CN108416814A - Quick positioning and recognition methods and the system on a kind of pineapple head - Google Patents
Quick positioning and recognition methods and the system on a kind of pineapple head Download PDFInfo
- Publication number
- CN108416814A CN108416814A CN201810139803.3A CN201810139803A CN108416814A CN 108416814 A CN108416814 A CN 108416814A CN 201810139803 A CN201810139803 A CN 201810139803A CN 108416814 A CN108416814 A CN 108416814A
- Authority
- CN
- China
- Prior art keywords
- image
- pineapple
- value
- color
- point
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/181—Segmentation; Edge detection involving edge growing; involving edge linking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of quick positioning on pineapple head and recognition methods and system, the method step to specifically include：Acquisition may include the RGB image of pineapple information；RGB image is converted into HSV images, to it into row threshold division, profile processing, the image-region of above-mentioned position in RGB image is extracted, as area-of-interest in location area maximum suspected target object area position；The color histogram for generating area-of-interest, it is matched with the pineapple head color histogram under preset varying environment, judges whether similarity reaches given threshold；Area-of-interest pineapple head fruit eye feature is extracted simultaneously, features described above is input to grader, judgement pineapple head fruit eye whether there is, and calculate the centre coordinate of fruit eye；Similarity and the feature of fruit eye are obtained through the above steps.The present invention can realize the quick positioning and identification to pineapple head, advantageously reduce risk injured in pineapple picking process, improve labor efficiency.
Description
Technical field
The invention mainly relates to the quick positioning on the identification of computer vision and positioning more particularly to a kind of pineapple head with
Recognition methods and system.
Background technology
Fruit is the necessity in our daily lifes, and fruit-picking how preferably to be gone to be asked at what we needed to solve
Topic.Pineapple is a kind of very strong fruit of picking season property, makes fruit peak period and there was only or so two weeks, so improving pineapple picking
Efficiency be a certainty developing direction.It is all manually to pick that the picking of pineapple is most of at present, does so and consumes largely
Manpower and materials, and pineapple has larger thorn, it is possible to orchard worker can be caused to damage.So the automatic picking of research pineapple,
For the loss for reducing labour, the quality for increasing pineapple picking, the harvest for stablizing pineapple and the sales rate for improving fruit, all
It is of great significance.
Invention content
It is an object of the invention to be related to a kind of pineapple head to be quickly positioned at recognition methods, and this method calculates simpler
And rapidly pineapple head can be identified, it solves the problems, such as automatic identification pineapple, pineapple can be effectively improved
Automatically the quality picked.
The purpose of the present invention can be realized by following technical solution：
A kind of quick positioning and recognition methods on pineapple head, specifically include：
S1, acquisition may include the RGB image of pineapple information；
S2, RGB image is converted to HSV images, to it into row threshold division, image preprocessing, it is doubtful obtains several
The image-region of above-mentioned position in RGB image is extracted, as sense in target object area, the maximum region position of location area
Interest region；
S3, the color histogram for generating area-of-interest, by itself and the pineapple head face under preset varying environment
Color Histogram is matched, and judges whether similarity reaches setting pre-value；
Meanwhile the pineapple head fruit eye feature of area-of-interest is extracted, features described above is input to grader, judges pineapple
Head fruit eye whether there is, and calculate the centre coordinate of fruit eye；
S4, the feature by exporting pineapple head color similarity and fruit eye that step S3 is obtained, judge the figure of acquisition
Seem it is no include pineapple head.
Further, the Threshold segmentation, specific steps include：
3 groups of threshold values are set according to the maximum color accounting of HSV image inspections target after conversion first, are tone threshold respectively
It is worth range [H_down, H_up], saturation degree threshold range [S_down, S_up], brightness threshold range [V_down, V_up]；
The tone of HSV images, saturation degree, the value of brightness and above-mentioned 3 groups of threshold ranges are compared respectively, specially：If falling
In threshold range, it is set as 255；If beyond outside threshold range, it is set as 0；Described 255 indicate the pixel of white；0 table
Show the pixel of black.
When further, in step s 2 to image into row threshold division, due to most of pineapple cultivation external environment out of office,
Therefore certain undulating value is had below different luminous intensities by the brightness-value on the pineapple head of camera crawl.Cause
This, after obtaining luminous intensity according to luminous intensity equations, needs to compensate the threshold value of brightness according to luminous intensity.
Therefore, 3 groups of threshold values [H_down, H_up], [S_down, S_up], [V_down, V_up] setting formula are as follows：
H_up=upper_H
H_down=lower_H
S_up=upper_S
S_down=lower_S
V_up=upper_V+I × b
V_down=lower_V+I × b
Wherein, upper_H, lower_H, upper_S, lower_S, upper_V, lower_V indicate tone, saturation respectively
Degree and the bound color value of brightness, check in all in accordance with the hsv color table of comparisons, and b is illumination compensation coefficient, for a setting
Value, I is intensity of illumination, is calculated by following formula：
I=hvN/ (At)
Wherein, h is planck constant, and v indicates that frequency, A indicate that irradiation area area, N indicate to be irradiated in time interval t
Number of photons on A.In above-mentioned expression formula, it is [0, I*b] that I × b was indicated, which is the fluctuation range of upper limit threshold and lower threshold,
It is not offered as strictly increasing this value of I × b.In real process, undulating value should be adjusted according to the effect of identification.
Further, in the image preprocessing after step S2 Threshold segmentations, gray-scale map is first converted images into, then carry out
Corresponding processing.The thickness of image can be reduced and improve the processing speed of image by being converted into gray-scale map.After Threshold segmentation
The conversion formula that image is converted to gray-scale map is as follows：
Gray=0.3414 × H+0.5478 × S+0.1108 × V
Image after the Threshold segmentation is binary map, is the image of triple channel, and color thickness is 3, transformed gray scale
It is 1 that the value of pixel, which only has 0 and 255, transformed color of image thickness, in figure, substantially increases arithmetic speed.
Further, it in described image pretreatment, is specifically included by the transformed respective handling of image：Be filtered with
And edge filling.
It is filtered using a kind of method for the discontinuous white point polishing continuous X-axis and Y-axis, is named as XY axis and does not connect
Continuous Denoising Algorithm.Further, the discontinuous Denoising Algorithm of XY axis specifically includes：
Preset a judgment value, respectively to discontinuous small particles in the X-axis and Y-axis of transformed gray level image into
Row polishes, if white point is less than judgment value in the continuation degree of X-axis or Y-axis, then the small particles in X-axis or Y-axis will be smoothed,
Become black；If white point is greater than or equal to judgment value in the continuation degree of X-axis or Y-axis, then the small particles in X-axis or Y-axis
It would not be smoothed, that is, be left white.By the discontinuous Denoising Algorithm of XY axis, can quickly remove extra non-after Threshold segmentation
The color point of object.
For image after carrying out denoising according to the discontinuous Denoising Algorithm of XY axis, target image edge can become incomplete, be not easy to catch
Catch profile, it is therefore desirable to which edge filling is carried out to the image after denoising.
Further, the edge filling, specifically includes：
The rough edge treated for carrying out X-axis to the image after being filtered first, using the image after thresholding, to pixel
Point carries out the right side and subtracts a left side, as a result takes absolute value, if the point of white subtracts each other or the point of black subtracts each other, the value after subtracting each other all is 0, when
The point of black takes absolute value after subtracting the point of white or the point of white subtracts the point of black, and obtained value is just 255；By rough side
Edge treated point be image marginal point；
Edge filling is carried out, selects a marginal point first, using it as the center of circle, radius is 1 pair and is filtered in rear image
Edge judged；If there is continuous point in the roundlet that radius is 1, i.e., the quantity of 255 pixel is more than or equal to 3, i.e.,
The radius of a filling circle is set to carry out edge filling then just using the marginal point of the selection as the center of circle for continuous boundary；Weight
It is complete to obtain edge filling after all marginal points of the obtained image of rough edge treated have all judged for multiple above-mentioned steps
At image.Accurate edge detection is carried out to the image after edge filling, profile is stored completely.
It is divided into multiple independent profiles to the profile of continuum to store, it is quasi- to carry out maximum rectangle to each profile
It closes, i.e., using profile length and wide maximum value as the length and width in maximum fitted rectangle region, positions the maximum fitting of each profile
Region.The image-region of above-mentioned position in RGB image is extracted in the maximum fitted area position of location area, emerging as feeling
Interesting region.
Further, pixel color is carried out to area-of-interest according to distribution of color and statistic algorithm in the step S3
Statistics, specifically include：
The Color Statistical that each pixel in area-of-interest indicates is got up, certain color dividing regions are set
Between, color histogram is generated, is matched with color histogram pre-set under varying environment, if similarity reaches threshold
Value, then can be expressed as the pineapple head of certain similarity.
Further, color histogram pre-set under the varying environment, specifically includes：
The number of colours magnitude in same color section under varying environment is acquired, Xi is expressed as the environment under i-th kind of environment
It is worth variable, with the number of colours magnitude in a color interval under F (X1, X2, X3....) the expressions varying environment.Pass through
The data for inputting a large amount of same color sections (such as yellow) allow neural network constantly to change its Weight value and biases values, and
Deviation is reduced, similar function is fitted, generates canonical function histogram.
Further, pineapple head fruit eye feature recognition is carried out to area-of-interest in the step S3, specifically included：
Pineapple head fruit eye sample is acquired, positive negative sample is classified as, trains grader to carry out positive and negative sample by OpenCV
This training, wherein positive sample is target sample to be checked, and negative sample is other any images.It sizes picture and positive and negative
The ratio of sample carries out feature extraction training by a large amount of pineapple head fruit eye data, sets the parameter of grader, for example,
Peak excursion degree, maximum rotation angle, modified weight values and the biases values of background color；Positive negative sample is placed on same
Catalogue is trained, and the feature of fruit eye is obtained after training.
The system of another object of the present invention is to provide a kind of quick positioning and recognition methods based on pineapple head, promotees
Into the research to the automatic picker system of pineapple.
A kind of quick positioning on pineapple head and identifying system, specifically include：Acquisition module, processing module, supporting module；
Further, acquisition module is used to carry out Image Acquisition to pineapple, and will acquire image transmitting back to processing
Module；
Further, processing module is used to transmit acquisition module image progress Color Statistical and to the head of pineapple into
The extraction and matching of row feature export the information after image procossing；Acquisition module is fixed in supporting module；
Further, supporting module specifically includes：Parallel-plate, supporting rack；Parallel-plate by two supporting racks be placed on away from
The place of certain altitude from the ground, acquisition module are placed on parallel-plate center close to ground side, and processing module is placed on parallel
Plate center is far from ground side；Supporting module is used to support processing module, and certain height and angle are provided for acquisition module
To carry out Image Acquisition.
Further, the embeded processor in the processing module, may be used the calculating with image-capable
The arbitrary combination of machine is substituted.
Further, the camera minute surface of the acquisition module is parallel to the ground；The camera of the acquisition module uses
Resolution ratio is 640X480, size is that 350X350 is acquired image；Under this resolution ratio and size, the camera can
The feature and details for accurately grabbing pineapple head, efficiently reduce processing time.
The present invention has advantageous effect below compared with prior art：
Comparison for calculation methods of the present invention is simple and arithmetic speed is fast, can realize to the quick fixed of pineapple head
Position with more accurately identify, and need not cost can be reduced using high speed processor in systems in practice.
Description of the drawings
Fig. 1 is a kind of flow chart of the quick positioning and recognition methods on pineapple head in the embodiment of the present invention；
Fig. 2 is thresholding treated image in the embodiment of the present invention；
Fig. 3 is the discontinuous Denoising Algorithm schematic diagram of XY axis in the embodiment of the present invention；
Fig. 4 is the figure judged Y-axis in the embodiment of the present invention；
Fig. 5 is in the embodiment of the present invention to the figure after X-axis and Y-axis denoising；
Fig. 6 is that transformed gray-scale map carries out the filtered image of the discontinuous Denoising Algorithm of XY axis in the embodiment of the present invention；
Fig. 7 is the schematic diagram that edge picture point is chosen in the embodiment of the present invention；
Fig. 8 is that the marginal point chosen in the embodiment of the present invention carries out the image before edge filling；
Fig. 9 is that the marginal point chosen in the embodiment of the present invention carries out the image after edge filling；
Figure 10 is to carry out the image after edge filling in the embodiment of the present invention to the image after being filtered；
Figure 11 is to carry out the maximum fitted area that maximum process of fitting treatment obtains to entire target image in the embodiment of the present invention
And the figure of maximum fitted area centre coordinate；
Figure 12 is the 24 by stages figure of color that Color Statistical is used in the embodiment of the present invention；
Figure 13 is pineapple sarcocarp eye pattern in the embodiment of the present invention.
Specific implementation mode
Technical solution of the present invention is further described with reference to the accompanying drawings and embodiments, but protection scope of the present invention
It is not limited to that.
Embodiment：
A kind of quick positioning on pineapple head is as shown in Figure 1 with the flow chart of recognition methods.
S1, acquisition may contain pineapple head image information RGB image；
Further, the camera that described image uses is placed on apart from ground level as on 1.2 meters of parallel-plate,
It is fitted in side of the parallel-plate center close to ground.
Further, the camera that described image acquisition uses uses resolution ratio, the size pair of 350X350 of 640X480
Image is acquired, and under above-mentioned resolution ratio and size, camera can accurately grab the feature and details on pineapple head,
Efficiently reduce processing time.
S2, RGB image is converted to HSV images, to it into row threshold division, image preprocessing, it is doubtful obtains several
The image-region of above-mentioned position in RGB image is extracted, as sense in target object area, the maximum region position of location area
Interest region；
Further, Threshold segmentation specifically includes：
3 groups of threshold values are set according to the maximum color accounting of HSV image inspections target after conversion first, are tone threshold respectively
It is worth range [H_down, H_up], saturation degree threshold range [S_down, S_up], brightness threshold range [V_down, V_up]；
The tone of HSV images, saturation degree, the value of brightness and above-mentioned 3 groups of threshold ranges are compared respectively, specially：If falling
In threshold range, it is set as 255；If beyond outside threshold range, it is set as 0；Described 255 indicate the pixel of white；0 table
Show the pixel of black.
Further, due to most of pineapple cultivation external environment out of office, by the bright of the pineapple head of camera crawl
Brightness value has certain undulating value below different luminous intensities.Therefore, it is shone according to luminous intensity equations
After intensity, need to compensate the threshold value of brightness according to luminous intensity.
Therefore, 3 groups of threshold values [H_down, H_up], [S_down, S_up], [V_down, V_up] setting formula are as follows：
H_up=upper_H
H_down=lower_H
S_up=upper_S
S_down=lower_S
V_up=upper_V+I × b
V_down=lower_V+I × b
Wherein, upper_H, lower_H, upper_S, lower_S, upper_V, lower_V indicate tone, saturation respectively
Degree and the bound color value of brightness, check in all in accordance with the hsv color table of comparisons, and b is illumination compensation coefficient, for a setting
Value, I is intensity of illumination, is calculated by following formula：
I=hvN/ (At)
Wherein, h is planck constant, and v indicates that frequency, A indicate that irradiation area area, N indicate to be irradiated in time interval t
Number of photons on A；The hsv color table of comparisons is as follows.
The 1 hsv color table of comparisons of table
In the present embodiment, indoors due to the place of Image Acquisition, undulating value can be set to：It shines strong
Degree * 0.5%；
Therefore, it sets brightness threshold value to [V_down, V_up], it is as follows specifically to set formula：
V_up=upper_V+I × 0.005
V_down=lower_V+I × 0.005
Wherein, upper_V, lower_V are respectively brightness upper limit color value and brightness lower limit color value, according to HSV
Color chart checks in, and it is intensity of illumination that illumination compensation coefficient, which is set as 0.005, I,.
The general intensity of illumination in interior ideally is 2000lx, therefore in the present embodiment, brightness upper limit threshold
It is set as 41+10, lower threshold is set as 153+10, this data is obtained by normal illumination in ideally room on daytime；Not
With under intensity of illumination, the value corresponding to bound in the hsv color table of comparisons can also change, therefore in reality, root
The value of specific illumination compensation value and upper limit color and lower limit color is determined according to actual conditions；In the present embodiment, by bright
The undulating value of brightness threshold value known to the threshold setting procedure of brightness can fluctuate in [0, I*0.005] range, therefore, according to
The effect of the hsv color table of comparisons and actual feedback can set the brightness threshold value on pineapple head in actual conditions to adjust
The brightness-value of image.
Fig. 2 is shown carries out the image that thresholding processing obtains later to original shooting image, and white area is maximum in Fig. 2
It is pineapple head, image can see that the small particles being much included in, these small particles are the non-targeted object in image
Noise need that the image after Threshold segmentation is further processed in order to eliminate these noises, that is, convert the image into ash
Image is filtered after degree figure and edge filling.
It is filtered using a kind of method for the discontinuous white point polishing continuous X-axis and Y-axis, is named as XY axis and does not connect
Continuous Denoising Algorithm.The discontinuous Denoising Algorithm schematic diagram of XY axis is as shown in Figure 3.
Further, the discontinuous Denoising Algorithm of XY axis specifically includes：
Preset a judgment value, respectively to discontinuous small particles in the X-axis and Y-axis of transformed gray level image into
Row polishes, if white point is less than judgment value in the continuation degree of X-axis or Y-axis, then the small particles in X-axis or Y-axis will be smoothed,
Become black；If white point X-axis or Y-axis continuation degree heavy rain or be equal to judgment value, then the small particles in X-axis or Y-axis
It would not be smoothed, that is, be left white.By the discontinuous Denoising Algorithm of XY axis, can quickly remove extra non-after Threshold segmentation
The color point of object.
As shown in figure 4, gray-scale map carries out denoising according to the discontinuous Denoising Algorithm of XY axis：Judgment value is set as 3, is located in X-axis
3 points meet judgment value, 2 in Y-axis point does not meet judgment value.Result such as Fig. 5 institutes after gray-scale map denoising in Fig. 4
Show：3 points in X-axis are not smoothed, and 2 in Y-axis point is smoothed.In the present embodiment, selection uses and sentences
Disconnected value carries out the discontinuous Denoising Algorithm denoising of XY axis for 2.
For image after carrying out denoising according to the discontinuous Denoising Algorithm of XY axis, target image edge can become incomplete, be not easy to catch
Profile is caught, as shown in Figure 6, it is therefore desirable to which edge filling is carried out to the image after denoising.
Further, the edge filling, specifically includes：
The rough edge treated for carrying out X-axis to the image after being filtered first, using the image after thresholding, to pixel
Point carries out the right side and subtracts a left side, as a result takes absolute value, if the point of white subtracts each other or the point of black subtracts each other, the value after subtracting each other all is 0, when
The point of black takes absolute value after subtracting the point of white or the point of white subtracts the point of black, and obtained value is just 255；By rough side
Edge treated point be image marginal point；
According to rough edge treated, the marginal point in image is obtained, according to the marginal point of gained, to the figure after being filtered
Picture, i.e. Fig. 6 in the present embodiment carry out edge filling.
Edge filling is carried out, selects a marginal point, the principle for choosing marginal point as shown in Figure 7 first；With the side of selection
Edge point is the center of circle, and radius is that the edge in 1 pair of image judges；If there is continuous point in the roundlet that radius is 1, i.e., 255
The quantity of pixel be more than or equal to 3, as continuous boundary, then just using the marginal point of the selection as the center of circle, the edge of selection
Image before point progress edge filling is as shown in figure 8, setting one fills round radius to carry out edge filling；In the present embodiment
In, it fills round radius and takes R=3；The pixel that all values are 0 is set to 255 in this circle, the marginal point of selection carries out
Image after edge filling is as shown in Figure 9；It repeats the above steps, until all edges of the obtained image of rough edge treated
After point has all judged, the image of edge filling completion is obtained.Image after being filtered to the present embodiment Fig. 6 carries out edge filling
The image obtained afterwards is as shown in Figure 10.
Accurate edge detection is carried out to the image after edge filling, profile is stored completely.
It is divided into multiple independent profiles to the profile of continuum to store, it is quasi- to carry out maximum rectangle to each profile
It closes, i.e., using profile length and wide maximum value as the length and width in maximum fitted rectangle region, to obtain the maximum of each profile
Fitted area.
The maximum rectangle fitting method, specially：The edge pixel point for finding a profile, defines the maximum of X-coordinate
Value is Max.x, minimum value is Min.x and the maximum value of Y-axis is Max.y, minimum value Min.y.Therefore, the width of profile is defined
Length for Wide, profile is Length and the centre coordinate of profile is (X, Y), the width of profile, the length of profile with
And the solution formula of the centre coordinate of profile is as follows：
Wide=Max.x-Min.x
Length=Max.y-Min.y
X=(Max.x+Min.x)/2
Y=(Max.y+Min.y)/2
Referring to Figure 11, maximum fitted area and maximum to be obtained to the maximum process of fitting treatment of entire target image progress are intended
Close the figure of regional center coordinate；Wherein, W indicates that maximum width, L indicate maximum length in figure, and black indicates in profile at center
Imago element.The maximum fitted area, for the rectangle of the maximum region fitting of area obtained after all contour fittings；Extraction
The image-region of above-mentioned maximum fitted area in RGB image, as area-of-interest；
S3, the color histogram for generating area-of-interest, by itself and the pineapple head face under preset varying environment
Color Histogram is matched, and judges whether similarity reaches setting pre-value；
The pineapple head fruit eye feature for extracting area-of-interest, grader, judgement pineapple head are input to by features described above
Fruit eye whether there is, and calculate the centre coordinate of fruit eye；
Further, the statistics for carrying out pixel color to area-of-interest according to distribution of color and statistic algorithm,
It specifically includes：
The Color Statistical that each pixel in area-of-interest indicates is got up, certain color dividing regions are set
Between, color histogram is generated, is matched with color histogram pre-set under varying environment, if similarity reaches threshold
Value, then can be expressed as the pineapple head of certain similarity.
Further, color demarcation interval can be divided into the section of the colors such as 12,24,46,72 point, what color interval divided
More, precision is higher, but operation efficiency is slower, therefore, in the present embodiment, in order to realize quick identification, for pineapple
Head feature identification uses 24 points of color interval.24 points of color interval is as shown in figure 12.
Further, color histogram pre-set under the varying environment, specifically includes：
The number of colours magnitude in same color section under varying environment is acquired, Xi is expressed as the environment under i-th kind of environment
It is worth variable, with the number of colours magnitude in a color interval under F (X1, X2, X3....) the expressions varying environment.Pass through
The data for inputting a large amount of same color sections (such as yellow) allow neural network constantly to change its Weight value and biases values, and
Deviation is reduced, similar function is fitted.
In the present embodiment, in order to reduce the complexity of calculating, just using intensity of illumination for 2000lx indoor environment as ring
Border variable.So just there are the reference axis of Y=F (X), input sample to generate data point, Function Fitting is carried out to data point.Not
After obtaining relatively good fitting function with color interval, the Color Statistical figure of standard comparing can be generated.It is carried out with object
Matching, according to similarity obtain whether be pineapple head.
Further, described that pineapple head fruit eye feature recognition is carried out to area-of-interest, it specifically includes：
Pineapple head fruit eye sample is acquired, positive negative sample is classified as, trains grader to carry out positive and negative sample by OpenCV
This training, wherein positive sample is target sample to be checked, and negative sample is other any images.It sizes picture and positive and negative
The ratio of sample carries out feature extraction training by a large amount of pineapple head fruit eye data, sets the parameter of grader, for example,
Peak excursion degree, maximum rotation angle, modified weight values and the biases values of background color, positive negative sample are placed on same
Catalogue is trained, and the feature of fruit eye is obtained after training.Pineapple head fruit eye is as shown in figure 13.
S4, the spy by exporting the pineapple head color similarity obtained after step S3 image recognition processings and fruit eye
Sign judges whether the image of acquisition includes pineapple head.
The foregoing is merely the protection domains of the preferred embodiment of the invention, but patent of the present invention to be not limited thereto, and appoints
What those familiar with the art is in the range disclosed in patent of the present invention, according to the inventive concept of patent of the present invention
Or technical solution is subject to equivalent substitution or change, belongs to the protection domain of patent of the present invention.
Claims (10)
1. the quick positioning and recognition methods on a kind of pineapple head, it is characterised in that：Specific steps include：
S1, acquisition may include the RGB image of pineapple information；
S2, RGB image is converted to HSV images, to it into row threshold division, image preprocessing, obtains several suspected targets
The image-region of above-mentioned position in RGB image is extracted, as interested in object area, the maximum region position of location area
Region；
S3, the color histogram for generating area-of-interest are straight with the pineapple head color under preset varying environment by it
Square figure is matched, and judges whether similarity reaches given threshold；
Meanwhile area-of-interest pineapple head fruit eye feature is extracted, features described above is input to grader, judgement pineapple head fruit
Eye whether there is, and calculate the centre coordinate of fruit eye；
S4, the feature by exporting pineapple head color similarity and pineapple head fruit eye that step S3 is obtained, judge to acquire
Image whether include pineapple head.
2. the quick positioning and recognition methods on a kind of pineapple head according to claim 1, it is characterised in that：The threshold value
Segmentation, specifically includes：
3 groups of threshold values are set according to the maximum color accounting of HSV image inspections target after conversion first, are hue threshold model respectively
Enclose [H_down, H_up], saturation degree threshold range [S_down, S_up], brightness threshold range [V_down, V_up]；By HSV
The tone of image, saturation degree, the value of brightness are compared respectively with above-mentioned 3 groups of threshold ranges, specially：If falling in threshold value model
In enclosing, it is set as 255；If beyond outside threshold range, it is set as 0；Described 255 indicate the pixel of white；Described 0 indicates black
Pixel.
3. the quick positioning and recognition methods on a kind of pineapple head according to claim 2, it is characterised in that：3 groups of threshold values
[H_down, H_up], [S_down, S_up], [V_down, V_up] setting formula are as follows：
H_up=upper_H
H_down=lower_H
S_up=upper_S
S_down=lower_S
V_up=upper_V+I × b
V_down=lower_V+I × b
Wherein, upper_H, lower_H, upper_S, lower_S, upper_V, lower_V indicate respectively tone, saturation degree with
And the bound color value of brightness, it is checked in all in accordance with the hsv color table of comparisons；Brightness-value has wave under different illumination intensity
Dynamic value, it is therefore desirable to the threshold value of brightness be compensated according to intensity of illumination；B is illumination compensation coefficient, is a setting value, I
For intensity of illumination, calculated by following formula：
I=hvN/ (At)
Wherein, h is planck constant, and v indicates that frequency, A indicate that irradiation area area, N indicate to be irradiated on A in time interval t
Number of photons.
4. the quick positioning and recognition methods on a kind of pineapple head according to claim 1, it is characterised in that：In the threshold
In image preprocessing after value segmentation, in order to reduce the thickness of image and improve the processing speed of image, need HSV images
Gray-scale map is converted to, conversion formula is as follows：
Gray=0.3414 × H+0.5478 × S+0.1108 × V
Wherein, Gray is gradation of image, and H, S, V are respectively tone, saturation degree and the value of brightness of image.
5. the quick positioning and recognition methods on a kind of pineapple head according to claim 4, it is characterised in that：Step S2's
In described image pretreatment, gray-scale map after conversion is filtered, method is as follows：
A judgment value is preset, discontinuous white point in the X-axis and Y-axis of transformed gray-scale map is polished respectively,
I.e.：If white point becomes black less than judgment value in the continuation degree of X-axis or Y-axis then the white point in X-axis or Y-axis will be smoothed
Color；If white point is greater than or equal to judgment value in the continuation degree of X-axis or Y-axis, then the white point in X-axis or Y-axis would not be smoothed,
It is left white.
6. the quick positioning and recognition methods on a kind of pineapple head according to claim 1, it is characterised in that：Step S2's
In image preprocessing, profile processing is carried out to image after being filtered, specific steps include：
First, the rough edge treated of X-axis is carried out to filtered treated image, it is right for the image after being filtered
Pixel carries out the right side and subtracts a left side, the operation as a result to take absolute value, if the point of white subtracts each other or the point of black subtracts each other, after subtracting each other
Value is all 0, takes absolute value after the point of black subtracts the point of white or the point of white subtracts the point of black, obtained value is just 255；
Point after rough edge treated is the marginal point of image；
Then, edge filling is carried out to the image after being filtered, selects a marginal point, using it as the center of circle, radius 1 first
Edge in image is judged；If thering is continuous point, i.e., the quantity of 255 pixel to be more than in the roundlet that radius is 1
Equal to 3, as continuous boundary sets the radius of a filling circle to carry out side then just using the marginal point of the selection as the center of circle
Edge is filled；It repeats the above steps, after all marginal points of the obtained image of rough edge treated have all judged, obtains side
The image that edge filling is completed；Accurate edge detection is carried out to the image after edge filling, profile is stored completely.
7. the quick positioning and recognition methods on a kind of pineapple head according to claim 1, it is characterised in that：Step S3
In, the Color Statistical that each pixel in area-of-interest indicates is got up, certain color demarcation interval is set, is generated
Color histogram is matched with color histogram pre-set under varying environment, if similarity reaches threshold value, then it represents that
It include pineapple head in the image of acquisition；
The feature recognition that pineapple head fruit eye is carried out to area-of-interest, specifically includes：
Pineapple head fruit eye sample is acquired, positive negative sample is classified as, trains grader to carry out positive negative sample by OpenCV
Training, wherein positive sample is target sample to be checked, and negative sample is other any images；It sizes picture and positive and negative sample
Ratio, by a large amount of pineapple head fruit eye data carry out feature extraction training, the parameter of grader is set, by positive negative sample
It is placed on same catalogue to be trained, the feature of fruit eye is obtained after training.
8. the quick positioning and recognition methods on a kind of pineapple head according to claim 7, it is characterised in that：The difference
Pre-set color histogram, specifically includes under environment：
The number of colours magnitude in same color section under varying environment is acquired, Xi indicates the environment value variable under i-th kind of environment,
F (X1, X2, X3....) indicates the number of colours magnitude in a color interval under varying environment；By inputting a large amount of same face
The data in color section allow neural network constantly to change its Weight value and biases values, and reduce deviation, fit similar
Function generates canonical function histogram.
9. a kind of quick positioning of quick positioning and recognition methods for realizing the pineapple heads any one of claim 1-8
With identifying system, it is characterised in that：The system comprises：
Acquisition module for carrying out Image Acquisition to pineapple, and will acquire image transmitting back to processing module；
Processing module, the image for being transmitted to acquisition module carry out Color Statistical and carry out the extraction of feature to the head of pineapple
And matching, export the information after image procossing；Acquisition module is fixed in supporting module；
Supporting module, including parallel-plate and supporting rack；Parallel-plate is placed on by two supporting racks apart from ground certain altitude
Place, acquisition module are placed on parallel-plate center close to ground side, and processing module is placed on parallel-plate center far from ground one
Side；Supporting module is used to support processing module, and provides certain height and angle for acquisition module to carry out Image Acquisition.
10. quick positioning according to claim 9 and identifying system, it is characterised in that：The camera of the acquisition module
Minute surface is parallel to the ground；The camera of the acquisition module use resolution ratio for 640X480, size be 350X350 to image into
Row acquisition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810139803.3A CN108416814B (en) | 2018-02-08 | 2018-02-08 | Method and system for quickly positioning and identifying pineapple head |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810139803.3A CN108416814B (en) | 2018-02-08 | 2018-02-08 | Method and system for quickly positioning and identifying pineapple head |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108416814A true CN108416814A (en) | 2018-08-17 |
| CN108416814B CN108416814B (en) | 2020-07-31 |
Family
ID=63128336
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810139803.3A Active CN108416814B (en) | 2018-02-08 | 2018-02-08 | Method and system for quickly positioning and identifying pineapple head |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108416814B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109376746A (en) * | 2018-10-25 | 2019-02-22 | 黄子骞 | A kind of image identification method and system |
| CN109376257A (en) * | 2018-10-24 | 2019-02-22 | 贵州省机电研究设计院 | Tealeaves recognition methods based on image procossing |
| CN109903275A (en) * | 2019-02-13 | 2019-06-18 | 湖北工业大学 | A kind of mouldy method for detecting area of fermented grain based on self-adapting multi-dimension filtering and histogram comparison |
| CN110415181A (en) * | 2019-06-12 | 2019-11-05 | 勤耕仁现代农业科技发展(淮安)有限责任公司 | Flue-cured tobacco RGB image intelligent recognition and grade determination method under a kind of open environment |
| CN112183230A (en) * | 2020-09-09 | 2021-01-05 | 上海大学 | Identification and central point positioning method for pears in natural pear orchard environment |
| CN113902909A (en) * | 2021-10-15 | 2022-01-07 | 江阴仟亿日化包装有限公司 | Cloud storage type appearance analysis system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN201600330U (en) * | 2009-09-23 | 2010-10-06 | 中国农业大学 | System for recognizing and locating mature pineapples |
| CN105095880A (en) * | 2015-08-20 | 2015-11-25 | 中国民航大学 | LGBP encoding-based finger multi-modal feature fusion method |
| US20160203591A1 (en) * | 2015-01-09 | 2016-07-14 | Umm Al-Qura University | System and process for monitoring the quality of food in a refrigerator |
-
2018
- 2018-02-08 CN CN201810139803.3A patent/CN108416814B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN201600330U (en) * | 2009-09-23 | 2010-10-06 | 中国农业大学 | System for recognizing and locating mature pineapples |
| US20160203591A1 (en) * | 2015-01-09 | 2016-07-14 | Umm Al-Qura University | System and process for monitoring the quality of food in a refrigerator |
| CN105095880A (en) * | 2015-08-20 | 2015-11-25 | 中国民航大学 | LGBP encoding-based finger multi-modal feature fusion method |
Non-Patent Citations (4)
| Title |
|---|
| F. GARCÍA ET.AL: "Fruit Classification by Extracting Color Chromaticity, Shape and Texture Features: Towards an Application for Supermarkets", 《IEEE LATIN AMERICA TRANSACTIONS》 * |
| LI BIN ET.AL: "Construction and in-field experiment of low-cost binocular vision platform for pineapple harvesting robot", 《农业工程学报》 * |
| MUHAMMAD AZMI AHMED NAWAWI ET.AL: "Simulation and Segmentation Techniques for Crop Maturity Identification of Pineapple Fruit", 《SPRINGER NATURE SINGAPORE PTE LTD》 * |
| 李斌 等: "基于单目视觉的田间菠萝果实识别", 《农业工程学报》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109376257A (en) * | 2018-10-24 | 2019-02-22 | 贵州省机电研究设计院 | Tealeaves recognition methods based on image procossing |
| CN109376746A (en) * | 2018-10-25 | 2019-02-22 | 黄子骞 | A kind of image identification method and system |
| CN109903275A (en) * | 2019-02-13 | 2019-06-18 | 湖北工业大学 | A kind of mouldy method for detecting area of fermented grain based on self-adapting multi-dimension filtering and histogram comparison |
| CN109903275B (en) * | 2019-02-13 | 2021-05-18 | 湖北工业大学 | Fermented grain mildewing area detection method based on self-adaptive multi-scale filtering and histogram comparison |
| CN110415181A (en) * | 2019-06-12 | 2019-11-05 | 勤耕仁现代农业科技发展(淮安)有限责任公司 | Flue-cured tobacco RGB image intelligent recognition and grade determination method under a kind of open environment |
| CN110415181B (en) * | 2019-06-12 | 2023-07-14 | 勤耕仁现代农业科技发展(淮安)有限责任公司 | Intelligent identification and grade judgment method for RGB (red, green and blue) images of flue-cured tobacco in open environment |
| CN112183230A (en) * | 2020-09-09 | 2021-01-05 | 上海大学 | Identification and central point positioning method for pears in natural pear orchard environment |
| CN113902909A (en) * | 2021-10-15 | 2022-01-07 | 江阴仟亿日化包装有限公司 | Cloud storage type appearance analysis system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108416814B (en) | 2020-07-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108416814A (en) | Quick positioning and recognition methods and the system on a kind of pineapple head | |
| CN105354865B (en) | The automatic cloud detection method of optic of multispectral remote sensing satellite image and system | |
| CN104483326B (en) | High-voltage line defects of insulator detection method and system based on depth belief network | |
| CN109584251A (en) | A kind of tongue body image partition method based on single goal region segmentation | |
| CN106651872A (en) | Prewitt operator-based pavement crack recognition method and system | |
| CN111915704A (en) | Apple hierarchical identification method based on deep learning | |
| CN103679173B (en) | Method for detecting image salient region | |
| CN110276386A (en) | A kind of apple grading method and system based on machine vision | |
| CN108549910A (en) | A kind of corn seed producing fruit ear image preliminary classification method based on convolutional neural networks | |
| CN105447825B (en) | Image defogging method and its system | |
| CN108664839A (en) | A kind of image processing method and equipment | |
| CN108470338A (en) | A kind of water level monitoring method | |
| CN109543518A (en) | A kind of human face precise recognition method based on integral projection | |
| CN111985435A (en) | Unmanned aerial vehicle water area monitoring and cruising method based on machine vision | |
| Mei et al. | Single image dehazing using dark channel fusion and haze density weight | |
| Ali et al. | Low cost solution for rice quality analysis using morphological parameters and its comparison with standard measurements | |
| CN103778428B (en) | Disease area-of-interest exacting method and system based on block labelling | |
| CN107545565A (en) | A kind of solar energy half tone detection method | |
| CN112288682A (en) | Electric power equipment defect positioning method based on image registration | |
| CN112101260A (en) | Method, device, equipment and storage medium for identifying safety belt of operator | |
| Yang et al. | Cherry recognition based on color channel transform | |
| CN109993228A (en) | Plant protection drone rice sheath blight disease recognition methods based on machine vision | |
| CN111882529B (en) | Mura defect detection method and device for display screen | |
| CN112337810B (en) | Vision guiding pearl sorting robot and sorting method thereof | |
| CN108007929A (en) | A kind of automatic judging method of beef physiological makeup |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |