CN112634327A - Tracking method based on YOLOv4 model - Google Patents
Tracking method based on YOLOv4 model Download PDFInfo
- Publication number
- CN112634327A CN112634327A CN202011516516.3A CN202011516516A CN112634327A CN 112634327 A CN112634327 A CN 112634327A CN 202011516516 A CN202011516516 A CN 202011516516A CN 112634327 A CN112634327 A CN 112634327A
- Authority
- CN
- China
- Prior art keywords
- picture
- target
- current
- frame
- storing
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 20
- 238000001514 detection method Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 5
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 4
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000001681 protective effect Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
Images
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
- G06T2207/20081—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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
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 multiple target tracking method based on a YOLOv4 model, which is used for detecting and tracking a target object in a video by utilizing the YOLOv4 model, has better tracking and de-duplication effects and is suitable for the multiple target tracking requirement in a complex scene video.
Description
Technical Field
The invention relates to a tracking method based on a YOLOv4 model.
Background
In a single picture, the deep learning model Yolov4 model can well identify and locate the target.
However, in video processing, there are many frames of pictures in the period from appearance to disappearance of the target, and therefore, many repetitions of the detected target appear. For this case, the repetitive problem can be solved to some extent using the tracking algorithm Deep SORT.
In the fields of intelligent transportation and the like, scenes of video data are complex, targets are many, large-area shielding exists, and the situations of target tracking failure and target loss still occur when Deep SORT is used according to a conventional method, so that a large number of repeated targets appear, and the detection efficiency is seriously influenced.
Therefore, it is urgently needed to provide a better method for solving the technical problem.
Disclosure of Invention
The invention aims to provide a tracking method based on a YOLOv4 model, which is used for detecting and tracking a target object in a video by utilizing the YOLOv4 model, has better tracking and de-duplication effects and meets the multi-target tracking requirement in a complex scene video.
In order to achieve the above object, the present invention provides a tracking method based on YOLOv4 model, including:
step 1, inputting an original picture into YOLOv4, outputting a detected target picture by YOLOv4, and numbering the target picture respectively to form an ID number; if the frame is the first frame, directly giving an ID (identity) and storing the picture and corresponding coordinate information and width and height;
step 2, if the frame is not the first frame, calculating the IOU intersection ratio, if the calculated value is larger than the threshold value, giving a new ID, storing information and storing the picture into newfile; if the calculated value is smaller than the threshold value, calculating the similarity, if the calculated value reaches the threshold value, determining that the picture of the previous frame and the detected picture of the current frame are the same target, deleting the picture of the previous frame and storing the picture of the current target into the newfile; then, traversing the next target of the current frame, and calculating by using the same method; if the calculated similarity value exceeds the threshold value, defining the current target as a new target, giving a new ID, and then independently storing the graph, the corresponding ID and the information;
step 3, when the target of the current frame is detected and the previous frame is found to have the undetected target, determining that the target is lost, and independently storing the graph, the corresponding ID and the information into the lossfile;
step 4, traversing the newly added picture in newfile and the lost picture in lossfile, and carrying out similarity calculation on the newly added picture and the lost picture in lossfile to judge whether the newly added picture and the lost picture are the same target; at this time, a threshold value is set through the ID or the frame number, whether the current ID is larger than the set ID or not is compared, if the current ID is larger than the set ID, the current ID is directly deleted, and if the current ID is not larger than the set ID, the current ID is reserved;
and 5, performing continuous cyclic detection according to the steps 1 to 4.
According to the technical scheme, the method is not used for carrying out the de-duplication on the basis of YOLOv4+ DeepsORT, but directly carries out the de-duplication on the basis of YOLOv4, and directly outputs the de-duplication to the method when the YOLOv4 detects the target object, so that the method can replace the DeepsORT and has better tracking and de-duplication effects.
Additional features and advantages of the invention will be set forth in the detailed description which follows.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
fig. 1 is a schematic flow chart of a tracking method based on the YOLOv4 model in the present invention.
Detailed Description
The following detailed description of embodiments of the invention refers to the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating the present invention, are given by way of illustration and explanation only, not limitation.
In the present invention, unless otherwise specified, the directional terms included in the terms merely represent the directions of the terms in a conventional use state or are colloquially known by those skilled in the art, and should not be construed as limiting the terms.
Referring to fig. 1, the present invention provides a tracking method based on YOLOv4 model, including:
step 1, inputting an original picture into YOLOv4, outputting a detected target picture by YOLOv4, and numbering the target picture respectively to form an ID number; if the frame is the first frame, directly giving an ID (identity) and storing the picture and corresponding coordinate information and width and height;
step 2, if the frame is not the first frame, calculating the IOU intersection ratio, if the calculated value is larger than the threshold value, giving a new ID, storing information and storing the picture into newfile; if the calculated value is smaller than the threshold value, calculating the similarity, if the calculated value reaches the threshold value, determining that the picture of the previous frame and the detected picture of the current frame are the same target, deleting the picture of the previous frame and storing the picture of the current target into the newfile; then, traversing the next target of the current frame, and calculating by using the same method; if the calculated similarity value exceeds the threshold value, defining the current target as a new target, giving a new ID, and then independently storing the graph, the corresponding ID and the information;
step 3, when the target of the current frame is detected and the previous frame is found to have the undetected target, determining that the target is lost, and independently storing the graph, the corresponding ID and the information into the lossfile;
step 4, traversing the newly added picture in newfile and the lost picture in lossfile, and carrying out similarity calculation on the newly added picture and the lost picture in lossfile to judge whether the newly added picture and the lost picture are the same target; at this time, a threshold value is set through the ID or the frame number, whether the current ID is larger than the set ID or not is compared, if the current ID is larger than the set ID, the current ID is directly deleted, and if the current ID is not larger than the set ID, the current ID is reserved;
and 5, performing continuous cyclic detection according to the steps 1 to 4.
Specifically, after an original picture is input into YOLOv4, YOLOv4 outputs detected target pictures, after the target pictures are taken, the original picture is respectively numbered, namely, an ID number (the first frame is all numbered and the target pictures are stored while the center coordinates (relative to a large picture) and the width and the height of the target pictures are stored), when a new frame of picture is obtained, a corresponding small detection target is output after the YOLO detection, then the corresponding center coordinates and the width and the height are respectively obtained, at this time, the target picture of the current frame is traversed, the values of the detection frame of the current frame and the target detection frame of the previous frame are obtained by an IOU (cross-over ratio) method, whether the candidate target is a current frame is determined by the value of the IOU, then, the similarity algorithm is used for calculation, if the similarity reaches a certain threshold value, the current detection target of the current frame appears in the previous frame, at this time, the picture of the current frame is stored to replace the picture of the previous frame, the ID remains unchanged and then the next target of the current frame is traversed, with the same aspect being calculated.
If the target of the current frame is found, the target of the previous frame is found through the IOU, or the similarity calculation exceeds a threshold value, the current target is defined as a new target, a new ID is given, and then the graph and the corresponding ID and information are independently stored.
When the target of the current frame is detected, and the previous frame is found to have the undetected target, the target is determined to be lost, and the graph, the ID and the corresponding information are independently stored.
In this case, the target may be out of the view of the video due to occlusion in the lost target, or the target may be out of the view of the video due to occlusion in the newly added target, and the target may appear again or a new target may enter the view of the video. Therefore, at this time, it is only necessary to calculate the similarity between the newly added pictures and the lost pictures to determine whether the pictures are the same target, but here, a threshold needs to be set through the ID (or the number of frames), that is, when the pictures are far away from the current frame, the pictures are saved, and the other pictures are discarded, so that the number of the lost pictures and the number of the newly added pictures can be controlled.
And then carrying out continuous cycle detection.
Through the technical scheme, the method is not used for carrying out de-duplication on the basis of YOLOv4+ DeepsORT, but directly carries out de-duplication on the basis of YOLOv4, and directly outputs the de-duplication to the method when a target object is detected by YOLOv4, so that the method can replace the DeepsORT and has better tracking and de-duplication effects.
The preferred embodiments of the present invention have been described in detail with reference to the accompanying drawings, however, the present invention is not limited to the specific details of the above embodiments, and various simple modifications can be made to the technical solution of the present invention within the technical idea of the present invention, and these simple modifications are within the protective scope of the present invention.
It should be noted that the various technical features described in the above embodiments can be combined in any suitable manner without contradiction, and the invention is not described in any way for the possible combinations in order to avoid unnecessary repetition.
In addition, any combination of the various embodiments of the present invention is also possible, and the same should be considered as the disclosure of the present invention as long as it does not depart from the spirit of the present invention.
Claims (1)
1. A tracking method based on a YOLOv4 model is characterized by comprising the following steps:
step 1, inputting an original picture into YOLOv4, outputting a detected target picture by YOLOv4, and numbering the target picture respectively to form an ID number; if the frame is the first frame, directly giving an ID (identity) and storing the picture and corresponding coordinate information and width and height;
step 2, if the frame is not the first frame, calculating the IOU intersection ratio, if the calculated value is larger than the threshold value, giving a new ID, storing information and storing the picture into newfile; if the calculated value is smaller than the threshold value, calculating the similarity, if the calculated value reaches the threshold value, determining that the picture of the previous frame and the detected picture of the current frame are the same target, deleting the picture of the previous frame and storing the picture of the current target into the newfile; then, traversing the next target of the current frame, and calculating by using the same method; if the calculated similarity value exceeds the threshold value, defining the current target as a new target, giving a new ID, and then independently storing the graph, the corresponding ID and the information;
step 3, when the target of the current frame is detected and the previous frame is found to have the undetected target, determining that the target is lost, and independently storing the graph, the corresponding ID and the information into the lossfile;
step 4, traversing the newly added picture in newfile and the lost picture in lossfile, and carrying out similarity calculation on the newly added picture and the lost picture in lossfile to judge whether the newly added picture and the lost picture are the same target; at this time, a threshold value is set through the ID or the frame number, whether the current ID is larger than the set ID or not is compared, if the current ID is larger than the set ID, the current ID is directly deleted, and if the current ID is not larger than the set ID, the current ID is reserved;
and 5, performing continuous cyclic detection according to the steps 1 to 4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011516516.3A CN112634327A (en) | 2020-12-21 | 2020-12-21 | Tracking method based on YOLOv4 model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011516516.3A CN112634327A (en) | 2020-12-21 | 2020-12-21 | Tracking method based on YOLOv4 model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112634327A true CN112634327A (en) | 2021-04-09 |
Family
ID=75320219
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011516516.3A Pending CN112634327A (en) | 2020-12-21 | 2020-12-21 | Tracking method based on YOLOv4 model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112634327A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113177931A (en) * | 2021-05-19 | 2021-07-27 | 北京明略软件系统有限公司 | Method and device for detecting and tracking key component |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018058595A1 (en) * | 2016-09-30 | 2018-04-05 | 富士通株式会社 | Target detection method and device, and computer system |
CN108985162A (en) * | 2018-06-11 | 2018-12-11 | 平安科技(深圳)有限公司 | Object real-time tracking method, apparatus, computer equipment and storage medium |
CN110717403A (en) * | 2019-09-16 | 2020-01-21 | 国网江西省电力有限公司电力科学研究院 | Face multi-target tracking method |
WO2020082258A1 (en) * | 2018-10-24 | 2020-04-30 | 深圳鲲云信息科技有限公司 | Multi-objective real-time tracking method and apparatus, and electronic device |
CN111429483A (en) * | 2020-03-31 | 2020-07-17 | 杭州博雅鸿图视频技术有限公司 | High-speed cross-camera multi-target tracking method, system, device and storage medium |
WO2020164282A1 (en) * | 2019-02-14 | 2020-08-20 | 平安科技(深圳)有限公司 | Yolo-based image target recognition method and apparatus, electronic device, and storage medium |
CN111582062A (en) * | 2020-04-21 | 2020-08-25 | 电子科技大学 | Re-detection method in target tracking based on YOLOv3 |
CN111626277A (en) * | 2020-08-03 | 2020-09-04 | 杭州智诚惠通科技有限公司 | Vehicle tracking method and device based on over-station inter-modulation index analysis |
CN112016445A (en) * | 2020-08-27 | 2020-12-01 | 重庆科技学院 | Monitoring video-based remnant detection method |
-
2020
- 2020-12-21 CN CN202011516516.3A patent/CN112634327A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018058595A1 (en) * | 2016-09-30 | 2018-04-05 | 富士通株式会社 | Target detection method and device, and computer system |
CN108985162A (en) * | 2018-06-11 | 2018-12-11 | 平安科技(深圳)有限公司 | Object real-time tracking method, apparatus, computer equipment and storage medium |
WO2020082258A1 (en) * | 2018-10-24 | 2020-04-30 | 深圳鲲云信息科技有限公司 | Multi-objective real-time tracking method and apparatus, and electronic device |
WO2020164282A1 (en) * | 2019-02-14 | 2020-08-20 | 平安科技(深圳)有限公司 | Yolo-based image target recognition method and apparatus, electronic device, and storage medium |
CN110717403A (en) * | 2019-09-16 | 2020-01-21 | 国网江西省电力有限公司电力科学研究院 | Face multi-target tracking method |
CN111429483A (en) * | 2020-03-31 | 2020-07-17 | 杭州博雅鸿图视频技术有限公司 | High-speed cross-camera multi-target tracking method, system, device and storage medium |
CN111582062A (en) * | 2020-04-21 | 2020-08-25 | 电子科技大学 | Re-detection method in target tracking based on YOLOv3 |
CN111626277A (en) * | 2020-08-03 | 2020-09-04 | 杭州智诚惠通科技有限公司 | Vehicle tracking method and device based on over-station inter-modulation index analysis |
CN112016445A (en) * | 2020-08-27 | 2020-12-01 | 重庆科技学院 | Monitoring video-based remnant detection method |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113177931A (en) * | 2021-05-19 | 2021-07-27 | 北京明略软件系统有限公司 | Method and device for detecting and tracking key component |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107633526B (en) | Image tracking point acquisition method and device and storage medium | |
CN105894464B (en) | A kind of medium filtering image processing method and device | |
CN111652902B (en) | Target tracking detection method, electronic equipment and device | |
WO2014136623A1 (en) | Method for detecting and tracking objects in sequence of images of scene acquired by stationary camera | |
KR102265315B1 (en) | Method and apparatus for estimating image optical flow | |
CN110648363A (en) | Camera posture determining method and device, storage medium and electronic equipment | |
CN111310759B (en) | Target detection inhibition optimization method and device for dual-mode cooperation | |
CN108898148B (en) | Digital image corner detection method, system and computer readable storage medium | |
CN110009662B (en) | Face tracking method and device, electronic equipment and computer readable storage medium | |
TW201917696A (en) | Object detection and tracking method and system | |
CN115423846A (en) | Multi-target track tracking method and device | |
CN114049383B (en) | Multi-target tracking method and device and readable storage medium | |
JP2006244074A (en) | Moving object close-up frame detection method and program, storage medium storing program, moving object close-up shot detection method, moving object close-up frame or shot detection method and program, and storage medium storing program | |
CN112634327A (en) | Tracking method based on YOLOv4 model | |
CN109961516B (en) | Surface acquisition method, device and non-transitory computer readable recording medium | |
CN116330658B (en) | Target tracking method, device and system based on depth image and image pickup equipment | |
CN110009683B (en) | Real-time on-plane object detection method based on MaskRCNN | |
KR102336284B1 (en) | Moving Object Detection Method and System with Single Camera | |
CN111860161B (en) | Target shielding detection method | |
US8179967B2 (en) | Method and device for detecting movement of an entity provided with an image sensor | |
KR101915402B1 (en) | Method for matching feature of image by progressive graph optimization | |
JPH03204783A (en) | Image tracking device | |
CN112634332A (en) | Tracking method based on YOLOv4 model and DeepsORT model | |
CN114677625B (en) | Object detection method, device, apparatus, storage medium, and program product | |
CN114115640B (en) | Icon determination method, device, equipment and storage medium |
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 |