CN110097578A - Plastic grains tracking - Google Patents
Plastic grains tracking Download PDFInfo
- Publication number
- CN110097578A CN110097578A CN201910385845.XA CN201910385845A CN110097578A CN 110097578 A CN110097578 A CN 110097578A CN 201910385845 A CN201910385845 A CN 201910385845A CN 110097578 A CN110097578 A CN 110097578A
- Authority
- CN
- China
- Prior art keywords
- plastic grains
- target
- plastic
- collection
- tracking
- 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
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
- G06T7/248—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving reference images or patches
-
- 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
- G06T2207/10016—Video; Image sequence
-
- 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
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- 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
- G06T2207/30108—Industrial image inspection
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of plastic grains trackings, obtain motion image sequence of the plastic grains in the detection visual field, plastic grains target detection is carried out respectively to every frame image in the motion image sequence, for the plastic grains target collection of previous frame image, extract the central area pixel collection of each plastic grains target, its matching characteristic point in a later frame image is sought based on optical flow method, obtain the matching characteristic point set of each central area pixel collection, ballot cost matrix is obtained according to the plastic grains target collection of matching characteristic point set and a later frame image, acquisition enables the highest plastic grains object matching combination of aggregate votes, to obtain the tracking result of plastic grains target in adjacent two field pictures.The present invention tracking problem of intensive plastic grains when colliding during " vibration moves ahead " when can effectively solve plastic grains defects detection, improves the collision robustness of tracking accuracy and tracking.
Description
Technical field
The invention belongs to Machine Vision Detection and target following technical fields, more specifically, are related to a kind of plastic cement
Grain tracking.
Background technique
Raw material product form of the plastic grains as numerous plasthetics, quality testing are the important of its production process
Link.Conventional detection generally using artificial sampling observation mode, i.e., manually presses certain rule extraction part plastic grains, and be molded into mark
Quasi- shape block, then the detection by means of special instrument progress correlated quality index.This is not only time-consuming bothersome, but also is unfavorable in time
It was found that quality problems in plastic grains production process, in some instances it may even be possible to cause it is secondary by the gross do over again or scrap, seriously affect productivity effect.
With the development of computer vision technique, directly the plastic grains of production line are carried out with the method for machine vision scarce
Falling into detection becomes a kind of important solutions.To improve detection efficiency, plastic grain to be checked is (logical by certain mode on production line
Often it is vibration delivery form) continuously enter in the detection visual field of detection device, thus using realizing of Robot Vision to modeling
The real-time defects detection and statistical analysis of glue particle.For the plastic grains object defect on accurate statistics and analysis production line
Situation, just must be comprising to entering the tracking for detecting plastic grains object within the vision in testing process.Currently, target following
Mainstream frame be the tracking frame based on target detection, main thought is passed through based on the object detection results of depth network
Matching is associated to the effective target of consecutive frame image, achievees the purpose that tracking.But since large number of plastic grains exist
Phenomena such as may colliding between each other during detection scene " vibration moves ahead ", beat, roll, the i.e. speed of particle movement
With direction and on-fixed and unanimously, it is easy to appear error hiding in consecutive frame effective target association matching process, leads to tracking accuracy
It is not high.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of plastic grains trackings, can be effective
Tracking of intensive plastic grains when colliding during " vibration moves ahead " is asked when solving plastic grains defects detection
Topic improves the collision robustness of tracking accuracy and tracking.
For achieving the above object, plastic grains tracking of the present invention the following steps are included:
S1: motion image sequence of the plastic grains in the detection visual field is obtained, to every frame figure in the motion image sequence
As carrying out plastic grains target detection respectively, each plastic grains target is surrounded into the coordinate of frame with its target to indicate, remembers phase
The plastic grains target collection that former frame and a later frame image detection obtain in adjacent two field pictures is respectively ApreAnd Acur, included
Plastic grains destination number distinguish M and N；
S2: for the plastic grains target collection A of previous frame imagepre, calculate the corresponding target of each plastic grains target
Surround the central point o of framemCoordinate (im,jm), m=1,2 ..., M obtain centre coordinate set Opre={ o1,o2,...,oM-1,
oM}；For the central point o of each plastic grainsm, extract with central point omCentered on, the rectangle frame that side length is 2d+1, wherein d
It is arranged according to actual needs, remembers the coordinate set of all pixels point in the rectangle frameMade
For the central area pixel collection of m-th of plastic grains target；
For each central area pixel collection pmIn each pixel, it is sought in image I based on optical flow methodcur
In matching characteristic point, by each central area pixel collection pmObtained matching characteristic point set is denoted as qm；
S3: the ballot cost matrix C that one size of setting is M × N, Elements C (m, n) indicate plastic grains target collection
ApreIn m-th of plastic grains target and plastic grains target collection AcurIn n-th of plastic grains target matching poll；It will throw
Ticket cost matrix C is initialized as null matrix；
For the matching characteristic point set q of m-th of plastic grains targetm, judge respectively wherein each pixel whether position
In plastic grains target collection AcurIn the target of n-th plastic grains target surround the inside of frame, n=1,2 ..., N, if
It is a ticket to be counted in the Elements C (m, n) that then coordinate is (m, n) in ballot cost matrix C, even C (m, n)=C (m, n)+1, no
Any operation is not made then；
S4: the ballot cost matrix C obtained according to step S3 is calculated and enables the highest plastic grains object set of aggregate votes
Close ApreAnd AcurIn plastic grains object matching combination, as the matching result of plastic grains target, to obtain phase
The tracking result of plastic grains target in adjacent two field pictures.
Plastic grains tracking of the present invention obtains motion image sequence of the plastic grains in the detection visual field, to the fortune
Every frame image in motion video sequence carries out plastic grains target detection respectively, for the plastic grains object set of previous frame image
It closes, extracts the central area pixel collection of each plastic grains target, it is sought in a later frame image based on optical flow method
Matching characteristic point obtains the matching characteristic point set of each central area pixel collection, according to matching characteristic point set with after
The plastic grains target collection of one frame image obtains ballot cost matrix, and acquisition enables the highest plastic grains object matching of aggregate votes
Combination, to obtain the tracking result of plastic grains target in adjacent two field pictures.The present invention can effectively solve plastic grains
Tracking problem of the intensive plastic grains when colliding during " vibration moves ahead ", improves tracking accuracy when defects detection
With the collision robustness of tracking.
Detailed description of the invention
Fig. 1 is the specific embodiment flow chart of plastic grains tracking of the present invention；
Fig. 2 is plastic grains target detection exemplary diagram in adjacent two field pictures in the present embodiment；
Fig. 3 is in the present embodiment based on the obtained matching characteristic point exemplary diagram of LK pyramid optical flow method；
Fig. 4 is plastic grains object matching result exemplary diagram in the present embodiment.
Specific embodiment
A specific embodiment of the invention is described with reference to the accompanying drawing, preferably so as to those skilled in the art
Understand the present invention.Requiring particular attention is that in the following description, when known function and the detailed description of design perhaps
When can desalinate main contents of the invention, these descriptions will be ignored herein.
Embodiment
Fig. 1 is the specific embodiment flow chart of plastic grains tracking of the present invention.As shown in Figure 1, plastic cement of the present invention
The specific steps of particle tracking include:
S101: plastic grains target detection:
Motion image sequence of the plastic grains in the detection visual field is obtained, to every frame image in the motion image sequence point
Not carry out plastic grains target detection, each plastic grains target is indicated with the coordinate that its target surrounds frame, note adjacent two
Former frame I in frame imagepreWith a later frame image IcurDetecting obtained plastic grains target collection is respectively ApreAnd Acur, wrapped
The plastic grains destination number difference M and N contained.
Fig. 2 is plastic grains target detection exemplary diagram in adjacent two field pictures in the present embodiment.As shown in Fig. 2, this implementation
In the plastic grains target detection of example, plastic grains enter the detection visual field from top, and the detection visual field is divided into deep, shallow two kinds of detections
Background area, the top half for detecting background area is black background, and for detecting lighter colored particles, lower half portion is white back
Scape, for detecting dark particle.When remembering target detection the length and width of image captured by camera be H and W (H:800 in the present embodiment,
W:896), the direction of length and width is respectively y-axis and x-axis direction, respectively indicates in image pixel along image x and y-axis using i and j
The coordinate position in direction.
Plastic grains object detection method is not emphasis of the invention, and specific method can select according to actual needs
It selects.The target detection network model based on deep learning is used in the present embodiment, obtains plastic grains in image captured by camera
Object object detection results.From figure 2 it can be seen that the plastic grains destination number that previous frame image detects in the present embodiment
M=14, the plastic grains destination number N=12 that a later frame image detection obtains.It is plastic cement that target, which surrounds frame, in the present embodiment
The minimum rectangle of grain target surrounds frame.
S102: Optical-flow Feature match point is calculated:
For the plastic grains target collection A of previous frame imagepre, calculate the corresponding target packet of each plastic grains target
The central point o of peripheral framemCoordinate (im,jm), m=1,2 ..., M obtain centre coordinate set Opre={ o1,o2,...,oM-1,
oM}.For the central point o of each plastic grainsm, extract with central point omCentered on, the rectangle frame that side length is 2d+1, wherein d
It is arranged according to actual needs, remembers the coordinate set of all pixels point in the rectangle frameMade
For the central area pixel collection of m-th of plastic grains target, it is clear that p in each setmIn include (2d+1) × (2d+1)
A pixel.
For each central area pixel collection pmIn each pixel, it is sought in a later frame figure based on optical flow method
As IcurIn matching characteristic point, by each central area pixel collection pmObtained matching characteristic point set is denoted as qm。
Optical flow method uses LK pyramid optical flow method in the present embodiment, by configuring Opencv computer vision open source library, leads to
Calling function cv::calcOpticalFlowPyrLK is crossed to calculate matching characteristic point.The parameter of LK pyramid optical flow method are as follows:
Window size is 8, the maximum pyramid number of plies 5, the number of iterations 5, and maximum number of iterations is that 20 search strategies are cv::
TermCriteria::MAX_ITER and cv::TermCriteria::EPS.Fig. 3 is in the present embodiment based on the light stream of LK pyramid
The obtained matching characteristic point exemplary diagram of method.As shown in figure 3, in order to preferably show black and white effect of visualization, to original detection back
Black background in scene area is removed.
S103: ballot cost matrix is calculated:
The ballot cost matrix C that one size is M × N is set, and Elements C (m, n) indicates plastic grains target collection Apre
In m-th of plastic grains target and plastic grains target collection AcurIn n-th of plastic grains target matching poll, that is, match
The matching of degree, poll two plastic grains targets of higher explanation is higher.Ballot cost matrix C is initialized as null matrix.
For the matching characteristic point set q of m-th of plastic grains targetm, judge respectively wherein each pixel whether position
In plastic grains target collection AcurIn the target of n-th plastic grains target surround the inside of frame, n=1,2 ..., N, if
It is a ticket to be counted in the Elements C (m, n) that then coordinate is (m, n) in ballot cost matrix C, even C (m, n)=C (m, n)+1, no
Any operation is not made then.
S104: plastic grains target following result is obtained:
The ballot cost matrix C obtained according to step S103 is calculated and enables the highest plastic grains object set of aggregate votes
Close ApreAnd AcurIn plastic grains object matching combination, as the matching result of plastic grains target, to obtain phase
The tracking result of plastic grains target in adjacent two field pictures.
The highest plastic grains target of aggregate votes is solved using maximum cum rights bipartite graph matching algorithm in the present embodiment
With combination.Maximum cum rights bipartite graph matching algorithm is a kind of common technology, and details are not described herein for specific calculating process.Fig. 4 is
Plastic grains object matching result exemplary diagram in the present embodiment.As shown in figure 4, in previous frame image and a later frame image, sequence
Number identical plastic grains target is same target.Plastic grains target collection ApreIn some plastic grains target in plastic cement
Particle target collection AcurIn matched plastic grains target is not present, then the visual plastic grains target has been moved off detection
The visual field.As plastic grains target collection AcurIn some plastic grains target in plastic grains target collection ApreIn there is no matching
Plastic grains target, then the plastic grains target can be considered as newly into particle.
In the present embodiment, due to plastic grains be enter from top detection the visual field can in order to simplify subsequent tracking process
A horizontal dividing lines to be arranged in the picture, as plastic grains target collection AcurIn some plastic grains target central point
Lower than the horizontal dividing lines, then the plastic grains target is considered as and leaves the detection visual field, by it from plastic grains target collection Acur
Middle deletion, using remaining plastic grains target as the target of next secondary tracking.It the position of horizontal dividing lines should be according to plastic cement
The image temporal interval of motion image sequence captured by the movement velocity and camera of particle is arranged.
Technical effect in order to better illustrate the present invention, using in the present embodiment the detection visual field and tracking, it is right
60 plastic grains consecutive frame images comprising a large amount of collision situations carry out plastic grains tracking and testing, through counting plastic grains mesh
The correct matching rate of target has reached 98.9%, it is seen that the present invention can be effectively applicable to the tracking of plastic grains object, have preferable
Collide robustness.
Although the illustrative specific embodiment of the present invention is described above, in order to the technology of the art
Personnel understand the present invention, it should be apparent that the present invention is not limited to the range of specific embodiment, to the common skill of the art
For art personnel, if various change the attached claims limit and determine the spirit and scope of the present invention in, these
Variation is it will be apparent that all utilize the innovation and creation of present inventive concept in the column of protection.
Claims (5)
1. a kind of plastic grains tracking, which comprises the following steps:
S1: motion image sequence of the plastic grains in the detection visual field is obtained, to every frame image in the motion image sequence point
Not carry out plastic grains target detection, each plastic grains target is indicated with the coordinate that its target surrounds frame, note adjacent two
The plastic grains target collection that former frame and a later frame image detection obtain in frame image is respectively ApreAnd Acur, the modeling that is included
Glue particle destination number distinguishes M and N；
S2: for the plastic grains target collection A of previous frame imagepre, calculate the corresponding target of each plastic grains target and surround
The central point o of framemCoordinate (im,jm), m=1,2 ..., M obtain centre coordinate set Opre={ o1,o2,...,oM-1,oM}；
For the central point o of each plastic grainsm, extract with central point omCentered on, side length be 2d+1 rectangle frame, wherein d according to
Actual needs setting, remembers the coordinate set of all pixels point in the rectangle frameAs m
The central area pixel collection of a plastic grains target；
For each central area pixel collection pmIn each pixel, it is sought in a later frame image based on optical flow method
Matching characteristic point, by each central area pixel collection pmObtained matching characteristic point set is denoted as qm；
S3: the ballot cost matrix C that one size of setting is M × N, Elements C (m, n) indicate plastic grains target collection Apre
In m-th of plastic grains target and plastic grains target collection ApreIn n-th of plastic grains target matching poll；It will ballot
Cost matrix C is initialized as null matrix；
For the matching characteristic point set q of m-th of plastic grains targetm, judge wherein whether each pixel is located at plastic cement respectively
Particle target collection AcurIn the target of n-th plastic grains target surround the inside of frame, n=1,2 ..., N, if it is, enabling C
Otherwise (m, n)=C (m, n)+1 does not make any operation；
S4: the ballot cost matrix C obtained according to step S3 is calculated and enables the highest plastic grains target collection of aggregate votes
ApreAnd AcurIn plastic grains object matching combination, as the matching result of plastic grains target, to obtain adjacent
The tracking result of plastic grains target in two field pictures.
2. plastic grains tracking according to claim 1, which is characterized in that plastic grains are from upper in the step S1
Side enters the detection visual field.
3. plastic grains tracking according to claim 2, which is characterized in that the step S4 further includes following step
It is rapid: a horizontal dividing lines are set in the picture, as plastic grains target collection AcurIn some plastic grains target central point
Lower than the horizontal dividing lines, then the plastic grains target is considered as and leaves the detection visual field, by it from plastic grains target collection Acur
Middle deletion, using remaining plastic grains target as the target of next secondary tracking.
4. plastic grains tracking according to claim 1, which is characterized in that optical flow method uses LK in the step S2
Pyramid optical flow method.
5. plastic grains tracking according to claim 1, which is characterized in that using maximum cum rights in the step S5
Bipartite graph matching algorithm combines to solve the highest plastic grains object matching of aggregate votes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910385845.XA CN110097578B (en) | 2019-05-09 | 2019-05-09 | Plastic particle tracking method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910385845.XA CN110097578B (en) | 2019-05-09 | 2019-05-09 | Plastic particle tracking method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110097578A true CN110097578A (en) | 2019-08-06 |
| CN110097578B CN110097578B (en) | 2021-08-17 |
Family
ID=67447550
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910385845.XA Expired - Fee Related CN110097578B (en) | 2019-05-09 | 2019-05-09 | Plastic particle tracking method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110097578B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103593679A (en) * | 2012-08-16 | 2014-02-19 | 北京大学深圳研究生院 | Visual human-hand tracking method based on online machine learning |
| CN106204640A (en) * | 2016-06-29 | 2016-12-07 | 长沙慧联智能科技有限公司 | A kind of moving object detection system and method |
| CN106845385A (en) * | 2017-01-17 | 2017-06-13 | 腾讯科技（上海）有限公司 | The method and apparatus of video frequency object tracking |
| CN107292911A (en) * | 2017-05-23 | 2017-10-24 | 南京邮电大学 | A kind of multi-object tracking method merged based on multi-model with data correlation |
| CN109146920A (en) * | 2018-06-29 | 2019-01-04 | 西北工业大学 | A kind of method for tracking target that insertion type is realized |
-
2019
- 2019-05-09 CN CN201910385845.XA patent/CN110097578B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103593679A (en) * | 2012-08-16 | 2014-02-19 | 北京大学深圳研究生院 | Visual human-hand tracking method based on online machine learning |
| CN106204640A (en) * | 2016-06-29 | 2016-12-07 | 长沙慧联智能科技有限公司 | A kind of moving object detection system and method |
| CN106845385A (en) * | 2017-01-17 | 2017-06-13 | 腾讯科技（上海）有限公司 | The method and apparatus of video frequency object tracking |
| CN107292911A (en) * | 2017-05-23 | 2017-10-24 | 南京邮电大学 | A kind of multi-object tracking method merged based on multi-model with data correlation |
| CN109146920A (en) * | 2018-06-29 | 2019-01-04 | 西北工业大学 | A kind of method for tracking target that insertion type is realized |
Non-Patent Citations (2)
| Title |
|---|
| 梁立东: "基于视觉跟踪的颗粒计数与缺陷检测系统研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 郭向威: "基于改进光流法的运动目标检测与跟踪研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110097578B (en) | 2021-08-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110136449B (en) | Deep learning-based traffic video vehicle illegal parking automatic identification snapshot method | |
| CN108986064B (en) | People flow statistical method, equipment and system | |
| CN108009543B (en) | License plate recognition method and device | |
| CN107016367B (en) | Tracking control method and tracking control system | |
| US11398084B2 (en) | Method, apparatus and application system for extracting a target feature | |
| CN108985169B (en) | Shop cross-door operation detection method based on deep learning target detection and dynamic background modeling | |
| CN107081755A (en) | A kind of robot monocular vision guides the automatic calibration device of system | |
| CN102706274B (en) | System for accurately positioning mechanical part by machine vision in industrially-structured scene | |
| TW202013252A (en) | License plate recognition system and license plate recognition method | |
| CN111339839B (en) | Intensive target detection metering method | |
| CN104966304A (en) | Kalman filtering and nonparametric background model-based multi-target detection tracking method | |
| CN105701476A (en) | Machine vision-based automatic identification system and method for production line products | |
| CN109196518A (en) | A kind of gesture identification method and device based on high light spectrum image-forming | |
| CN105069429A (en) | People flow analysis statistics method based on big data platform and people flow analysis statistics system based on big data platform | |
| CN104835143A (en) | Rapid projector system parameter calibration method | |
| US20220279131A1 (en) | Camera Assembly and Monitoring Camera | |
| CN106845338B (en) | Pedestrian detection method and system in video stream | |
| CN107609475B (en) | Pedestrian detection false detection extraction method based on light field camera | |
| CN109948725A (en) | Based on address-event representation neural network object detecting device | |
| CN106067031B (en) | Based on artificial mechanism for correcting errors and deep learning network cooperation machine vision recognition system | |
| CN109389085B (en) | Lip language recognition model training method and device based on parameterized curve | |
| CN114140745A (en) | Method, system, device and medium for detecting personnel attributes of construction site | |
| CN107346547A (en) | Real-time foreground extracting method and device based on monocular platform | |
| CN112637519A (en) | Panoramic stitching algorithm for multi-path 4K quasi-real-time stitched video | |
| WO2021036243A1 (en) | Method and apparatus for recognizing lane, and computing device |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210817 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |