the outside of the image that the possibility of the preferential periphery of selecting to take coin 23 is high is promptly away from the point of line central authorities.Thus, even take place irregularly on the illumination of Lighting Division 15, the periphery of coin 23 is taken when very dark or the like, also can detect the position of coin with good precision.
In Figure 13, the part of a plurality of periphery candidate points 113 is represented with white point, stain.On every line, in the past select 2 the mid point total of any to obtain 9 groups in half candidate point and the later half candidate point.6 lines are added up to the highest point of the frequency of 54 groups the mid point frequency in distributing as the Provisional Center coordinate, it is obtained as Provisional Center 114 for and vertical direction parallel with direction of passage.So the distance of calculating Provisional Center 114 and each periphery candidate point 113 (whole 72 points) is obtained the highest distance of frequency from the frequency distribution of these distances.In addition, in the periphery candidate point 113 of the distance of obtaining in certain permissible range, select each and have only any to belong to same 2 the later half group of preceding half-sum that detects line, as the periphery point of in Figure 13, representing 115 with stain.Such as, the distance of periphery candidate point 116 and Provisional Center 114 in permissible range, and because at the same periphery candidate point 113 that not have on the line 112 apart from permissible range that detects, not as the periphery point.
At last, obtain to be positioned at the parallel average coordinates of the direction of passage that detect a plurality of and periphery point 115 line 111 on parallel and to obtain and be positioned at the vertical average coordinates of the direction of passage that detect a plurality of and periphery point 115 line 112 on vertical as the center with direction of passage with direction of passage.And obtain the mean distance of this center and periphery point 115, with its 2 times of values as external diameter.
In the present embodiment, select a plurality of candidate points suitable to be decided to be interim center with periphery.Obtain the distance at each candidate point and interim center then.And by from being judged to be the position of obtaining the center among a plurality of candidate points suitable and detecting coin 23 reliably with periphery.Set a plurality of lines 111,112 that detect like this.And, carry out the smoothing rim detection, when revising under the preferential situation in the image outside, select a plurality of to the periphery candidate point 113 that respectively detects on the line 111,112.Smoothing and rim detection any one, when the center external diameter detected, not conference was subjected to the dirt of coin 23 and path 13 and the 15 irregular influences of throwing light on.In addition, not only in rim detection, in emphasizing, the edge also can obtain same effect.Carry out smoothing and rim detection or emphasize by the array data that utilizes table 1 and table 2, can detect the center external diameter at a high speed.In addition, be occasion beyond circular at identifying object, such as, for rectangular bank note etc., also can obtain long limit and minor face size together to center and periphery.Also be not restricted to these, the position that detects identifying object to detect as reference point as the shape in the hole of feature etc. and floral designs etc. also is fine.
In addition, utilize a plurality of technology that line detects periphery point that detect, such as, open in Japanese kokai publication hei 9-17821 number and Japanese kokai publication hei 10-63852 number.Some in the present embodiment, are the candidate points of selecting a plurality of peripheries for this, utilize and the distance of Provisional Center verifies that whether candidate point is periphery point and obtain the periphery point.Therefore, as shown in figure 13, even because when taking bottom surface 22, there is noise to exist, can wrong choice external diameter point yet, so correctly inspection center and external diameter.
Afterwards, carry out the value of money kind and judge (S43).In S43, judge that according to interim from the external diameter of center external diameter test section 85 outputs the coin 23 that drops into is approaching with any value of money.The interim detection unit 86 of value of money kind, according to predetermined various value of money kinds each the external diameter reference value and the shooting multiplying power of storage, two kinds of values of money that the external diameter of selecting and detecting is approaching.But, this value of money kind is a kind of or a kind of when following in the permissible range of storage in advance, just selects this value of money kind.The interim detection unit 86 of value of money kind can be easy to judge the value of money kind according to external diameter temporarily.
If when determining that the value of money kind is a kind of, then, easily the value of money kind is judged by accident because coin 23 is very near by still being from its far influence of passing through of passing through the position from image pickup part 16 in path 13.In the present embodiment, not to concentrate on a kind of value of money kind but two kinds, accuracy of identification is improved.Such as, the hundred yuan of coins (22.6mm) of Japan and the external diameter difference of ten yuan of coins (23.5mm) have only 4%.Particularly very big by the influence of position in the occasion of using wide-angle lens, such as under the situation of the variation of pass through the position of ± 0.5mm,, the danger of erroneous judgement is arranged owing to the external diameter on the image has ± 2% variation.In addition, if be judged to be 3 kinds of value of money kinds or more than 3 kinds temporarily, following data processing, the anglec of rotation of particularly following floral designs comparison operations (S26) detects (S50) needs a lot of times.Therefore, select two kinds of value of money kind outputs at most.In addition, the value of money kind output of selecting more than three kinds and three kinds is also passable.Like this, the upper limit is set, makes data processing time be no more than the certain upper limit time by number to the kind judged temporarily.
Be defined as the value of money kind a kind of and carry out technology corresponding to the processing of this value of money kind, such as, open in Japanese kokai publication hei 10-63852 number.To this, in the present embodiment,, can prevent because the erroneous judgement of the caused value of money kind of influence of passing through position etc. of coin 23 by interim judgement to a plurality of value of money kinds.
Afterwards, carry out concavo-convex calculating (S44).In S44, detect the concavo-convex degree that portion 83 detects coin 23 by concavo-convex according to first, second view data of input in concavo-convex input operation (S23).For this reason, to each concavo-convex line 102-104 that detects, the absolute value with the difference of each pixel of first view data and second view data carries out addition to these concavo-convex whole pixels that detect on the line.When each pixel being asked difference, in order to give up the influence that above-mentioned coin 23 moves, obtain data poor of pixel of the 1 pixel back (exporting 18 sides) of itself and second view data for the pixel of first view data of importing earlier.Like this, the concavo-convex portion 83 that detects is calculated 3 concavo-convex line 102-104 that detect, and the maximal value in 3 is exported.By setting the many concavo-convex line 102-104 that detect, not conference is subjected to the influence of the dirt of coin 23 and path 13, even ridge design pattern coin 23 seldom also can detect concavo-convex degree reliably.
Carry out floral designs comparison operation (S26) afterwards again.In S26, the view data of many-valued digital value is being carried out rim detection (S45) afterwards, in the correction of multiplying power of making a video recording (S46), carry out from rectangular coordinate to polar coordinate transform (S47).On the basis with its binaryzation (S49), detection is to the anglec of rotation (S50) of the view data of the coin 23 of the input of benchmark image.So, under the situation of the anglec of rotation of considering detection, carry out comparison (S51) with benchmark image.These operations will be carried out two sides in each table of judging interim two kinds of value of money kinds judging in (S43) in the value of money kind temporarily.Herein, such as, as Euro coin, be general at floral designs simultaneously, and each different occasion of 15 countries of another side can add up to and carry out 16 processing.
Each work to floral designs comparison operations (S26) is described in detail according to rim detection (S45) below.Rim detection portion 84 carries out detecting (S42) with the center external diameter to each pixel data of the 3rd view data and similarly is also used as the rim detection of smoothing and the view data after the rim detection is exported as the 4th view data.In specific words, the data of 9 pixels of 8 pixels around the pixel of gazing at for comprising, obtain with the array data of table 1 and 9 elements that each arranges element accumulation and absolute value.And obtain the array data of table 2 and each arrange element accumulation 9 elements and absolute value.So, obtain these absolute values and as the data after the rim detection of the pixel of being gazed at.In addition, rim detection (S45) if be benchmark with the center of detecting detection in (S42) at the center external diameter, is carried out being positioned at as the pixel of the external diameter scope of the maximum outside diameter coin of identifying object, and is just enough.By view data being carried out rim detection (S45), be not only catoptrical lightness (brightness) itself, and can utilize with the luminance difference of adjacent part and judge the edge.Therefore not conference is subjected to the overall variation of brightness of view data and local irregular influence and can correctly discerns.In addition, be not only rim detection, carrying out also can obtaining same effect when the edge is emphasized.The overall variation of the brightness of view data and local irregular being caused by the integral body of coin 23 materials and color, path 13 or coin 23 or a part of local dirt etc. are taken place.By carrying out above-mentioned such processing, can reduce these influence.
Below, before coordinate transform (S47) is illustrated in multiplying power correction (S46).Coordinate converting section 88 is polar coordinates and the view data after the coordinate transform is exported as the 5th view data from rectangular coordinates transformation the 4th view data.In the present embodiment, as identifying object, the optical design of 500 yen coin of maximum outside diameter is N * N pixel (such as, 100 * 100 pixels) with Japanese coin.In addition, because the deviation of the characteristics of components that passes through position and package size, lens 28 etc. of coin, the multiplying power of will making a video recording is predefined for 0.9 to 1.1.In other words, 500 yen coin when goods or coin drop into, all carries out from minimum 0.9N * 0.9N to maximum 1.1N * 1.1N pixel the shooting of (such as, from 90 * 90 to 110 * 110 pixels).Herein, for the sake of simplicity, explanation be the coordinate transform (S47) and the multiplying power correction (S46) of maximum 7 * 7 pixels.
The 4th view data is shown to Figure 14 A pattern.Direction that will be parallel with the direction of passage of coin 23 is as X coordinate 122, and direction that will be vertical with the direction of passage of coin 23 will detect the coordinate at the center of detecting in (S42) 121 as (0,0) at the center external diameter as Y coordinate 123.So, 49 pixels of expression from rectangular coordinate (3,3) to (3,3).The 5th view data after the coordinate transform is shown to Figure 14 B pattern.With polar radial direction as R coordinate 124, with circumferencial direction as S coordinate 125.Herein, R coordinate 124 carries out four with unequal interval and is divided into 0-3, and S coordinate 125 is divided into 0-7 uniformly-spaced to carry out eight.In other words, be to (3,7) from polar coordinates (0,0) shown in the example.This coordinate transform utilizes the conversion of table 3 and table 4 to carry out with table.The X coordinate and the corresponding polar S coordinate of Y coordinate of table 3 expression and rectangular coordinate, and the corresponding polar R coordinate of Y coordinate of table 4 expression and rectangular coordinate.
(table 3)
| Circumferencial direction S | The X coordinate | |||||||
| ????-3 | ????-2 | ????-1 | ????0 | ????1 | ????2 | ????3 | ||
| The Y coordinate | ????-3 | ????7 | ????7 | ????7 | ????0 | ????0 | ????0 | ????1 |
| ????-2 | ????6 | ????7 | ????7 | ????0 | ????0 | ????1 | ????1 | |
| ????-1 | ????6 | ????6 | ????7 | ????0 | ????1 | ????1 | ????1 | |
| ????0 | ????6 | ????6 | ????6 | ????0 | ????2 | ????2 | ????2 | |
| ????1 | ????5 | ????5 | ????5 | ????4 | ????3 | ????2 | ????2 | |
| ????2 | ????5 | ????5 | ????4 | ????4 | ????3 | ????3 | ????2 | |
| ????3 | ????5 | ????4 | ????4 | ????4 | ????3 | ????3 | ????3 | |
(table 4)
| Radial direction R | The X coordinate | |||||||
| ????-3 | ????-2 | ????-1 | ????0 | ????1 | ????2 | ????3 | ||
| The Y coordinate | ??-3 | ????4 | ????4 | ????3 | ????2 | ????3 | ????4 | ????4 |
| ??-2 | ????4 | ????2 | ????1 | ????1 | ????1 | ????2 | ????4 | |
| ??-1 | ????3 | ????1 | ????0 | ????0 | ????0 | ????1 | ????3 | |
| ??0 | ????2 | ????1 | ????0 | ????0 | ????0 | ????1 | ????2 | |
| ??1 | ????3 | ????1 | ????0 | ????0 | ????0 | ????1 | ????3 | |
| ??2 | ????4 | ????2 | ????1 | ????1 | ????1 | ????2 | ????4 | |
| ??3 | ????4 | ????4 | ????3 | ????2 | ????3 | ????4 | ????4 | |
Such as, the pixel of rectangular coordinate (0,0) is transformed to polar (0,0), and the pixel of (1 ,-3) of rectangular coordinate is transformed to polar coordinates (3,7).So just can carry out conversion, the suitable occasion of a plurality of rectangular coordinates be arranged, just adopt the mean value of these pixels at same polar coordinates to whole pixels of rectangular coordinate.In addition, the polar coordinates that all do not have of pixel of this rectangular coordinate polar data by duplicating adjacency or the like are set according to contiguous data.Be transformed to polar coordinates by coordinate transform, become and be easy to simultaneously making the following anglec of rotation detect (S50) and comparison (S51), can cut down data volume and make processing speed become high speed.In addition, in the radial direction table (table 4) that polar coordinate transform is used, the R coordinate of radial direction coordinate has the coordinate that becomes than peaked 3 bigger 4.This is illustrated in outside the external diameter scope of coin 23, be after obsolete pixel.In addition, the R coordinate of the arrangement by will being used for the view data after the storing coordinate conversion is prepared to reach 4, is then judging whether the R coordinate is not need to carry out conversion when (comprising 3) below 3 when coordinate transform.Thus, can carry out conversion at a high speed to the common processing of whole pixels.
Utilize polar coordinate transform, change with the distance of pixel (coordinate) around the adjacency.Therefore, if carry out rim detection (S45) afterwards in coordinate transform (S47), then pixel of being gazed at and rim detection or emphasize the distance of the adjacent pixels around employed are according to the coordinate of the pixel of being gazed at or the coordinate of adjacent pixels, different rather than certain.To this, in the present embodiment, carry out coordinate transform (S47) afterwards in rim detection (S45).Thus, the floral designs of faithful to coin 23 detect the edge, and accuracy of identification improves.
In multiplying power correction (S46), the correction of the shooting multiplying power that carry out because the assembly error that passes through position and image pickup part 16 in the path 13 of coin 23 etc. causes.This correction is that basis is at the center external diameter
Detect the external diameter and the value of money kind that detect in (S42) and judge the interim value of money kind of judging in (S43) temporarily, undertaken by the value of upgrading the radial direction table (table 4) that polar coordinate transform uses.
The radial direction table is tried to achieve the upper distance limit table of the upper limit of the distance of each radial direction coordinate by the distance table and the such expression of table 6 of the distance at such each pixel of expression of table 5 and center.
(table 5) (table 6)
| Distance R ' | The X coordinate | ||||
| ????0 | ????1 | ????2 | ????3 | ||
| The Y coordinate | ????0 | ????0.00 | ????1.00 | ????2.00 | ????3.00 |
| ????1 | ????1.00 | ????1.41 | ????2.24 | ????3.16 | |
| ????2 | ????2.00 | ????2.24 | ????2.83 | ????3.61 | |
| ????3 | ????3.00 | ????3.16 | ????3.61 | ????4.24 | |
| Radial direction R | Upper distance limit |
| ????0 | ??1.75 |
| ????1 | ??2.47 |
| ????2 | ??3.03 |
| ????3 | ??3.50 |
| ????4 | ??- |
Such as, establishing the X coordinate is x, the Y coordinate is y, and establish apart from tabular value be R ' (x, y), then because R ' (0,0) be 0.00, less than 1.75, just radial direction tabular value R (0,0) is decided to be 0.Because R ' (3,1) is 3.16,, just R (3,1) is decided to be 3 greater than 3.03 less than 3.50.In addition, the distance table of the part of X coordinate＜0 or Y coordinate＜0, the symmetry of utilization table and being omitted.So, in x＜0 and the occasion of y 〉=0 utilize R (x, y)=R (x, y), in x 〉=0 and the occasion of y＜0 utilize R (x, y)=R (x ,-y), and in x＜0 and the occasion of y＜0 utilize R (x, y)=(x ,-y) calculating formula is obtained radial direction coordinate R to R.By such formation map table, can cut down data storage capacity and processing time.
Such as, judge temporarily that in the value of money kind one of interim value of money kind of judging is 500 yuan of yen in (S43), detecting the external diameter (such as 90) that detects in (S42) at the center external diameter is 0.9 times occasion of 500 yuan nominal outer diameter (such as 100), and the shooting multiplying power is 0.9 times.In multiplying power correction (S46), as shown in table 7 based on table 6, the value of the upper distance limit of upper distance limit table is updated to 0.9 times of standard.
(table 7)
| Radial direction R | Upper distance limit R ' max |
| ????0 | ????1.58 |
| ????1 | ????2.23 |
| ????2 | ????2.73 |
| ????3 | ????3.15 |
| ????4 |
So the value of upgrading the radial direction table that polar coordinate transform uses according to this upper distance limit table (table 7) and distance table (table 5) obtains table 8.
(table 8)
| Radial direction R | The X coordinate | |||||||
| ??-3 | ??-2 | ??-1 | ??0 | ??1 | ??2 | ??3 | ||
| The Y coordinate | ??-3 | ??4 | ??4 | ??4 | ??3 | ??4 | ??4 | ??4 |
| ??-2 | ??4 | ??3 | ??2 | ??1 | ??2 | ??3 | ??4 | |
| ??-1 | ??4 | ??2 | ??0 | ??0 | ??0 | ??2 | ??4 | |
| ??0 | ??3 | ??1 | ??0 | ??0 | ??0 | ??1 | ??3 | |
| ??1 | ??4 | ??2 | ??0 | ??0 | ??0 | ??2 | ??4 | |
| ??2 | ??4 | ??3 | ??2 | ??1 | ??2 | ??3 | ??4 | |
| ??3 | ??4 | ??4 | ??4 | ??3 | ??4 | ??4 | ??4 | |
Carry out coordinate transform according to conversion like this with table, just view data can enlarged on radial direction in 1.11 times, realize coordinate transform, the multiplying power correction of can making a video recording at an easy rate.By carrying out the multiplying power correction, can not be subjected to coin 23 the deviation of passing through position and package size characteristics of components influence and can correctly discern.In addition, can store all different benchmark image of each goods, commonization of benchmark image be suppressed the deviation of the recognition performance of each goods, can enlarge the permissible range of assembling and characteristics of components simultaneously.
In the present embodiment, the value of the upper distance limit of upper distance limit table, radius is closely-spaced more wide more, and radius large-spacing is narrow more more, setting be unequal interval.But, also can be as shown in table 9 be set at uniformly-spaced.
(table 9)
| Radial direction R | Upper distance limit R ' max |
| ????0 | ????0.88 |
| ????1 | ????1.75 |
| ????2 | ????2.63 |
| ????3 | ????3.50 |
| ????4 | ????- |
Like this, on each polar area, can produce very big difference, the pixel count that is transformed to same polar rectangular coordinate when coordinate transform becomes inhomogeneous, when comparison, each polar weighting meeting generation deviation, so, make each polar coordinates have equal area if upper distance limit set, each polar weighting will equate.In other words, by being set with unequal interval, the radial direction coordinate can improve recognition performance.Particularly, the coordinate of radial direction is more little, cuts apart at interval widely more, and by carrying out polar coordinate transform, the difference of area with before the coordinate transform that the coordinate behind the polar coordinate transform can be produced diminishes.
In addition, obtain the image of the object that becomes benchmark, obtain that magnification error is revised and the technology of carrying out image interception is open in Japanese kokai publication hei 9-245213.In this communique, as benchmark to as if fixing value of money kind, under adjustment modes,, use very difficult but drop into multiple value of money kind time-like in that only to drop into this value of money kind time-like no problem.To this, in the present embodiment,, also can be applicable to the occasion that a plurality of value of money kinds drop into by a plurality of value of money kinds are judged temporarily.
