CN111428642A - Multi-target tracking algorithm, electronic device and computer readable storage medium - Google Patents
Multi-target tracking algorithm, electronic device and computer readable storage medium Download PDFInfo
- Publication number
- CN111428642A CN111428642A CN202010213267.4A CN202010213267A CN111428642A CN 111428642 A CN111428642 A CN 111428642A CN 202010213267 A CN202010213267 A CN 202010213267A CN 111428642 A CN111428642 A CN 111428642A
- Authority
- CN
- China
- Prior art keywords
- target
- tracking
- algorithm
- position information
- tracking algorithm
- 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
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 63
- 238000001514 detection method Methods 0.000 claims abstract description 31
- 238000001914 filtration Methods 0.000 claims abstract description 9
- 238000004590 computer program Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 abstract description 7
- 238000011160 research Methods 0.000 abstract description 5
- 238000000034 method Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 241000195940 Bryophyta Species 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
- G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The present invention relates to the field of computer vision research. The invention discloses a multi-target tracking algorithm, which comprises the following steps: s1, detecting current frame position information of one or more targets by using a target detection algorithm; s2, predicting predicted position information of a next frame of the one or more targets using a KCF tracker; s3, filtering the predicted position information through the IOU, and then transmitting the filtered predicted position information into a deepsort network for tracking; and S4, updating the depsort network corresponding to the successfully tracked target, updating the KCF tracker at the same time, and otherwise, judging. The multi-target tracking algorithm can effectively eliminate the condition of missing detection in the current target detection algorithm and simultaneously improve the tracking effect of the target tracking algorithm.
Description
Technical Field
The present invention relates to the field of computer vision research, and more particularly to a multi-target tracking algorithm.
Background
Target tracking technology is one of the hot spots in the field of computer vision research. With the development of decades, the target tracking has been developed. From classical tracking methods such as particle filtering and kalman filtering, to detection-based or correlation filtering methods, to deep learning correlation methods that have emerged in recent years. The target tracking is widely applied, such as automatic target tracking of unmanned aerial vehicles, human target tracking, target tracking in traffic monitoring systems, and the like.
The current mainstream target tracking algorithm is mainly based on target detection. However, under the condition that the target detection algorithm has missed detection, the actual application effect of the target tracking algorithm has a gap relative to the effect of the target tracking competition. At present, the most mainstream target detection algorithms include Yolo, fasterrnnn, SSD, etc., which all have missed detection, and this may cause the same tracking target to be regarded as two targets for tracking.
The KCF is fully called a Kernel Correlation Filter Kernel Correlation filtering algorithm. In 2014, the algorithm proposed by joaof, henriques, Rui Caseiro, Pedro Martins, and Jorge Batista, when the algorithm is started after coming out, the algorithm has a very bright appearance in both tracking effect and tracking speed, so that a large number of scholars are led to research the algorithm and the industry applies the algorithm to actual scenes successively. The related filtering algorithm is discriminant tracking, and mainly uses a given sample to train a discriminant classifier to judge whether a tracked target or surrounding background information exists. The method mainly uses a rotation matrix to collect samples, and uses fast Fourier change to perform accelerated calculation on an algorithm.
Disclosure of Invention
Aiming at the condition of missing detection in the prior art, the invention provides the technical scheme that: a multi-target tracking algorithm combining deepsort and KCF is provided. The method comprises the following specific steps:
A multi-target tracking algorithm is applied to an electronic device and comprises the following steps:
S1, detecting current frame position information of one or more targets by using a target detection algorithm;
S2, predicting predicted position information of a next frame of the one or more targets using a KCF tracker;
S3, filtering the predicted position information through an IOU (cross-over-parallel ratio), and then transmitting the filtered predicted position information into a deepsort network for tracking;
And S4, updating the depsort network corresponding to the successfully tracked target, updating the KCF tracker at the same time, and otherwise, judging.
Preferably, wherein the step S4 further comprises: and judging the target candidate item which is not tracked successfully, if the target candidate item is the current frame position information of one or more detected targets, initializing a new Kalman tracking filter and a new KCF tracker for the target candidate item position, and otherwise, not processing.
Preferably, the algorithm further comprises:
And S5, judging that the target of the untracked Kalman tracking filter is lost and setting a continuous lost frame number threshold.
Preferably, when the number of target lost frames is greater than the threshold, it is determined that the target disappears, and the kalman tracking filter and the KCF tracker corresponding to the deppsort tracking network are deleted.
The present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and capable of running the multi-target tracking algorithm as described above on the processor.
The present invention also provides a computer readable storage medium having stored thereon a computer program for execution by a processor of the steps in the multi-target tracking algorithm as described above.
