CN104978547B - Articles detection system - Google Patents
Articles detection system Download PDFInfo
- Publication number
- CN104978547B CN104978547B CN201410129820.0A CN201410129820A CN104978547B CN 104978547 B CN104978547 B CN 104978547B CN 201410129820 A CN201410129820 A CN 201410129820A CN 104978547 B CN104978547 B CN 104978547B
- Authority
- CN
- China
- Prior art keywords
- image
- characteristic point
- point
- vector
- module
- 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.)
- Expired - Fee Related
Links
Landscapes
- Image Analysis (AREA)
Abstract
The present invention provides a kind of articles detection system, including storage unit, image collection module, feature search module, computing module and confirmation module.Image collection module reads first, second and the third image that first, second and third time point are corresponded respectively in video streaming from storage unit, and wherein the time difference at first and second time point is less than or equal to the time difference of second with third time point.Feature search module searches characteristic point from the first image.When computing module judging characteristic point is located at the detection zone of short distance, according to the first image and the second image, the motion-vector of characteristic point is calculated；Conversely, computing module calculates the motion-vector of characteristic point according to first and second image one of them and third image.Confirmation module confirms that moving article whether there is according to the motion-vector of characteristic point.
Description
Technical field
The invention relates to a kind of articles detection systems, and in particular to a kind of moving article detection system.
Background technique
As sale of automobile amount is constantly grown up, auto electronic industry also flourishes, and wherein auto electronic industry includes
Automotive safety system, bodywork system, driver information system, suspention chassis system, engine transmission system, safety system etc. six is big
Aspect, and its using automotive safety system year multiple growth rate as highest.
However, the probability that road accident occurs also increases year by year under automobile quantity is growing.According to statistics
It has been shown that, as long as driver obtains early warning before 0.5 second to collide, it can avoid at least 60% the car accident that knocks into the back,
30% head-on car accident and 50% road surface related accidents, and if have 1 second pre-warning time, can avoid 90% thing
Therefore.Therefore, the importance of automotive safety system has some idea of.
With vision facilities cost be greatly reduced and computerized image identification technique make rapid progress, based on computerized image
Vehicle and pedestrian warn technology to have become main direction of studying.Such technology is mainly onboard installed based on computer vision
Detection system, continuously shot images in vehicle traveling, then with the detection of the technologies such as image procossing, computer vision and pattern recognition
Road surface, roadside, vehicle, pedestrian etc. be static in other travelings and multidate information, and then accurately senses environmental aspect, to provide
The more running informations of driver.
Detection system based on computer vision has many advantages, such as diversity application, economy and using elasticity.However, for
For the algorithm of various image procossings, the accuracy and efficiency difference of detection.Therefore, how a kind of Gao Zhun is provided
True rate and dynamical detection system one of are a topic of concern to those skilled in the art in fact.
Summary of the invention
The present invention provides a kind of articles detection system, can accurately and in real time detect the motive objects in video streaming
Part.
The present invention provides a kind of articles detection system, including storage unit, image collection module, feature search module, meter
Calculate module and confirmation module.Storage unit includes at least the video streaming of multiple images to store.Image collection module, coupling
Storage unit is connect, corresponds to the first image of first time point in video streaming, corresponding to second to read from storage unit
Second image at time point and third image corresponding to third time point, wherein first time point and the second time point when
Between difference be less than or equal to the second time point and third time point time difference.Feature search module from the first image to search
At least one characteristic point.Computing module is to judge that the characteristic point is located at the detection zone or detection at a distance of short distance
Region.When the characteristic point is located at the detection zone of short distance, computing module is calculated according to the first image and the second image
The motion-vector of each characteristic point；When the characteristic point is located at remote detection zone, computing module is according to the first figure
Picture and the second image one of them and third image, calculate the motion-vector of each characteristic point.Confirmation module is to root
According to the motion-vector of each characteristic point, confirm that moving article whether there is.
In one embodiment of this invention, the characteristic point that above-mentioned feature search module is searched is that a 1x1 is drawn
Plain or multiple picture elements are constituted.
In one embodiment of this invention, above-mentioned computing module is to utilize multiple pixel characteristic Comparison Methods or optical flow method
Calculate the motion-vector of each characteristic point.
In one embodiment of this invention, above-mentioned computing module calculates each according to the motion-vector of each characteristic point
The displacement of the characteristic point, to calculate the movement speed of each characteristic point, and confirmation module judges each characteristic point
Movement speed whether be greater than the speed threshold value with each characteristic point distance dependent.When confirmation module judges at least one feature
When the movement speed of point is greater than speed threshold value, confirmation module confirms that moving article exists.
The present invention provides another articles detection system, including storage unit, image collection module, feature search module,
Computing module and confirmation module.Storage unit includes at least the video streaming of multiple images to store.Image collection module,
Storage unit is coupled, corresponds to the first image of first time point in video streaming to be read from storage unit, correspond to the
Second image at two time points, the third image corresponding to third time point and the 4th image corresponding to the 4th time point,
Wherein the time difference at first time point and the second time point is less than or equal to the time difference at third time point and the 4th time point.It is special
Search module is levied to search at least one characteristic point from the first image.Computing module is to judge that it is close that the characteristic point is located at
The detection zone of distance or remote detection zone.When the characteristic point is located at the detection zone of short distance, mould is calculated
Root tuber calculates the motion-vector of each characteristic point according to the first image and the second image；When the characteristic point is located at a distance
Detection zone when, computing module calculates the motion-vector of each characteristic point according to third image and the 4th image.Confirm mould
Block confirms that moving article whether there is to the motion-vector according to each characteristic point.
In one embodiment of this invention, the characteristic point that above-mentioned feature search module is searched is that a 1x1 is drawn
Plain or multiple picture elements are constituted.
In one embodiment of this invention, above-mentioned computing module is to utilize multiple pixel characteristic Comparison Methods or optical flow method
Calculate the motion-vector of each characteristic point.
In one embodiment of this invention, above-mentioned computing module calculates each according to the motion-vector of each characteristic point
The displacement of the characteristic point, to calculate the movement speed of each characteristic point, and confirmation module judges each characteristic point
Movement speed whether be greater than the speed threshold value with each characteristic point distance dependent.When confirmation module judges at least one feature
When the movement speed of point is greater than speed threshold value, confirmation module confirms that moving article exists.
Based on above-mentioned, articles detection system provided by the present invention is calculated using image different in video streaming
The motion-vector of characteristic point.When characteristic point is located at the detection zone of far and near distance, using time difference lesser two images come
The motion-vector of characteristic point is calculated, to shorten the time of object detection；When characteristic point is located at remote detection zone, utilize
Time difference biggish two images calculate the motion-vector of characteristic point, to improve the accuracy rate of object detection.In addition to this, originally
Invention also by the motion-vector of characteristic point, calculates the displacement of characteristic point, to calculate the movement speed of characteristic point, is moved with confirming
The presence of animal part.Base this, articles detection system of the invention can more accurately grasp the information of moving article, and can reach
The efficiency of scan picture, to apply on the consumer electrical product of low cost, the enhancing present invention is in practical applications
Applicability.
To make the foregoing features and advantages of the present invention clearer and more comprehensible, special embodiment below, and it is detailed to cooperate attached drawing to make
Carefully it is described as follows.
Detailed description of the invention
Fig. 1 is the block diagram of articles detection system shown by an embodiment according to the present invention；
Fig. 2 is the flow chart according to object detection method shown by first embodiment of the invention；
Fig. 3 A is the schematic diagram of the first image shown by an embodiment according to the present invention；
Fig. 3 B is the schematic diagram of step S209 in object detection method shown by an embodiment according to the present invention；
Fig. 4 is the flow chart according to object detection method shown by second embodiment of the invention.
Description of symbols:
100: articles detection system；
110: storage unit；
120: image collection module；
130: feature search module；
140: computing module；
150: confirmation module；
S201~S211, S401~S411: the process of object detection method；
30: object；
310: the first images；
320: the second images；
330: third image；
312,332: remote detection zone；
314,324: the detection zone of short distance.
Specific embodiment
In general, in detecting dynamic image when the mobile message of moving article, no matter moving article position
Distance is often detected using identical one group of image.In the case, when moving article is located at farther away detection zone
When domain, then the mobile message estimated is possible to that great error can be generated with practical mobile message.
For example, the example of table one is to be located at closely to move with remote using identical one group of image detection
The speed estimation error of animal part:
Table one
In every image, the reality of real space corresponding to the actual displacement picture element of each picture element (actual space)
Displacement is different.In this example, it is assumed that in short distance, actual displacement picture element corresponding to each actual displacement picture element is
50cm；When remote, actual displacement picture element corresponding to each actual displacement picture element is 100cm.It is deposited in estimation displacement picture element
In the case where 1 pixel ratio is to error (One-Pixel Matching Error), for short distance with speed (5,5)
The velocity error that cm/ms mobile moving article is estimated is 20%, and at a distance equally with speed (5,5) cm/ms movement
The velocity error estimated of moving article be then 40%.Therefore, if coming to carry out speed to moving article with identical one group of image
Estimation, can generate biggish speed estimation error positioned at remote moving article.Base this, idea of the invention is first to judge
The distance of moving article position obtains to readaptability the mobile message of moving article using different groups of images.
Next section Example of the invention will cooperate attached drawing to be described in detail, element cited in description below
Symbol will be regarded as the same or similar element when identical component symbol occur in different attached drawings.These embodiments are the present invention
A part, all embodiments of the invention are not disclosed.More precisely, these embodiments are patent of the invention
The example of apparatus and method in application range.
Fig. 1 is the block diagram of articles detection system shown by an embodiment according to the present invention, but this is merely for convenience
Illustrate, is not intended to limit the invention.Fig. 1 first first introduces all components and configuration relation of articles detection system, in detail
Function will cooperate Fig. 2 to disclose together.
Fig. 1 is please referred to, articles detection system 100 includes storage unit 110, image collection module 120, feature search module
130, computing module 140 and confirmation module 150.Articles detection system 100 can be built-in or be external in PC, notebook
Computer, digital camera, digital code camera, network camera, smart phone, tablet computer, drive recorder, vehicle video-audio system
Deng the electronic device with shooting function, invention is not limited thereto.
Storage unit 110 may, for example, be any form of fixed or packaged type random access memory (Random
Access Memory, abbreviation RAM), read-only memory (Read-Only Memory, abbreviation ROM), flash memory (Flash
Memory), the combination of hard disk or other similar device or these devices, to store image data.
Image collection module 120, feature search module 130, computing module 140 and confirmation module 150 can be by software journeys
Sequence, hardware circuit or combinations thereof are realized, to execute the function of object detection.Software is, for example, source code, application software, drive
Dynamic program is specifically intended for realizing software module or function of specific function etc..Hardware is, for example, central processing unit (Central
Processing Unit, abbreviation CPU), programmable controller, digital signal processor (Digital Signal
Processor, abbreviation DSP) or other programmables general service or specific use microprocessor
(Microprocessor) etc..For example, image collection module 120, feature search module 130, computing module 140 and really
Recognizing module 150 is, for example, computer program or instruction, can be examined by the processor of articles detection system 100 thereby executing object
The function of survey.
Fig. 2 is the flow chart according to object detection method shown by first embodiment of the invention.Referring to figure 2., this reality
The method for applying example is suitable for the articles detection system 100 of Fig. 1, and each item i.e. in collocation articles detection system 100 is said below
The detailed step of bright object detection method.It in the present embodiment, will be by taking articles detection system 100 be coupled to driving recorder as an example
To be illustrated.Therefore, before executing object detection method, the storage unit 110 of articles detection system 100 has been stored in advance
It include the video streaming of multiple images acquired in bicycle register, wherein the content of video streaming is driving recorder institute
The Chinese herbaceous peony image of shooting.
Referring to Fig. 1 and Fig. 2, image collection module 120 is read in video streaming from storage unit 110 to be corresponded to
First image of first time point, the second image corresponding to the second time point and the third figure corresponding to third time point
Picture, it is assumed that first time point t1Earlier than the second time point t2, the second time point t2Earlier than third time point t3, then first time point
It is less than the time difference (step S201) of first time point and third time point with the time difference at the second time point.In other words, first
Time point t1, the second time point t2And third time point t3The relationship of inequality (1) will be met:
|t1-t2|≤|t1-t3| inequality (1)
Then, feature search module 130 searches at least one characteristic point (step S203) in the first image.In detail,
The calculation method (Feature Detection Algorithm) that feature search module 130 is detected according to existing characteristic point is searched
The characteristic point in the first image is sought, to recognize the shape or structure of object appeared in the first image.And feature search module
130 characteristic points searched are made of a 1x1 picture element or multiple picture elements.For example, in the present embodiment, feature
The object feature to be searched of search module 130 can be vehicles or pedestrians, and the characteristic point is to constitute vehicles or pedestrians
Edge contour, angle point or block etc., the present invention does not limit herein.
In the present embodiment, feature search module 130 can utilize histograms of oriented gradients (Histogram of
Oriented Gradient, abbreviation HOG) search the characteristic point in the first image.Feature search module 130 is first by the first figure
As being divided into multiple pane locations (cell), and by the gradient value of pixel each in pane location and with gradient direction come
As the foundation of identification characteristic point.This mode is stronger for the object identification capability with obvious boundary, and can avoid part
Modification or small-sized foreign matter are influenced for caused by image, and it is upper to be suitable for the application such as pedestrian detection.
In another embodiment, the calculation mode of characteristic point detection may, for example, be edge detection (Edge Detection)
Algorithm is searched the boundary point of the first objects in images and background by the variation of grayscale, or detects (Corner using angle
Detection) algorithm, by the way that there is point corresponding to maximum curvature on local bright spot or dim spot, line segment terminal, curve, with
Search the angle point of the first objects in images.Feature search module 130 recognizes the profile that all boundary points or angle point are constituted
No is the feature of e.g. vehicle or pedestrian.In another embodiment, feature search module 130 can also be detected according to block
(Blob Detection) algorithm, detected using color and grayscale with the discrepant block of surrounding, to recognize this block
Whether be e.g. vehicle or pedestrian feature.The present invention does not limit in the calculation method of characteristic point detection.
When feature search module 130 can not search characteristic point in the first image, articles detection system 100 will terminate
Object testing process in first image.And when feature search module 130 hunts out at least one characteristic point in the first image
When, computing module 140 judges that the characteristic point is located at the detection zone or remote detection zone (step of short distance
S205).The detection zone of short distance in this is that the position (i.e. driving recorder) of distance the first image of shooting is closer
Region, and detection zone is the farther away region in position of distance the first image of shooting at a distance.
For example, Fig. 3 A is the schematic diagram of the first image shown by an embodiment according to the present invention.
A referring to figure 3., articles detection system 100 is from after the first image 310 of acquirement of storage unit 110, feature search module
130 can search characteristic point from the first image 310.In the present embodiment, the characteristic point is the feature of composition object 30, and this
Object 30 is vehicle that is positive and coming.Computing module 140 is to judge the characteristic point for constituting object 30 positioned at remote
The detection zone 312 of detection zone 314 or short distance.And in the first image 310 of Fig. 3 A, the characteristic point is then located at
Remote detection zone 314.
When computing module 140 judges that the characteristic point is located at the detection zone of short distance, according to the first image and
Two images calculate the motion-vector (step S207) of each characteristic point；When computing module 140 judges that the characteristic point is located at far
When the detection zone of distance, according to the first image and the second image one of them and third image, each feature is calculated
The motion-vector (step S209) of point.In detail, when the characteristic point is located at the detection zone of short distance, characteristic point is constituted
Profile or block it is shared in each image area it is larger.Under same time and movement speed, the characteristic point
Displacement it is more significant, therefore computing module 140 can be calculated respectively using time difference lesser first image and the second image
The motion-vector of the characteristic point, to reduce the search time of characteristic point and reach processing (Real-time in real time
Processing) the effect of.On the other hand, when the characteristic point is located at remote detection zone, what characteristic point was constituted
Profile or block area shared in each image are smaller.Under same time and movement speed, the characteristic point
Be displaced it is less significant, therefore computing module 140 can be calculated using time difference biggish first image and third image it is each described
The motion-vector of characteristic point, to more precisely compute the motion-vector for being located at characteristic point described in a distant place.
In another embodiment, as first time point t1, the second time point t2And third time point t3Meet inequality
(1) and when the relationship of following inequality (1.5):
|t1-t2|≤|t2-t3| inequality (1.5)
When computing module 140 judges that the characteristic point is located at the detection zone of short distance, according to the first image and
Two images calculate the motion-vector of each characteristic point；When computing module 140 judges that the characteristic point is located at detection at a distance
When region, then it can be calculated each according to time difference biggish first image and third image or the second image and third image
The motion-vector of the characteristic point.
Fig. 3 B is the schematic diagram of step S209 in object detection method shown by an embodiment according to the present invention.
B referring to figure 3., it is assumed that when image collection module 120 is read in video streaming from storage unit 110 corresponding to first
Between point t1The first image 310, correspond to the second time point t2The second image 320 and correspond to third time point t3?
Three images 330.If Fig. 3 A is sayed, the first image 310 includes the detection zone 314 and remote detection zone of short distance
312.When computing module 140 judges that preceding feature point is located at the detection zone 314 of short distance, computing module 140 then can basis
The detection zone 324 of the detection zone 314 of the short distance of first image 310 and the short distance of the second image 320 calculates each institute
State the motion-vector of characteristic point.
On the other hand, when computing module 140 judges that preceding feature point is located at remote detection zone 312, in this reality
It applies in example, computing module 140 then can be according to the remote of the remote detection zone 312 of the first image 310 and third image 330
The detection zone 332 of distance calculates the motion-vector of each characteristic point.In another embodiment, computing module 140 also can root
According to the remote detection zone (not shown) of the second image 320 and the remote detection zone 332 of third image 330,
Calculate the motion-vector of each characteristic point.
The example of table two is to detect to be located at the closely speed with remote moving article respectively using two groups of different images
Spend estimation error:
Table two
In every image, the actual displacement of real space corresponding to the actual displacement picture element of each picture element is different.Herein
In example, it is assumed that in short distance, actual displacement picture element corresponding to each actual displacement picture element is 50cm；When remote,
Actual displacement picture element corresponding to each actual displacement picture element is 100cm.When the characteristic point is located at remote detection zone
When, in the case where same 1 pixel ratio is to error, the present invention is using the first image and third image at a distance with speed
It with the velocity error of table one is 40% to compare that the velocity error that the moving article of (5cm/ms, 5cm/ms) is estimated, which is 20%, is improved
2 times.
In addition, computing module 140 can utilize pixel characteristic Comparison Method (Pixel-matching Algorithm) or light
Stream method (Optical Flow Algorithm) calculates the motion-vector of each characteristic point.
In detail, it is assumed that feature search module 130 is located at the detection zone of short distance from the characteristic point that the first image searches
Behind domain, computing module 140 utilizes pixel characteristic Comparison Method, to find out the spy for being similar to the first image 310 in the second image 320
Sign point.In the present embodiment, computing module 140 can correspond to the place of the characteristic point of the first image 310 in the second image 320
Centered on position, it is located at the second figure from the neighbouring picture element search of surrounding using features such as the colors of characteristic point of the first image 310
As 320 characteristic point.Then, computing module 140 using the characteristic point in the first image 310 and the second image 320 institute
Motion-vector of each characteristic point in continuous time is calculated in position.
In another embodiment, it is assumed that feature search module 130 is located at closely from the characteristic point that the first image searches
Detection zone, and after computing module 140 searches corresponding characteristic point in the second image, using institute each in optical flow method
The invariance of characteristic point brightness value in different images is stated, to calculate its motion-vector.
Later, confirmation module 150 is according to the motion-vector of each characteristic point, and confirmation moving article is with the presence or absence of (step
S211).In detail, mobile speed of the computing module after the motion-vector for obtaining each characteristic point, when using shooting image
Degree, frame per second (Frame Rate) and motion-vector calculate the displacement for respectively stating characteristic point, and to calculate each characteristic point
Movement speed.Whether the movement speed for judging each characteristic point is greater than a speed threshold value by confirmation module 150.It is in this
Movement speed when speed threshold value then must be in view of shooting image.In other words, confirmation module 150 can determine whether each characteristic point
Movement speed and shooting image when movement speed between difference, if be greater than speed threshold value.In one embodiment, position
It can with speed threshold value corresponding to the characteristic point positioned at remote detection zone in the characteristic point of the detection zone of short distance
With difference, to reach more accurate judgement.
When confirmation module 150 judges the movement speed of at least one characteristic point greater than speed threshold value, described in representative
Characteristic point is the feature of the moving article such as other driving vehicles or pedestrian, and confirmation module 150 will confirm that moving article is deposited
?.Conversely, representing the feature that the characteristic point is the e.g. static object such as road, building or roadside parked vehicles, really
Recognizing module 150 will confirm that moving article is not present.
In one embodiment, articles detection system 100 further includes alarm module (not shown), when confirmation module 150 confirms
In the presence of moving article, alarm module can provide such as one or a combination set of prompt text, the sound or light, be driven with reminding
The presence of moving article.
Fig. 4 is the flow chart according to object detection method shown by second embodiment of the invention.Referring to figure 4., this reality
The method for applying example is also applied for the articles detection system 100 of Fig. 1.It in the present embodiment, similarly will be with articles detection system 100
It is coupled to for driving recorder to be illustrated.Therefore, before executing object detection method, articles detection system 100
Storage unit 110 has stored the video streaming acquired in preparatory bicycle register comprising multiple images, wherein video streaming
Content be Chinese herbaceous peony image captured by driving recorder.
Referring to Fig. 1 and Fig. 4, image collection module 120 is read in video streaming from storage unit 110 to be corresponded to
First image of first time point, the second image corresponding to the second time point, the third image corresponding to third time point with
And the 4th image corresponding to the 4th time point, it is assumed that first time point t1Earlier than the second time point t2, the second time point t2Earlier than
Third time point t3, third time point t3Earlier than the 4th time point t4, and first time point and the time difference at the second time point are small
In or equal to third time point and the 4th time point time difference (step S401).In other words, first time point t1, the second time
Point t2, third time point t3And the 4th time point t4The relationship of inequality (2) will be met:
|t1-t2|≤|t3-t4| inequality (2)
Then, feature search module 130 searches at least one characteristic point (step S403) in the first image.When feature is searched
When characteristic point can not be searched in the first image by seeking module 130, articles detection system 100 will terminate the object in the first image
Testing process.And when feature search module 130 hunts out at least one characteristic point in the first image, computing module 140 is sentenced
The characteristic point of breaking is located at the detection zone or remote detection zone (step S405) of short distance.When computing module 140
When judging that the characteristic point is located at the detection zone of short distance, according to the first image and the second image, each feature is calculated
The motion-vector (step S407) of point；When computing module 140 judges that the characteristic point is located at remote detection zone, according to
Third image and the 4th image calculate the motion-vector (step S409) of each characteristic point.Later, confirmation module 150
According to the motion-vector of each characteristic point, confirm that moving article whether there is (step S411).
The embodiment of the embodiment and Fig. 2 of Fig. 4 the difference is that for amount of images needed for calculating motion-vector.So
And the concept that the two embodiment is used is similar, when characteristic point is located at the detection zone of short distance, computing module 140 is utilized
Time difference lesser two images calculate the motion-vector of characteristic point；When characteristic point is located at remote detection zone, meter
Calculate the motion-vector that module 140 calculates characteristic point using time difference biggish two images.Above-mentioned steps S403~step
The detailed content of S411, those skilled in the art can refer to the related description of Fig. 1 to Fig. 3 B and analogize it, and details are not described herein.
In conclusion articles detection system provided by the present invention, is calculated using image different in video streaming
The motion-vector of characteristic point.When characteristic point is located at remote detection zone, counted using time difference biggish two images
The motion-vector of characteristic point is calculated, to shorten the time of object detection；When characteristic point is located at remote detection zone, when utilization
Between biggish two images of difference calculate the motion-vector of characteristic point, to improve the accuracy rate of object detection.In addition to this, this hair
It is bright to calculate the displacement of characteristic point also by the motion-vector of characteristic point, so that the movement speed of characteristic point is calculated, to confirm movement
The presence of object.Therefore, articles detection system of the invention can more accurately grasp the information of moving article, and can reach reality
When image procossing efficiency, to apply on the consumer electrical product of low cost, the enhancing present invention in practical applications suitable
The property used.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into
Row equivalent replacement；And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme.
Claims (8)
1. a kind of articles detection system characterized by comprising
The video streaming that storage unit, at least storage include multiple images；
Image collection module, couples the storage unit, and responsible read in the video streaming being stored in the storage unit corresponds to
The first image in first time point, the second image corresponding to the second time point and the third figure corresponding to third time point
Picture, wherein the time difference at the first time point and second time point is less than or equal to second time point and the third time point
Time difference；
Feature search module searches at least one characteristic point in first image；
Computing module judges that the characteristic point is located at the detection zone or remote detection zone of short distance, as the spy
When sign point is located at the detection zone of the short distance, according to first image and second image, each characteristic point is calculated
Motion-vector, when the characteristic point is located at the remote detection zone, wherein according to first image and second image
One and the third image, calculate the motion-vector of each characteristic point；And
Confirmation module confirms that a moving article whether there is according to the motion-vector of each characteristic point.
2. articles detection system according to claim 1, which is characterized in that the spy that this feature search module is searched
Sign point is made of a 1x1 picture element or multiple picture elements.
3. articles detection system according to claim 1, which is characterized in that the computing module is to utilize multiple pixel characteristics
Comparison Method or optical flow method calculate the motion-vector of each characteristic point.
4. articles detection system according to claim 1, which is characterized in that the computing module is according to each characteristic point
Motion-vector calculates the displacement of each characteristic point, to calculate the movement speed of each characteristic point, and the confirmation module
Judge whether the movement speed of each characteristic point is greater than the speed threshold value with each characteristic point distance dependent, when the confirmation mould
When block judges that the movement speed of at least one characteristic point is greater than the speed threshold value, which confirms that the moving article is deposited
?.
5. a kind of articles detection system characterized by comprising
The video streaming that storage unit, at least storage include multiple images；
Image collection module, couples the storage unit, and responsible read in the video streaming being stored in the storage unit corresponds to
The first image in first time point, the second image corresponding to the second time point, the third image corresponding to third time point
And the 4th image corresponding to the 4th time point, wherein the time difference at the first time point and second time point is less than or waits
In the time difference at the third time point and the 4th time point；
Feature search module searches multiple characteristic points in first image；
Computing module judges that the characteristic point is located at the detection zone or remote detection zone of short distance, as the spy
When sign point is located at the detection zone of the short distance, according to first image and second image, each characteristic point is calculated
Motion-vector, according to the third image and the 4th image, is calculated when the characteristic point is located at the remote detection zone
The motion-vector of each characteristic point；And confirmation module confirms a moving article according to the motion-vector of each characteristic point
It whether there is.
6. articles detection system according to claim 5, which is characterized in that the spy that this feature search module is searched
Sign point is made of a 1x1 picture element or multiple picture elements.
7. articles detection system according to claim 5, which is characterized in that the computing module calculates each characteristic point
The calculation method of motion-vector is to utilize multiple pixel characteristic Comparison Methods or optical flow method.
8. articles detection system according to claim 5, which is characterized in that the computing module is according to each characteristic point
Motion-vector obtains the displacement of each characteristic point, to calculate the movement speed of each characteristic point, and the confirmation module
Judge whether the movement speed of each characteristic point is greater than the speed threshold value with each characteristic point distance dependent, when the confirmation mould
When block judges that the movement speed of at least one characteristic point is greater than the speed threshold value, which confirms that the moving article is deposited
?.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410129820.0A CN104978547B (en) | 2014-04-01 | 2014-04-01 | Articles detection system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410129820.0A CN104978547B (en) | 2014-04-01 | 2014-04-01 | Articles detection system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104978547A CN104978547A (en) | 2015-10-14 |
| CN104978547B true CN104978547B (en) | 2019-06-18 |
Family
ID=54275037
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410129820.0A Expired - Fee Related CN104978547B (en) | 2014-04-01 | 2014-04-01 | Articles detection system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104978547B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001126184A (en) * | 1999-10-29 | 2001-05-11 | Matsushita Electric Ind Co Ltd | Automatic license plate recognizing device and vehicle speed measuring method |
| CN103413325A (en) * | 2013-08-12 | 2013-11-27 | 大连理工大学 | Vehicle speed identification method based on vehicle body feature point positioning |
-
2014
- 2014-04-01 CN CN201410129820.0A patent/CN104978547B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001126184A (en) * | 1999-10-29 | 2001-05-11 | Matsushita Electric Ind Co Ltd | Automatic license plate recognizing device and vehicle speed measuring method |
| CN103413325A (en) * | 2013-08-12 | 2013-11-27 | 大连理工大学 | Vehicle speed identification method based on vehicle body feature point positioning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104978547A (en) | 2015-10-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Niu et al. | Robust lane detection using two-stage feature extraction with curve fitting | |
| Wu et al. | Lane-mark extraction for automobiles under complex conditions | |
| Huang et al. | Vehicle detection and inter-vehicle distance estimation using single-lens video camera on urban/suburb roads | |
| Lee et al. | Stereo vision–based vehicle detection using a road feature and disparity histogram | |
| TWI595450B (en) | Object detection system | |
| US9965677B2 (en) | Method and system for OCR-free vehicle identification number localization | |
| JP2013164853A (en) | System and method for traffic signal recognition | |
| Arenado et al. | Monovision‐based vehicle detection, distance and relative speed measurement in urban traffic | |
| US20090110286A1 (en) | Detection method | |
| Lin et al. | Lane departure and front collision warning using a single camera | |
| CN106951898B (en) | Vehicle candidate area recommendation method and system and electronic equipment | |
| Tanaka et al. | Vehicle Detection Based on Perspective Transformation Using Rear‐View Camera | |
| Hajimolahoseini et al. | Robust vehicle tracking algorithm for nighttime videos captured by fixed cameras in highly reflective environments | |
| CN109960959B (en) | Method and apparatus for processing image | |
| Jang et al. | Pothole detection using spatio‐temporal saliency | |
| Hsieh et al. | Vehicle make and model recognition using symmetrical SURF | |
| CN113780480A (en) | Method for constructing multi-target detection and category identification model based on YOLOv5 | |
| Phelawan et al. | A new technique for distance measurement of between vehicles to vehicles by plate car using image processing | |
| CN117392638A (en) | Open object class sensing method and device for serving robot scene | |
| CN112308801A (en) | Road traffic tracking management method and system based on big data image acquisition | |
| Abbas et al. | Recognition of vehicle make and model in low light conditions | |
| Munajat et al. | Vehicle detection and tracking based on corner and lines adjacent detection features | |
| CN104978547B (en) | Articles detection system | |
| US20220318456A1 (en) | Simulation method based on three-dimensional contour, storage medium, computer equipment | |
| Yakimov | Traffic signs detection using tracking with prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190618 Termination date: 20200401 |