Then, the threshold setting (S48) that is used for binaryzation (S49).Whether binaryzation portion 90 is for each pixel (coordinate) as the 5th view data after the coordinate transform of many-valued digital value, according to being to be transformed to 1 or 0 bi-level digital value more than the threshold value.Then the view data after the binaryzation is exported as the 6th view data.By carrying out binaryzation (S49), in view data, when reducing The noise, can cut down image data amount and carry out high speed processing.In addition, by carrying out rim detection (S45) before in binaryzation (S49) or the edge is emphasized, coordinate transform (S47), utilize the correct rim detection of multi-value data or emphasize with coordinate transform be possible, accuracy of identification can improve.
For obtaining threshold value, such as, the frequency that makes data value and pixel count from the 5th view data distribute (histogram).Afterwards, obtain from the big side's of data cumulative frequency and become 50% value as threshold value.In addition, the 5th view data that threshold setting uses is defined in the inside that coin partly is an external diameter certainly.Therefore, by carrying out multiplying power correction and polar coordinate transform, the radial direction coordinate can only use the following part of radius of predetermined this value of money kind, in addition, in the present embodiment, will owing to the state of coin 23 and issue annual difference make the edge portion of the different outer perimeter holes of each coin 23, not be the inside in hole of coin 23 but background from the use of threshold setting except.In addition, even same coin 23 because the difference of the anglec of rotation and position, drop into each time have different latent image fine rule floral designs part also from the use of threshold setting except.Figure 15 for the benchmark image (following) of new 500 Japanese yen coins is gone up with oblique line except the mode chart of the rectangular coordinate inverse transformation that marks of part, the edge portion 131 of periphery shown in it and latent image floral designs portion 132, fine rule floral designs portion 133.
The edge portion of outer perimeter holes is easy to take place wearing and tearing and changes.The inside in hole can obtain background.Speckle (Moire fringe) appears in the angle that latent image fine rule floral designs portion understands owing to the floral designs of the line of pixels column direction of sensor 26 and coin, and reflectance varies is remarkable.So if use these parts to carry out threshold setting, the influence that is subjected to the image of these parts can cause the obstacle that carries out correct binaryzation to integral body.Because same reason, for the different title of an emperor's reign part of each coin floral designs preferably also except.
Which in the 5th view data, can use partly carry out threshold setting, can make the table of the bi-level digital value of representing with polar form each of each value of money kind face.So, by the storage in advance of this table just can be carried out suitable stable binaryzation to each floral designs of each value of money kind face.
In addition, be the rotational symmetric occasion of center right and wrong of relative coin in the part that is used for this threshold setting, each anglec of rotation threshold value of coin is all changed.Therefore, detect in (S50), make the view data rotation all need to attach most importance at every turn and ask threshold value to carry out a large amount of processing in the following anglec of rotation.In the present embodiment, the part of using in the threshold setting is set at do not comprise except part and relatively the center of coin be that the rotation symmetry is a concentric circles.Therefore, detect in (S50), make the view data rotation not need heavily to ask threshold value at every turn in the anglec of rotation.In addition, the storage table that is used for the part of threshold setting also is to use one-dimensional data (below be called the threshold setting area data) to get final product for the coordinate of radial direction.Like this, in the 6th view data, the setting of threshold value is will the value of making be 1 pixel count becomes confirmable ratio to the ratio of whole pixels, in other words, to the setting of threshold value be frequency after the binaryzation is distributed become certain.Thus, can not be subjected to about the deviation of state, the anglec of rotation, position or Lighting Division 15 and package size of material, edge portion, surface etc. of coin 23 and illumination difference and carry out stable identification.
In addition, calculate the technology of setting the threshold value of binaryzation in the zone in specific threshold value, such as, open in TOHKEMY 2002-109596 number.But, in this communique, openly do not remove the technology of the specific part in the coin of the obstacle that causes binaryzation.In other words, even openly will and not issue edge portion or the same coin 23 that annual difference makes the different outer perimeter holes of each coin 23 owing to the state of coin 23, because the difference of the anglec of rotation and position, drop into the technology except the part with different latent image fine rule floral designs each time.In the present embodiment, by setting threshold in the zone except these zones, can carry out stable threshold setting and binaryzation.
In the anglec of rotation detected (S50), anglec of rotation test section 91 detected the anglec of rotation of the coin 23 that drops into and with its output.In other words, anglec of rotation test section 91 according to each value of money kind face being stored in benchmark image in the storage part 92 and the 6th view data of coin 23, detects the anglec of rotation of the 6th view data relative datum image.Otherwise, the anglec of rotation that also can relative the 6th view data of detection reference image.In addition, the benchmark image data, the same with the 6th view data, be to store in advance as the table of the bi-level digital value of representing with polar form.For the anglec of rotation detects, on one side the rotation of the 6th view data is compared with the benchmark image data on one side.So, detect the most consistent angle just as the anglec of rotation.
The 6th view data is with polar coordinate representation.So, to misplace in a circumferential direction for making the rotation of the 6th view data, promptly change the circumferencial direction coordinate.Utilize this point can make it be easy to rotation.With B (r, s) (r is polar radial direction, and s is a circumferencial direction) expression the 6th view data, and with R (r, s) expression benchmark image data.(number of cutting apart of polar circumferencial direction represents with Ns for A, B) expression, and then Δ s is each to be changed 2 (r, (r s) compares for moD (s+ Δ s, Ns)) and R to B when 0 changes to Ns-1 making if the remainder that removes A with B is with moD.Consistent degree is judged with the 6th view data pixel consistent with the value of benchmark image data (coordinate) number.Because the 6th view data and benchmark image data all are the bi-level digital values, get these XOR value, utilizing the result is that 0 pixel count is easy to judge consistent degree.Δ s is changed with each 2, is in order to shorten event detection time, to make its each change 2 can obtain the highest Δ s of consistent degree.So, also can obtain the anglec of rotation to Δ s adjacency, promptly moD (Δ s-1, Ns) and moD (Δ s+1, the Ns) consistent degree in are selected the highest anglec of rotation of consistent degree in 3 anglecs of rotation.Thus, even sampling also can precision detect the anglec of rotation well between the anglec of rotation.In addition, also be not limited to each method of 2 that changes, if consistent degree is low, variable quantity can be big, and if the consistent degree height, variable quantity can be little.Calculate consistent degree by the anglec of rotation being carried out an extraction sample, can be when keeping precision the high speed detection anglec of rotation.
In addition, in the present embodiment, the part of the edge portion of the outer perimeter holes of the coin 23 in threshold setting (S48) except from use, the inside in hole, latent image fine rule floral designs is calculated in (anglec of rotation detects and the floral designs comparison) also at consistent degree except.Its reason identical during with threshold setting.In specific words, will represent that the anglec of rotation detects the data of the pixel (zone) of use in (S50) and the floral designs comparisons (S51) (below be called the comparison area numeric field data) and stores in advance.Herein, the comparison area numeric field data, the same with the benchmark image data with the 6th view data, be to store as the table of the bi-level digital value of representing with polar form.So, get the XOR value of the 6th view data and benchmark image data and comparison area numeric field data with value, the result is that 1 inconsistent pixel count is few more, just judges that consistent degree is high more.
In addition, in the present embodiment, be value with the comparison area numeric field data be 1 pixel count as the comparison pixels number, deduct inconsistent pixel count from the comparison pixels number, its difference is removed as consistent degree with the comparison pixels number.Thus, the consistent degree of expression is not relative, but absolute, and the comparison pixels number does not compare consistent degree simultaneously easily yet.
In floral designs comparison (S51), the 6th view data and the reference data of utilizing the anglec of rotation that anglec of rotation test section 91 detects are compared.At first, the size of the anglec of rotation that the 6th view data is staggered in a circumferential direction detect with anglec of rotation test section 91 promptly changes the circumferencial direction coordinate and is rotated the correction of angle and makes it as the 7th view data.Afterwards, get the distance value of the 7th view data and benchmark image data, and get with anglec of rotation comparison in the comparison area numeric field data used " with " value is as the comparison result data.The value of the 7th view data and the inconsistent pixel of benchmark image data (coordinate) becomes distance value 1, and if in the comparison zone, with the comparison area numeric field data " with " value becomes 1.Therefore, the result is that 1 pixel count is represented inconsistent comparison pixels number.In addition, because quotient representation the 7th view data that the difference that the value of comparison area numeric field data is 1 pixel count represents comparison pixels number and the inconsistent pixel count of comparison pixels number is removed with the comparison pixels number and the consistent degree of benchmark image.
In the present embodiment, not the consistent degree of obtaining whole coin, but the coin zone be divided into a plurality of zones that the number of obtaining the zone of consistent degree more than certain value in each zone is consistent number of regions.By consistent degree and certain value are compared, be easy to obtain consistent number of regions herein.In addition, can reduce the influence of the dirt of coin 23 and path 13.
Utilize below that Figure 16 A, B specify that the radial direction pixel count is 32, the circumferencial direction pixel count is 16 example.Figure 16 B is for carrying out the mode chart of Region Segmentation to the view data shown in Figure 16 A.On polar radial direction 141, view data is cut apart (cutting apart number in this example is 4) with the approximate number of radial direction pixel count.On circumferencial direction 142, equally also cut apart (cutting apart number in this example is 8) with the approximate number of circumferencial direction pixel count.So, can obtain a plurality of (being 32 in this example) zone 143.Utilize polar coordinates, can be when easily view data being carried out Region Segmentation, if uniformly-spaced cut apart, the weighting in each zone 143 can be impartial.
Afterwards, to each zone, utilize the pixel that belongs in this zone to obtain consistent degree.In addition, will be to the ratio of number of regions more than the predetermined lower limit of each regional consistent degree and Zone Full number as consistent regional percentage, and by floral designs comparison portion 93 with this value output.In addition, narrow in comparison zone, promptly in the zone that the comparison pixels number waits less, be used to calculate the quantity of information deficiency of the consistent degree in each zone.Such zone is difficult to the correct consistent degree of judging.Therefore, consistent zone can correctly be judged not as object in the zone by the comparison pixels number is not reached certain value.
Like this, not the consistent degree of obtaining the floral designs of whole coin, but the coin zone is divided into a plurality of regional 143, and compare according to each regional consistent degree.Thus, even a part of part of coin 23 and path 13 exists dirt also not too can be affected and can correctly discern.In addition, as long as change the permissible range of the consistent regional percentage of the lower limit of each the regional consistent degree be judged to be consistent zone and comparing section 97, just can be easy to change degree of priority to the tolerance and the eliminating counterfeit money performance of dirt.Such as, make and can not correctly read floral designs even 1/3 part of coin is attached with dirt etc., for make identification may, the permissible range that can make consistent regional percentage is more than 2/3.
When on coin 23 and path 13 dirt being arranged, consistent degree reduces, and consistent number of regions reduces, and can not judge that the possibility of genuine note increases.This is because all almost same low event of any one anglec of rotation that the consistent degree in entire comparison zone detects in the anglec of rotation.In the present embodiment, be that the consistent degree that each is regional is that number of regions more than the predetermined lower limit is as consistent number of regions.In addition, when each value of money kind face is determined this lower limit, be not fixed value but variable value if make it, then can reduce coin 23 and path 13 dirt influence and may carry out correct identification.Such as, use the consistent degree in entire comparison zone to become consistent degree under the minimum anglec of rotation, utilize the difference of maximum consistent degree and minimum consistent degree.Add on the consistent degree of minimum that perhaps certain value as lower limit or the like, can improve recognition performance.In this example, significantly to making correct identification become very at need, maximum consistent degree becomes low-down value, and the danger that produces mistake identification is arranged at dirt.Yet,, can pre-determine the lower limit that can not judge genuine note if the absolute value of maximum consistent degree is low.If perhaps adding lower limit that certain value is calculated on the consistent degree of minimum is predetermined certain value when following, just with this lower limit as this certain value.So just can avoid mistake identification.
Floral designs comparisons (S51) are at first carried out according to this reference data and comparison area numeric field data on the surface that temporarily is judged to be the first approximate value of money kind in (S43) judge in the value of money kind temporarily.The threshold setting area data because each face is all different, repeats (S52) from threshold setting (S48) until floral designs are compared (S51) to the inside too.In addition, the multiplying power correction because each value of money kind is all different, from multiplying power correction (S46) until S52, to judge temporarily that in the value of money kind being judged to be the second approximate value of money kind in (S43) carries out (S53) temporarily.So floral designs comparison portion 93 exports the consistent regional percentage of 2 value of money kind * 2 amount.
In relatively (S61), detect with being input to thickness among each comparing section 94-97 respectively that portion 74, material detect portion 78, concavo-convex each output that detects portion 83 and floral designs comparison portion 93 compares with the benchmark that is stored in the storage part 92.Comparing section 94-97, in permissible range when consistent, the signal of the kind of this genuine note of output expression then, and with all inconsistent occasion of the reference value of any kind of, then the output expression is the signal of counterfeit money.
Judging in (S62), detection unit 99 is so long as be the occasion of the kind of same genuine note, the just signal of the kind of this genuine note of output expression all from the signal indication of comparing section 94-97 of input, and the occasion beyond it, then the signal of output expression counterfeit money is as decision signal 100.
Utilizing thickness transducer input to detect the occasion of ("Yes" in S16) and utilizing the material sensor to drop into to detect the occasion of ("Yes" in S17), be the occasion during dewfall detects.In this occasion, in comparing (S71), the thickness that is input to each comparing section 94,95 respectively detects portion 74, material detects each output of portion 78 and the benchmark that is stored in the storage part 92 compares.In other words, can ignore the output that be associated with image pickup part 16 concavo-convex detects the comparing section 96,97 that portion 83 and floral designs comparison portion 93 be connected thereto.Comparing section 94,95, in permissible range when consistent, the signal of the kind of this genuine note of output expression then, and with all inconsistent occasion of the reference value of any kind of, then the output expression is the signal of counterfeit money.
Judging the output of ignoring comparing section 96,97 in (S72) too.Detection unit 99 is the occasion of the kind of same genuine note at the signal indication of comparing section 94,95 from input all, the signal of the kind of this genuine note of output expression just, and the occasion beyond it, and then the signal of output expression counterfeit money is as decision signal 100.
Detection unit 99, in order to judge whether to carry out any one judgement among S62 and the S72, detection unit 99 is connected with the output that dewfall detects portion 82.In addition, on detection unit 99 also input the output of the switching part 98 of switch etc. is arranged.Utilize these some, detection unit 99 only carries out the identification project (thickness and material) the identification project (concavo-convex and floral designs) that the judgement in dewfall is detected is associated with image pickup part 16.Therefore, in dewfall, also can preferentially accept coin and prevent to lose as the vending machine of automatic vending machine etc.Or, in dewfall, because identification difficulty in image pickup part 16, do not force the coin accepting to drop into, the counterfeit money discharge function is preferential.Though not shown in Fig. 9, utilize switching part 98 can carry out this selection.Like this owing to have switching part 98, can select coin accept preferential or the counterfeit money discharge function preferential, and also can change selection at an easy rate at the scene that is provided with of automatic vending machine etc.In addition, have and adopt the dewfall of heat generating part to prevent portion 101, prevent portion's 101 work, just can when saving power consumption, can carry out the identification that is associated with image pickup part 16 forever by when dewfall detects, making dewfall.As heat generating part, needn't use special-purpose parts, but the big electronic unit of dual-purpose thermal value in this purpose it is disposed at path 13 near, perhaps can adopt cheap well heater.
Sometimes with image pickup part 16 uncorrelated identification projects (being thickness and material in the present embodiment) in be the genuine note of same kind, and in the identification project that is associated with image pickup part 16 (concavo-convex and floral designs), be counterfeit money or different types of genuine note then.In this occasion, not shown among Fig. 9, judging in (S62), be judged as and be attached with dirt or the like in the path 13 and make image quality low.And, when the occasion of following is all discerned project and all is the genuine note of same kind, can judge that image quality is no problem, can detect the low degree of image quality.Detecting the low occasion of image quality, can communicate by letter with the control part of automatic vending machine, the image quality that shows to the operator of automatic vending machine report is low, impels it to clean.In other words, these formation has the function of the portion that detects as image quality.
In addition, by the second switching part (not shown) of separating with switching part 98, also can select still not carry out the identification (below be called image recognition) itself in image pickup part 16.By second switching part is set, at image pickup part 16 and optical transparent body 14 breaks down and the such occasion of great damage dirt, can select not carry out image recognition.So, not shown in Fig. 9, do not carry out dewfall and detect that (S14) and input in image pickup part 16 detect (S15) and concavo-convex and floral designs detect (S21).Thus, can prevent because the action of the goods that misoperation produces stops and recognition performance low.
In addition, the inspection method of coin input, utilize access structure, the dewfall of Fig. 2 and Fig. 4 explanation to detect and dewfall prevents and image quality detects, also can be applied in the floral designs recognition device of existing structure.Can drop into the correct identification that detects, is not subjected to the reliable identification that dirt influences and reduce the influence of dewfall and dirt respectively reliably.
As mentioned above, according to present embodiment, utilize following formation work can obtain following effect respectively.Discern from the non-incident portion of the normal reflection light 29 of coin 23 by being used to, can prevent from image, to produce irregular.In addition, image pickup part 16 is set, can reduces dirt adhering to image pickup part 16 by vertical upper side at relative path 13.Because image pickup part 16 is made up of the sensor 26 of image that can output, can read the view data of specific region 66 at short notice.
Image pickup part 16, by the array with the photo detector that is configured as the ranks shape, the voltage that will be used for sensitivity control is supplied with the control circuit 68 of the row of this array, and processing is formed from the sensor 26 of the neural network 69 of the ground current of the row of array.Therefore, in sensor 26, can carry out the accumulation computing of wave filter ranks and image.In addition, by possess near the path that detects the image pickup part 16 13 whether the dewfall of dewfall detect portion 82, can detect whether dewfall of path 13.So,, can prevent because the misoperation that abnormal view data causes makes action of the dewfall portion of preventing and report that dewfall take place owing to the control and treatment that saves image pickup part 16 is just not all right.By utilizing image pickup part 16 to detect the input of coin 23, can detect the input of coin 23 reliably.By the image that reads according to image pickup part 16 in advance reading conditions is adjusted, can be obtained the view data of certain image quality.Be provided with and have almost same wavelength characteristic carries out illumination from different directions to coin 23 Lighting Division 36 and Lighting Division 37.So,, can detect the concavo-convex of coin 23 by according to coming leisure to utilize the reflected light of the coin 23 when Lighting Division 36 shines and coming leisure to utilize the catoptrical difference of the coin 23 when Lighting Division 37 shines that coin 23 is discerned.By verifying that a plurality of periphery candidate points 116 detect the point (center 114) of benchmark of the position of the coin 23 that becomes the 3rd image, can detect the position of coin 23 reliably.Discern by a kind of and more than one value of money kind are judged temporarily, can prevent from the value of money kind to be judged by accident owing to be subjected to the influence of passing through the position of coin 23.By image pickup part 16 being arranged at the vertical upper side of path 13, can reduce dirt adhering to image pickup part 16 grades.Emphasize that by detecting in the 3rd edge of image laggard row-coordinate conversion discerns, floral designs that can faithful to coin 23 detect the edge.By according to the image setting threshold value of the specific part of removing coin 23 and with its binaryzation, can carry out stable binaryzation.Carry out comparison with benchmark image by image, can carry out stable comparison according to the specific part that removes cloud coin 23.Detect an anglec of rotation by the anglec of rotation being carried out an extraction sample, can keep the precision former state constant in the high-speed anglec of rotation that detects.Because with the image segmentation of coin 23 is a plurality of regional 143, and discern, so can reduce the influence of the dirt of coin 23 and path 13 according to each consistent degree of benchmark image and each zone.
Utilize these characteristics, present embodiment can correctly be discerned the floral designs of coin.Therefore, can obtain to prevent the high floral designs recognition device of recognition performance of the illegal use of counterfeit money.
In addition, the floral designs recognition device of present embodiment be mint-mark floral designs with coin as identifying object, but the present invention also can be applied to the mint-mark floral designs etc. that identifying object is printing floral designs, metal and the mark etc. of bank note and card etc.And floral designs not only can be in surface, the inside of coin etc., also can be applied to the floral designs of side.In addition, the object of identification is not limited to utilize as present embodiment during gravity rotates, also can in falling naturally, utilize in the way that trucking department such as travelling belt moves and utilization stops portion etc. and makes identifying object stop at it to discern in stopping.
In addition, in embodiments of the present invention, image pickup part and Lighting Division are to be disposed at the same side for path, are used to that the reflected light of self-identifying object discerns, but image pickup part and Lighting Division also can be disposed at contrary side for path, and the transmitted light that is used to the self-identifying object is discerned.
Claims (26)
1. floral designs recognition device comprises:
Lighting Division with the rayed identifying object reaches
Acceptance is by any image pickup part in the light of above-mentioned identifying object reflection or transmission;
When the image of obtaining according to above-mentioned image pickup part is discerned the floral designs of above-mentioned identifying object, judge that temporarily above-mentioned identifying object is one or more.
2. floral designs recognition device as claimed in claim 1 is wherein revised the shooting multiplying power corresponding to the kind of the above-mentioned identifying object of interim judgement.
3. floral designs recognition device as claimed in claim 2 is wherein revised the shooting multiplying power by the table that the coordinate transform of rewriting the image of being obtained by above-mentioned image pickup part is used.
4. floral designs recognition device as claimed in claim 3, the table that wherein above-mentioned coordinate transform is used utilize symmetry to constitute.
5. floral designs recognition device as claimed in claim 1, wherein the profile of above-mentioned identifying object is circular, judges according to the external diameter of above-mentioned identifying object temporarily.
6. floral designs recognition device as claimed in claim 1, wherein the number to the kind judged temporarily is provided with the upper limit.
7. floral designs recognition device as claimed in claim 6 is limited to two kinds on wherein above-mentioned.
8. floral designs recognition device as claimed in claim 1, wherein the profile of above-mentioned identifying object is circular, the image that will be obtained by above-mentioned image pickup part is cut apart on radial direction with unequal interval and is carried out polar coordinates and change and discern.
9. floral designs recognition device as claimed in claim 8, wherein the more for a short time interval of cutting apart of the coordinate of radial direction is wide more and carry out polar coordinate transform.
10. floral designs recognition device as claimed in claim 9, each coordinate time that wherein carries out after polar coordinate transform makes the preceding area of coordinate transform and utilizes polar coordinate transform equates.
11. floral designs recognition device as claimed in claim 1, wherein the point to the benchmark of the position of the above-mentioned identifying object that becomes the image of being obtained by above-mentioned image pickup part detects by verifying a plurality of candidate points.
12. floral designs recognition device as claimed in claim 11, the center that the above-mentioned point that wherein becomes benchmark is above-mentioned identifying object.
13. floral designs recognition device as claimed in claim 11, wherein a plurality of above-mentioned candidate points are the candidates that become the above-mentioned point of benchmark.
14. floral designs recognition device as claimed in claim 11, the candidate point of the periphery that wherein above-mentioned candidate point is above-mentioned identifying object.
15. floral designs recognition device as claimed in claim 14 wherein utilizes the candidate point of verifying above-mentioned periphery with the distance of the above-mentioned point that becomes reference point.
16. floral designs recognition device as claimed in claim 11 is wherein preferentially selected above-mentioned candidate point with the outside of the image of above-mentioned identifying object.
17. floral designs recognition device as claimed in claim 16 is wherein by selecting above-mentioned candidate point to the image correction of above-mentioned identifying object.
18. floral designs recognition device as claimed in claim 11 wherein carries out the selection of above-mentioned candidate point with many straight lines.
19. floral designs recognition device as claimed in claim 11 wherein carries out smoothing by the image to above-mentioned identifying object and selects above-mentioned candidate point.
20. floral designs recognition device as claimed in claim 11 wherein carries out any in emphasizing of rim detection and edge by the image to above-mentioned identifying object and selects above-mentioned candidate point.
21. floral designs recognition device as claimed in claim 20 wherein carries out any in emphasizing of smoothing and rim detection or edge simultaneously to the image of above-mentioned identifying object.
22. floral designs recognition device as claimed in claim 11 wherein becomes the detecting of above-mentioned point of benchmark in the anticipation position of above-mentioned identifying object.
23. floral designs recognition device as claimed in claim 22 wherein becomes the detecting of above-mentioned point of benchmark in the anticipation position based on the above-mentioned identifying object of the position of being stored.
24. floral designs recognition device as claimed in claim 1 wherein detects and the edge is emphasized carries out coordinate transform after any and discern in the edge of image of carrying out above-mentioned identifying object.
25. floral designs recognition device as claimed in claim 1 wherein detects and the edge is emphasized carries out binaryzation after any and discern in the edge of image of carrying out above-mentioned identifying object.
26. floral designs recognition device as claimed in claim 1 wherein carries out binaryzation and discerns after the coordinate transform of the image that carries out above-mentioned identifying object.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2002322255A JP3945379B2 (en) | 2002-11-06 | 2002-11-06 | Pattern recognition device |
| JP322255/2002 | 2002-11-06 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1499435A true CN1499435A (en) | 2004-05-26 |
| CN100530232C CN100530232C (en) | 2009-08-19 |
Family
ID=32802495
Family Applications (4)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2003101038243A Expired - Fee Related CN100530233C (en) | 2002-11-06 | 2003-11-06 | Figured pattern identifier |
| CNA2003101038192A Pending CN1499436A (en) | 2002-11-06 | 2003-11-06 | Figured pattern identifier |
| CNB2003101038262A Expired - Fee Related CN100552711C (en) | 2002-11-06 | 2003-11-06 | The floral designs recognition device |
| CNB2003101038188A Expired - Fee Related CN100530232C (en) | 2002-11-06 | 2003-11-06 | Figured pattern identifier |
Family Applications Before (3)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2003101038243A Expired - Fee Related CN100530233C (en) | 2002-11-06 | 2003-11-06 | Figured pattern identifier |
| CNA2003101038192A Pending CN1499436A (en) | 2002-11-06 | 2003-11-06 | Figured pattern identifier |
| CNB2003101038262A Expired - Fee Related CN100552711C (en) | 2002-11-06 | 2003-11-06 | The floral designs recognition device |
Country Status (2)
| Country | Link |
|---|---|
| JP (1) | JP3945379B2 (en) |
| CN (4) | CN100530233C (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101819692A (en) * | 2010-04-12 | 2010-09-01 | 高新现代智能***股份有限公司 | Coin image identification method and device |
| CN101789146B (en) * | 2009-01-23 | 2012-09-05 | 北京凌云光视数字图像技术有限公司 | Face-to-face photographing type coin inspection machine and coin inspection method |
| CN102663849A (en) * | 2012-05-09 | 2012-09-12 | 苏州爱锐邦智能科技有限公司 | Coin selector with appearance anti-fake function |
| CN103635941A (en) * | 2011-04-29 | 2014-03-12 | 加拿大皇家铸币厂 | Method and apparatus for authentication of a coin or other manufactured item |
| CN106683259A (en) * | 2017-03-06 | 2017-05-17 | 深圳怡化电脑股份有限公司 | Method and device for identifying paper money |
| US9922486B2 (en) | 2013-01-24 | 2018-03-20 | Arjo Solutions | Unique identification of coin or other object |
| CN111599080A (en) * | 2019-02-20 | 2020-08-28 | 深圳怡化电脑股份有限公司 | Spliced paper money detection method and device, financial machine tool equipment and storage medium |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4620435B2 (en) * | 2004-11-26 | 2011-01-26 | 富士電機ホールディングス株式会社 | Pattern matching method |
| JP6214036B2 (en) * | 2013-09-20 | 2017-10-18 | グローリー株式会社 | Coin identification device and coin identification method |
| JP6357897B2 (en) * | 2014-06-13 | 2018-07-18 | オムロン株式会社 | Medal identification device and medal identification method |
| JP6751686B2 (en) * | 2017-03-31 | 2020-09-09 | ローレル精機株式会社 | Coin identification device, coin processing device and coin identification method |
| CN108198226B (en) * | 2018-01-10 | 2022-04-26 | 广东辰宜信息科技有限公司 | Ceramic color identification method, electronic equipment, storage medium and device |
| TWI673486B (en) * | 2018-02-13 | 2019-10-01 | 特銓股份有限公司 | Optical detecting device and its light source edge finding mechanism |
| CN110164788A (en) * | 2018-02-14 | 2019-08-23 | 特铨股份有限公司 | Optical detection apparatus and its light source Xun Bian mechanism |
| CN113591467B (en) * | 2021-08-06 | 2023-11-03 | 北京金堤征信服务有限公司 | Event main body recognition method and device, electronic equipment and medium |
-
2002
- 2002-11-06 JP JP2002322255A patent/JP3945379B2/en not_active Expired - Fee Related
-
2003
- 2003-11-06 CN CNB2003101038243A patent/CN100530233C/en not_active Expired - Fee Related
- 2003-11-06 CN CNA2003101038192A patent/CN1499436A/en active Pending
- 2003-11-06 CN CNB2003101038262A patent/CN100552711C/en not_active Expired - Fee Related
- 2003-11-06 CN CNB2003101038188A patent/CN100530232C/en not_active Expired - Fee Related
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101789146B (en) * | 2009-01-23 | 2012-09-05 | 北京凌云光视数字图像技术有限公司 | Face-to-face photographing type coin inspection machine and coin inspection method |
| CN101819692A (en) * | 2010-04-12 | 2010-09-01 | 高新现代智能***股份有限公司 | Coin image identification method and device |
| CN101819692B (en) * | 2010-04-12 | 2012-08-15 | 高新现代智能***股份有限公司 | Coin image identification method and device |
| CN103635941A (en) * | 2011-04-29 | 2014-03-12 | 加拿大皇家铸币厂 | Method and apparatus for authentication of a coin or other manufactured item |
| US9558544B2 (en) | 2011-04-29 | 2017-01-31 | Monnaie Royale Canadienne/Royal Canadian Mint | Method and apparatus for authentication of a coin or other manufactured item |
| CN102663849A (en) * | 2012-05-09 | 2012-09-12 | 苏州爱锐邦智能科技有限公司 | Coin selector with appearance anti-fake function |
| US9922486B2 (en) | 2013-01-24 | 2018-03-20 | Arjo Solutions | Unique identification of coin or other object |
| CN106683259A (en) * | 2017-03-06 | 2017-05-17 | 深圳怡化电脑股份有限公司 | Method and device for identifying paper money |
| CN106683259B (en) * | 2017-03-06 | 2019-04-26 | 深圳怡化电脑股份有限公司 | A kind of discrimination method and device of bank note |
| CN111599080A (en) * | 2019-02-20 | 2020-08-28 | 深圳怡化电脑股份有限公司 | Spliced paper money detection method and device, financial machine tool equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| JP3945379B2 (en) | 2007-07-18 |
| CN100530232C (en) | 2009-08-19 |
| CN1499437A (en) | 2004-05-26 |
| CN1499438A (en) | 2004-05-26 |
| CN1499436A (en) | 2004-05-26 |
| CN100530233C (en) | 2009-08-19 |
| JP2004157727A (en) | 2004-06-03 |
| CN100552711C (en) | 2009-10-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1499435A (en) | Figured pattern identifier | |
| US9256912B2 (en) | Method of measuring measurement target | |
| US20020057428A1 (en) | Optical member inspecting apparatus and method of inspection thereof | |
| CN1271575C (en) | Coin discriminating method and device | |
| CN1711559A (en) | Characteristic region extraction device, characteristic region extraction method, and characteristic region extraction program | |
| WO2001084128A1 (en) | Inspection method for foreign matters inside through hole | |
| CN1300994A (en) | Device and method for image pick-up | |
| JP4321344B2 (en) | Pattern recognition device | |
| JP2006172080A (en) | Parameter setup method of light source, and recognition method of identification symbol | |
| JP4321343B2 (en) | Pattern recognition device | |
| JP3985733B2 (en) | Pattern recognition device | |
| JP3945397B2 (en) | Pattern recognition device | |
| JP4161790B2 (en) | Pattern recognition device | |
| JP2005083776A (en) | Grain classifier | |
| JP4161766B2 (en) | Pattern recognition device | |
| JP4148032B2 (en) | Pattern recognition device | |
| JP4148016B2 (en) | Pattern recognition device | |
| JP3982450B2 (en) | Pattern recognition device | |
| CN113933300A (en) | Garbage misdelivery detection method, device and system based on machine vision | |
| JP3982422B2 (en) | Pattern recognition device | |
| JP2005316552A (en) | Pattern identifying device | |
| JP2004157974A (en) | Pattern identifying device | |
| JP2004157962A (en) | Pattern identifying device | |
| JP2004157972A (en) | Pattern identifying device | |
| JP2004157975A (en) | Pattern identifying device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20090819 Termination date: 20141106 |
|
| EXPY | Termination of patent right or utility model |