Aiming at the condition that the target detection algorithm has missing detection, the invention provides a multi-target tracking algorithm combining deepsort and KCF. The detection missing condition of the target detection algorithm is made up by using the characteristic that the KCF can accurately predict the position information of the next frame of the target, and the tracking effect of the target tracking algorithm is improved.
Drawings
Hereinafter, the present invention will be explained in more detail with reference to the accompanying drawings. The same reference numbers will be used throughout the drawings to refer to the same or like features:
FIG. 1 is a flow chart of a processing method of a multi-target tracking algorithm incorporating depsort and KCF in accordance with the present invention;
Fig. 2 is a block diagram of an electronic device according to the present invention.
Detailed Description
To further illustrate the various embodiments, the invention provides the accompanying drawings. The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the embodiments. Those skilled in the art will appreciate still other possible embodiments and advantages of the present invention with reference to these figures. Elements in the figures are not drawn to scale and like reference numerals are generally used to indicate like elements.
The invention will now be further described with reference to the accompanying drawings and detailed description.
The invention provides a multi-target tracking algorithm combining deepsort and KCF, an electronic device and a storage medium. The method comprises a target prediction stage and a target tracking stage, and relates to a depsort model and a KCF model, wherein the depsort model only uses a tracking part such as a Kalman tracking filter to confirm tracks and the like, the KCF model is used for predicting target position information of a current frame, and the target position information are combined to screen out an approximate target position. Not only effectively improves the detection effect, but also can avoid certain missed targets.
FIG. 1 shows a flow chart of a processing method of a multi-target tracking algorithm combining deepsort and KCF according to the present invention.
As shown in fig. 1, a multi-target tracking algorithm combining deepsort and KCF includes the following steps:
S1, detecting current frame position information of one or more targets by using a target detection algorithm. Currently, the most mainstream algorithms of Yolo and SSD are all one-stage algorithms, and only one CNN network is used to directly predict the categories and positions of different targets, which has the advantages of high speed but low accuracy, so we need to combine the KCF tracker to perform prediction inference on the basis to ensure the "missed detection" situation.
S2, predicting the predicted position information of the next frame of the one or more targets using the KCF tracker. The correlation filter is improved according to the previous MOSSE algorithm, and can be said to be a nasal ancestor of a tracker such as the later CSK, STC, Color Attributes and the like. The correlation filter is derived from the signal processing field and is then applied to the aspects of image classification and the like. The concept of applying the Correlation Filter to the tracking aspect is as follows: correlation is a measure of the similarity of two signals, and if two signals are more similar, the correlation is higher, and in tracking applications, a filter template needs to be designed so that when it acts on a tracked target, the obtained response is maximum, and the position of the maximum response value is the position of the target, so that the position information of the next frame of the target can be "predicted".
And S3, filtering the predicted position information through the IOU, and then transmitting the filtered predicted position information into a deepsort network for tracking. The region of interest is a region that is delineated by machine vision, image processing, and the like as needed, and is not described in detail herein.
And S4, updating the depsort network corresponding to the successfully tracked target, updating the KCF tracker at the same time, and otherwise, judging.
And judging the target candidate item which is not successfully tracked. If the target candidate item is the target position information detected by the target detection algorithm, initializing a new Kalman tracker and a new KCF tracker (namely judging that the target candidate item is a newly entered target) aiming at the target candidate item position; if the target candidate is the target position information predicted by the KCF (it is considered that the target is likely to go out of view or be occluded), no processing is performed on the target position information. And judging that the target of the unscented Kalman tracker is lost, if the number of continuous lost frames is greater than a set threshold value, judging that the target disappears, and deleting the Kalman tracker and the KCF tracker corresponding to the deepsort tracking network.
The invention provides the above technical solution combining deepsort and KCF for solving the problem of missing detection in multi-target tracking, which is based on the research findings of the inventors of the present application. The inventor of the application finds that the existing target detection algorithms have the problem of missing detection when researching the deepsort multi-target tracking algorithm, so that the problem that one target is tracked for multiple times or is tracked wrongly in one video is caused. The KCF tracking algorithm is a single-target tracking algorithm and has the characteristic of high speed and accurate tracking, so that the KCF can be used for predicting the target position in a new frame of video and combined with the target detected by the target detection algorithm, the problem that the target detection algorithm is missed is solved, and the accuracy of the depsort tracking algorithm is improved.
The difficulty in applying both in combination is how accurately duplicate targets can be excluded for targets predicted by KCF and targets detected by target detection algorithms. Therefore, in the solution algorithm in the present application, whether the same target is determined is further determined by the IOU. And if the target is the same target, deleting the target predicted by the KCF. The IOU between the target predicted by the KCF and the target detected by the target detection algorithm is smaller than the set threshold, if the target detected by the target detection algorithm is a new target entering the video, and if the target predicted by the KCF is a target missed by the target detection algorithm, the new target is judged.
Therefore, the technical problem of the application is solved, the multi-target tracking algorithm combining deepsort and KCF provided by the invention makes up the condition that the target detection algorithm in the prior art has missing detection, and improves the tracking effect of the target tracking algorithm.
Fig. 2 is a block diagram of an electronic device according to the present invention. The electronic device includes: a memory, a processor and a computer program stored in said memory and capable of running on said processor a multi-target tracking algorithm incorporating depsort and KCF as described above.
The present invention also provides a computer readable storage medium storing a computer program for execution by a processor of the steps in the combined deepsort and KCF multi-target tracking algorithm as described above.
While the invention has been particularly shown and described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (6)
1. A multi-target tracking algorithm is characterized by comprising the following steps:
S1, detecting current frame position information of one or more targets by using a target detection algorithm;
S2, predicting predicted position information of a next frame of the one or more targets using a KCF tracker;
S3, filtering the predicted position information through the IOU, and then transmitting the filtered predicted position information into a deepsort network for tracking;
And S4, updating the depsort network corresponding to the successfully tracked target, updating the KCF tracker at the same time, and otherwise, judging.
2. The multi-target tracking algorithm of claim 1, wherein the step S4 further comprises: and judging the target candidate item which is not tracked successfully, if the target candidate item is the current frame position information of one or more detected targets, initializing a new Kalman tracking filter and a new KCF tracker for the target candidate item position, and otherwise, not processing.
3. The multi-target tracking algorithm of claim 1, further comprising:
And S5, judging that the target of the untracked Kalman tracking filter is lost and setting a continuous lost frame number threshold.
4. The multi-target tracking algorithm according to claim 3, wherein when the number of target lost frames is greater than a threshold, it is determined that the target disappears, and the Kalman tracking filter and the KCF tracker corresponding to the depsort tracking network are deleted.
5. An electronic device, comprising: a memory, a processor and a computer program stored in the memory and capable of running the multi-target tracking algorithm of any of claims 1-4 on the processor.
6. A computer-readable storage medium storing a computer program for execution by a processor of the steps in the multi-target tracking algorithm of any one of claims 1-4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010213267.4A CN111428642A (en) | 2020-03-24 | 2020-03-24 | Multi-target tracking algorithm, electronic device and computer readable storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010213267.4A CN111428642A (en) | 2020-03-24 | 2020-03-24 | Multi-target tracking algorithm, electronic device and computer readable storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111428642A true CN111428642A (en) | 2020-07-17 |
Family
ID=71548606
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010213267.4A Pending CN111428642A (en) | 2020-03-24 | 2020-03-24 | Multi-target tracking algorithm, electronic device and computer readable storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111428642A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111814734A (en) * | 2020-07-24 | 2020-10-23 | 南方电网数字电网研究院有限公司 | Method for identifying state of knife switch |
CN111986237A (en) * | 2020-09-01 | 2020-11-24 | 安徽炬视科技有限公司 | Real-time multi-target tracking algorithm irrelevant to number of people |
CN112581507A (en) * | 2020-12-31 | 2021-03-30 | 北京澎思科技有限公司 | Target tracking method, system and computer readable storage medium |
CN112581506A (en) * | 2020-12-31 | 2021-03-30 | 北京澎思科技有限公司 | Face tracking method, system and computer readable storage medium |
CN112700469A (en) * | 2020-12-30 | 2021-04-23 | 武汉卓目科技有限公司 | Visual target tracking method and device based on ECO algorithm and target detection |
CN112734809A (en) * | 2021-01-21 | 2021-04-30 | 高新兴科技集团股份有限公司 | Online multi-pedestrian tracking method and device based on Deep-Sort tracking framework |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107346538A (en) * | 2016-05-06 | 2017-11-14 | 株式会社理光 | Method for tracing object and equipment |
CN108053427A (en) * | 2017-10-31 | 2018-05-18 | 深圳大学 | A kind of modified multi-object tracking method, system and device based on KCF and Kalman |
CN108198209A (en) * | 2017-12-22 | 2018-06-22 | 天津理工大学 | It is blocking and dimensional variation pedestrian tracking algorithm |
CN109448027A (en) * | 2018-10-19 | 2019-03-08 | 成都睿码科技有限责任公司 | A kind of adaptive, lasting motion estimate method based on algorithm fusion |
US20190213420A1 (en) * | 2018-01-09 | 2019-07-11 | Qualcomm Incorporated | Adaptive object detection and recognition |
CN110222667A (en) * | 2019-06-17 | 2019-09-10 | 南京大学 | A kind of open route traffic participant collecting method based on computer vision |
CN110516556A (en) * | 2019-07-31 | 2019-11-29 | 平安科技(深圳)有限公司 | Multi-target tracking detection method, device and storage medium based on Darkflow-DeepSort |
CN110751674A (en) * | 2018-07-24 | 2020-02-04 | 北京深鉴智能科技有限公司 | Multi-target tracking method and corresponding video analysis system |
-
2020
- 2020-03-24 CN CN202010213267.4A patent/CN111428642A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107346538A (en) * | 2016-05-06 | 2017-11-14 | 株式会社理光 | Method for tracing object and equipment |
CN108053427A (en) * | 2017-10-31 | 2018-05-18 | 深圳大学 | A kind of modified multi-object tracking method, system and device based on KCF and Kalman |
CN108198209A (en) * | 2017-12-22 | 2018-06-22 | 天津理工大学 | It is blocking and dimensional variation pedestrian tracking algorithm |
US20190213420A1 (en) * | 2018-01-09 | 2019-07-11 | Qualcomm Incorporated | Adaptive object detection and recognition |
CN110751674A (en) * | 2018-07-24 | 2020-02-04 | 北京深鉴智能科技有限公司 | Multi-target tracking method and corresponding video analysis system |
CN109448027A (en) * | 2018-10-19 | 2019-03-08 | 成都睿码科技有限责任公司 | A kind of adaptive, lasting motion estimate method based on algorithm fusion |
CN110222667A (en) * | 2019-06-17 | 2019-09-10 | 南京大学 | A kind of open route traffic participant collecting method based on computer vision |
CN110516556A (en) * | 2019-07-31 | 2019-11-29 | 平安科技(深圳)有限公司 | Multi-target tracking detection method, device and storage medium based on Darkflow-DeepSort |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111814734A (en) * | 2020-07-24 | 2020-10-23 | 南方电网数字电网研究院有限公司 | Method for identifying state of knife switch |
CN111814734B (en) * | 2020-07-24 | 2024-01-26 | 南方电网数字电网研究院有限公司 | Method for identifying state of disconnecting link |
CN111986237A (en) * | 2020-09-01 | 2020-11-24 | 安徽炬视科技有限公司 | Real-time multi-target tracking algorithm irrelevant to number of people |
CN112700469A (en) * | 2020-12-30 | 2021-04-23 | 武汉卓目科技有限公司 | Visual target tracking method and device based on ECO algorithm and target detection |
CN112581507A (en) * | 2020-12-31 | 2021-03-30 | 北京澎思科技有限公司 | Target tracking method, system and computer readable storage medium |
CN112581506A (en) * | 2020-12-31 | 2021-03-30 | 北京澎思科技有限公司 | Face tracking method, system and computer readable storage medium |
CN112734809A (en) * | 2021-01-21 | 2021-04-30 | 高新兴科技集团股份有限公司 | Online multi-pedestrian tracking method and device based on Deep-Sort tracking framework |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111428642A (en) | Multi-target tracking algorithm, electronic device and computer readable storage medium | |
De Geest et al. | Online action detection | |
US9213896B2 (en) | Method for detecting and tracking objects in image sequences of scenes acquired by a stationary camera | |
US20110052002A1 (en) | Foreground object tracking | |
US20180268556A1 (en) | Method for detecting moving objects in a video having non-stationary background | |
US10671887B2 (en) | Best image crop selection | |
CN109766867A (en) | Travel condition of vehicle determines method, apparatus, computer equipment and storage medium | |
CN109711332B (en) | Regression algorithm-based face tracking method and application | |
CN112750150A (en) | Vehicle flow statistical method based on vehicle detection and multi-target tracking | |
CN114879696A (en) | Track matching method, device, equipment and medium | |
CN104077776A (en) | Visual background extracting algorithm based on color space self-adapting updating | |
CN114708300A (en) | Anti-blocking self-adaptive target tracking method and system | |
CN111639570B (en) | Online multi-target tracking method based on motion model and single-target clue | |
CN111161324B (en) | Target tracking method based on self-adaptive multimode updating strategy | |
CN112037257A (en) | Target tracking method, terminal and computer readable storage medium thereof | |
CN111428626B (en) | Method and device for identifying moving object and storage medium | |
WO2017077261A1 (en) | A monocular camera cognitive imaging system for a vehicle | |
CN112015960A (en) | Clustering method of vehicle-mounted radar measurement data, storage medium and electronic device | |
CN111582025B (en) | Method and device for identifying moving object and storage medium | |
CN114879177B (en) | Target analysis method and device based on radar information | |
CN112906424B (en) | Image recognition method, device and equipment | |
US11386656B2 (en) | Device and method of handling video content analysis | |
TW202139134A (en) | Device and method of handling video content analysis | |
Su et al. | A Survey of Long-term Object Tracking Methods | |
CN115797410A (en) | Vehicle tracking method and system |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200717 |