CN111986237A - Real-time multi-target tracking algorithm irrelevant to number of people - Google Patents
Real-time multi-target tracking algorithm irrelevant to number of people Download PDFInfo
- Publication number
- CN111986237A CN111986237A CN202010906003.7A CN202010906003A CN111986237A CN 111986237 A CN111986237 A CN 111986237A CN 202010906003 A CN202010906003 A CN 202010906003A CN 111986237 A CN111986237 A CN 111986237A
- Authority
- CN
- China
- Prior art keywords
- target
- tracking
- targets
- people
- chain
- 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
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
-
- 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
- G06T2207/30232—Surveillance
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to multi-target tracking, in particular to a real-time multi-target tracking algorithm independent of the number of people, which utilizes a target detection algorithm to identify targets of an input image, filters improper targets and establishes a position chain for recording the moving track of the targets, and carries out KCF tracking operation on the targets in parallel in a multi-thread mode, if the operation result meets the tracking condition, the targets are continuously tracked and corresponding position chains are updated, otherwise, the targets are released, each target is compared, repeated tracking targets are released, the position information of the filtered targets in the next frame of image is predicted by utilizing the KCF tracking operation, a prediction chain for recording the predicted position information of the targets is established, the predicted position information is filtered, and the filtered predicted position information is input into the prediction chain for tracking; the technical scheme provided by the invention can effectively overcome the defect that the multi-target tracking can not be accurately and effectively carried out in the prior art.
Description
Technical Field
The invention relates to multi-target tracking, in particular to a real-time multi-target tracking algorithm irrelevant to the number of people.
Background
The target tracking is an important content in video monitoring and video analysis, and has wide application in the aspects of intelligent monitoring, man-machine interaction, robot navigation, medical diagnosis and the like. Object tracking mainly refers to determining the position of a moving object of interest in each image of a video and corresponding the same object in different frames.
The existing target tracking algorithms are mainly divided into two categories, namely batch processing methods and online methods. The batch processing method processes the detection results of all frames, and links the tracks that are interrupted due to blocking, such as "Continuous Energy Minimization for Multi-Target Tracking" (Continuous Energy Minimization of Multi-Target Tracking) published by a.andriyenko et al in 2014 IEEE TPAMI at 35, and after obtaining the detection result of each frame, the detection results are connected to obtain a Tracking segment, and then the Tracking segment is globally linked to obtain a long Tracking segment. Therefore, the key to this approach is global association (global association), whereas batch processing approaches exhibit deficiencies when there is a long time of blocking because it is difficult to distinguish between different targets. Meanwhile, these methods require detection of the image sequence in advance, and are computationally expensive, and thus are difficult to apply in real-time applications.
In addition, there are some Online methods, such as "Online Multi-person tracking-by-detection from a single, uncalibrated camera" (Online Multi-person tracking using a single uncalibrated camera) published by m.d. breitenstein in 2011 IEEE TPAMI, and "Multi-target tracking on consistency maps: application of one person tracking" published by f.poiesi et al in 2013 CVIU, and a target trajectory is established using information accumulated to the current frame according to the relation between frames, and thus can be used to implement real-time application. However, this method is difficult to deal with objects that are not accurately detected because they are occluded, and online methods produce tracking segments that are often intermittent, which can miss objects when they are occluded.
Disclosure of Invention
Technical problem to be solved
Aiming at the defects in the prior art, the invention provides a real-time multi-target tracking algorithm irrelevant to the number of people, which can effectively overcome the defect that the multi-target tracking can not be accurately and effectively carried out in the prior art.
(II) technical scheme
In order to achieve the purpose, the invention is realized by the following technical scheme:
a real-time multi-target tracking algorithm irrelevant to people number comprises the following steps:
s1, performing target identification on the input image by using a target detection algorithm, filtering out unsuitable targets, and establishing a position chain for recording target movement tracks;
s2, performing KCF tracking operation on the target in parallel through multiple threads, if the operation result meets the tracking condition, continuing to track the target, updating the corresponding position chain, and otherwise, releasing the target;
s3, comparing each target and releasing the repeatedly tracked target;
s4, predicting the position information of the filtered target in the next frame image by utilizing KCF tracking operation, and establishing a prediction chain for recording the predicted position information of the target;
s5, filtering the predicted position information, and inputting the filtered predicted position information into a prediction chain for tracking;
s6, extracting feature vectors of the updated position chain and the prediction chain respectively, and constructing a weighted undirected graph by combining the target and the feature vectors;
s7, clustering the undirected graph to obtain clusters, and updating the current position of the target according to the clusters;
and S8, inputting the next frame of image, comprehensively judging the current positions of the target and acquiring new target position information by the target detection algorithm, judging whether a new tracking target is established or not, and repeating the step S2.
Preferably, the performing object recognition on the input image by using an object detection algorithm in S1 includes:
s11, scaling the input image to 320x240, and performing normalization operation;
s12, selecting an interested area for the input image, and selecting the directional gradient histogram feature for the interested area.
Preferably, in S2, the performing KCF tracking operation on the target in parallel by multiple threads includes: acquiring the CPU core number of a computer, starting thread numbers with the same number as the CPU core number, simultaneously performing tracking operation, completing the operation of one target by the thread, and supplementing a new target.
Preferably, comparing each target in S3 and releasing the repeated tracking target comprises:
s31, calculating the coincidence ratio of the positions of the targets, wherein the coincidence ratio is the ratio of the areas of the two rectangular frames;
s32, if the superposition ratio is more than 0.65, repeated tracking occurs;
and S33, reserving the target with the minimum distance between the rectangular frame and the bottom edge of the video, and releasing the rest repeated tracking targets.
Preferably, the filtered predicted location information in S5 is input into a prediction chain for tracking, including: when the target is not successfully tracked by the prediction chain, the initialization of the KCF tracker is carried out according to the position of the target in the current frame image.
Preferably, if the target is not successfully tracked by the prediction chain after the KCF tracker is initialized for the position of the target in the current frame image, the number of consecutive lost frames of the lost target is determined;
when the number of the continuous lost frames of the lost target is greater than the threshold value, judging that the lost target disappears, and inputting the disappearing information into a prediction chain; otherwise, judging that the lost target disappears in the lost frame number image, and inputting the disappearance information into the prediction chain.
Preferably, the step of constructing a weighted undirected graph by combining the target and the feature vector in S6 includes: and taking the target and the feature vector as nodes, and taking the similarity between the nodes as the weight of the edge.
Preferably, a weighted undirected graph is constructed in S6 by combining the target and the feature vector, and includes two constraints:
A. the nodes of the same type do not calculate the similarity and are not connected with the edge;
B. when the coordinate distance between the nodes exceeds a threshold, the edge is not connected.
Preferably, the updating the target current position according to the cluster in S7 includes: and fusing the position coordinates and the prediction coordinates in the same cluster as the current position of the target in the current frame image.
(III) advantageous effects
Compared with the prior art, the real-time multi-target tracking algorithm irrelevant to the number of people provided by the invention has the advantages that the KCF tracking operation is carried out on the target in parallel by utilizing multiple threads, the multi-target is tracked in real time, the follow-up moving position of the tracked target is effectively predicted by establishing the prediction chain, and the current position of the target is updated by combining the position chain and the prediction chain, so that the condition of missing detection can be effectively avoided, and the continuity and the accuracy of multi-target detection are ensured.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It is obvious that the drawings in the following description are only some embodiments of the invention, and that for a person skilled in the art, other drawings can be derived from them without inventive effort.
FIG. 1 is a schematic diagram of a multi-target tracking process according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
A real-time multi-target tracking algorithm irrelevant to the number of people is disclosed, as shown in figure 1, a target detection algorithm is used for carrying out target recognition on an input image, filtering out unsuitable targets and establishing a position chain for recording the moving track of the targets.
Before the target detection algorithm is used for carrying out target identification on the input image, the purpose of meeting the use requirement can be achieved only by carrying out parameter adjustment on the input image, and each target is subjected to initialization identification. The method for identifying the target of the input image by using the target detection algorithm comprises the following steps:
s11, scaling the input image to 320x240, and performing normalization operation;
s12, selecting an interested area for the input image, and selecting the directional gradient histogram feature for the interested area.
And performing KCF tracking operation on the targets in parallel by multiple threads, if the operation result meets the tracking condition, continuing to track the targets, updating the corresponding position chains, and otherwise, releasing the targets.
In the technical scheme, the tracking adopts multi-thread parallel operation, and a plurality of targets can be ensured to be tracked simultaneously. The method for performing KCF tracking operation on the target in parallel by multiple threads comprises the following steps: acquiring the CPU core number of a computer, starting thread numbers with the same number as the CPU core number, simultaneously performing tracking operation, completing the operation of one target by the thread, and supplementing a new target.
And comparing each target, and releasing the repeated tracking target.
In the technical scheme, the problem of repeated tracking of the same target may exist, so that not only can computing resources be wasted, but also the accuracy of tracking the target can be influenced. Wherein, comparing each target, releasing the repeated tracking target, comprising:
s31, calculating the coincidence ratio of the positions of the targets, wherein the coincidence ratio is the ratio of the areas of the two rectangular frames;
s32, if the superposition ratio is more than 0.65, repeated tracking occurs;
and S33, reserving the target with the minimum distance between the rectangular frame and the bottom edge of the video, and releasing the rest repeated tracking targets.
And predicting the position information of the filtered target in the next frame of image by utilizing KCF tracking operation, and establishing a prediction chain for recording the predicted position information of the target.
In the technical scheme of the application, the position information of the filtered target in the next frame image can be predicted by adopting a relevant filter, and the principle 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 object, the obtained response is maximum, and the position of the maximum response value is the position of the object, so that the position information in the next frame image can be "predicted".
And filtering the predicted position information, and inputting the filtered predicted position information into a prediction chain for tracking.
In the multi-target tracking process, the tracking target may disappear from the video image. For the situation that the tracking target disappears, in the technical scheme of the application, the filtered predicted position information is input into a prediction chain for tracking, and the method comprises the following steps: when the target is not successfully tracked by the prediction chain, the initialization of the KCF tracker is carried out according to the position of the target in the current frame image.
And if the target is not successfully tracked by the prediction chain after the initialization of the KCF tracker is carried out on the position of the target in the current frame image, judging the number of the continuously lost frames of the lost target.
When the number of the continuous lost frames of the lost target is greater than the threshold value, judging that the lost target disappears, and inputting the disappearing information into a prediction chain; otherwise, judging that the lost target disappears in the lost frame number image, and inputting the disappearance information into the prediction chain.
And respectively extracting the feature vectors of the updated position chain and the prediction chain, and constructing a weighted undirected graph by combining the target and the feature vectors.
The method for constructing the weighted undirected graph by combining the target and the feature vector comprises the following steps: and taking the target and the feature vector as nodes, and taking the similarity between the nodes as the weight of the edge. While the creation of an undirected graph includes two constraints:
A. the nodes of the same type do not calculate the similarity and are not connected with the edge;
B. when the coordinate distance between the nodes exceeds a threshold, the edge is not connected.
And clustering the undirected graph to obtain clusters, and updating the current position of the target according to the clusters.
Wherein updating the current location of the target according to the cluster comprises: and fusing the position coordinates and the prediction coordinates in the same cluster as the current position of the target in the current frame image. Because the undirected graph contains the characteristic vectors of the position chain and the prediction chain, the undirected graph can be combined with the position chain and the prediction chain to update the current position of the target, thereby effectively avoiding the condition of missing detection and ensuring the continuity and accuracy of multi-target detection.
Inputting the next frame of image, comprehensively judging the target detection algorithm to obtain new target position information and the current positions of all targets, judging whether to establish a new tracking target, and repeating multithreading and performing KCF tracking operation on the target in parallel.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.
Claims (9)
1. A people-number-independent real-time multi-target tracking algorithm is characterized in that: the method comprises the following steps:
s1, performing target identification on the input image by using a target detection algorithm, filtering out unsuitable targets, and establishing a position chain for recording target movement tracks;
s2, performing KCF tracking operation on the target in parallel through multiple threads, if the operation result meets the tracking condition, continuing to track the target, updating the corresponding position chain, and otherwise, releasing the target;
s3, comparing each target and releasing the repeatedly tracked target;
s4, predicting the position information of the filtered target in the next frame image by utilizing KCF tracking operation, and establishing a prediction chain for recording the predicted position information of the target;
s5, filtering the predicted position information, and inputting the filtered predicted position information into a prediction chain for tracking;
s6, extracting feature vectors of the updated position chain and the prediction chain respectively, and constructing a weighted undirected graph by combining the target and the feature vectors;
s7, clustering the undirected graph to obtain clusters, and updating the current position of the target according to the clusters;
and S8, inputting the next frame of image, comprehensively judging the current positions of the target and acquiring new target position information by the target detection algorithm, judging whether a new tracking target is established or not, and repeating the step S2.
2. The people-independent real-time multi-target tracking algorithm of claim 1, wherein: in S1, performing object recognition on the input image by using an object detection algorithm, including:
s11, scaling the input image to 320x240, and performing normalization operation;
s12, selecting an interested area for the input image, and selecting the directional gradient histogram feature for the interested area.
3. The people-independent real-time multi-target tracking algorithm of claim 1, wherein: in S2, the multithreading parallel KCF tracking operation on the target includes: acquiring the CPU core number of a computer, starting thread numbers with the same number as the CPU core number, simultaneously performing tracking operation, completing the operation of one target by the thread, and supplementing a new target.
4. The people-independent real-time multi-target tracking algorithm of claim 1, wherein: comparing each target in S3, and releasing the repeated tracking targets, wherein the steps comprise:
s31, calculating the coincidence ratio of the positions of the targets, wherein the coincidence ratio is the ratio of the areas of the two rectangular frames;
s32, if the superposition ratio is more than 0.65, repeated tracking occurs;
and S33, reserving the target with the minimum distance between the rectangular frame and the bottom edge of the video, and releasing the rest repeated tracking targets.
5. The people-independent real-time multi-target tracking algorithm of claim 1, wherein: inputting the filtered predicted position information into a prediction chain for tracking in S5, including: when the target is not successfully tracked by the prediction chain, the initialization of the KCF tracker is carried out according to the position of the target in the current frame image.
6. The people-independent real-time multi-target tracking algorithm of claim 5, wherein: if the target is initialized in the current frame image according to the position of the target, and the target is still not successfully tracked by the prediction chain, judging the continuous lost frame number of the lost target;
when the number of the continuous lost frames of the lost target is greater than the threshold value, judging that the lost target disappears, and inputting the disappearing information into a prediction chain; otherwise, judging that the lost target disappears in the lost frame number image, and inputting the disappearance information into the prediction chain.
7. The people-independent real-time multi-target tracking algorithm of claim 1, wherein: and S6, combining the target and the feature vector to construct a weighted undirected graph, which comprises the following steps: and taking the target and the feature vector as nodes, and taking the similarity between the nodes as the weight of the edge.
8. The people-independent real-time multi-target tracking algorithm of claim 7, wherein: and S6, constructing a weighted undirected graph by combining the target and the feature vector, wherein the undirected graph comprises two constraints:
A. the nodes of the same type do not calculate the similarity and are not connected with the edge;
B. when the coordinate distance between the nodes exceeds a threshold, the edge is not connected.
9. The people-independent real-time multi-target tracking algorithm of claim 1, wherein: in S7, updating the current position of the target according to the cluster includes: and fusing the position coordinates and the prediction coordinates in the same cluster as the current position of the target in the current frame image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010906003.7A CN111986237A (en) | 2020-09-01 | 2020-09-01 | Real-time multi-target tracking algorithm irrelevant to number of people |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010906003.7A CN111986237A (en) | 2020-09-01 | 2020-09-01 | Real-time multi-target tracking algorithm irrelevant to number of people |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111986237A true CN111986237A (en) | 2020-11-24 |
Family
ID=73447790
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010906003.7A Pending CN111986237A (en) | 2020-09-01 | 2020-09-01 | Real-time multi-target tracking algorithm irrelevant to number of people |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111986237A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113052116A (en) * | 2021-04-06 | 2021-06-29 | 深圳华声医疗技术股份有限公司 | Ultrasonic video data processing method and device, ultrasonic equipment and storage medium |
| CN113221808A (en) * | 2021-05-26 | 2021-08-06 | 新疆爱华盈通信息技术有限公司 | Dinner plate counting statistical method and device based on image recognition |
| CN113538515A (en) * | 2021-07-19 | 2021-10-22 | 安徽炬视科技有限公司 | High-voltage switch cabinet abnormal movement detection algorithm based on combination of semantic segmentation and target detection tracking |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109993769A (en) * | 2019-03-07 | 2019-07-09 | 安徽创世科技股份有限公司 | A kind of multiple-target system of deep learning SSD algorithm combination KCF algorithm |
| CN110363791A (en) * | 2019-06-28 | 2019-10-22 | 南京理工大学 | A kind of online multiple target tracking algorithm merging monotrack result |
| WO2019237536A1 (en) * | 2018-06-11 | 2019-12-19 | 平安科技（深圳）有限公司 | Target real-time tracking method and apparatus, and computer device and storage medium |
| CN111428642A (en) * | 2020-03-24 | 2020-07-17 | 厦门市美亚柏科信息股份有限公司 | Multi-target tracking algorithm, electronic device and computer readable storage medium |
-
2020
- 2020-09-01 CN CN202010906003.7A patent/CN111986237A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019237536A1 (en) * | 2018-06-11 | 2019-12-19 | 平安科技（深圳）有限公司 | Target real-time tracking method and apparatus, and computer device and storage medium |
| CN109993769A (en) * | 2019-03-07 | 2019-07-09 | 安徽创世科技股份有限公司 | A kind of multiple-target system of deep learning SSD algorithm combination KCF algorithm |
| CN110363791A (en) * | 2019-06-28 | 2019-10-22 | 南京理工大学 | A kind of online multiple target tracking algorithm merging monotrack result |
| CN111428642A (en) * | 2020-03-24 | 2020-07-17 | 厦门市美亚柏科信息股份有限公司 | Multi-target tracking algorithm, electronic device and computer readable storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 齐美彬 等: "基于广义关联聚类图的分层关联多目标跟踪", 《自动化学报》, vol. 43, no. 1, pages 152 - 160 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113052116A (en) * | 2021-04-06 | 2021-06-29 | 深圳华声医疗技术股份有限公司 | Ultrasonic video data processing method and device, ultrasonic equipment and storage medium |
| CN113052116B (en) * | 2021-04-06 | 2022-02-22 | 深圳华声医疗技术股份有限公司 | Ultrasonic video data processing method and device, ultrasonic equipment and storage medium |
| CN113221808A (en) * | 2021-05-26 | 2021-08-06 | 新疆爱华盈通信息技术有限公司 | Dinner plate counting statistical method and device based on image recognition |
| CN113538515A (en) * | 2021-07-19 | 2021-10-22 | 安徽炬视科技有限公司 | High-voltage switch cabinet abnormal movement detection algorithm based on combination of semantic segmentation and target detection tracking |
| CN113538515B (en) * | 2021-07-19 | 2024-06-07 | 安徽炬视科技有限公司 | High-voltage switch cabinet abnormal state detection method based on combination of semantic segmentation and target detection tracking |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107516321B (en) | Video multi-target tracking method and device | |
| CN111986237A (en) | Real-time multi-target tracking algorithm irrelevant to number of people | |
| CN101325691B (en) | Method and apparatus for tracing a plurality of observation model with fusion of differ durations | |
| Simonnet et al. | Re-identification of pedestrians in crowds using dynamic time warping | |
| CN106570490B (en) | A kind of pedestrian's method for real time tracking based on quick clustering | |
| CN106934817B (en) | Multi-attribute-based multi-target tracking method and device | |
| CN103971386A (en) | Method for foreground detection in dynamic background scenario | |
| CN113192105B (en) | Method and device for indoor multi-person tracking and attitude measurement | |
| CN107564035B (en) | Video tracking method based on important area identification and matching | |
| CN105894530A (en) | Detection and tracking solution scheme aiming at motion target in video | |
| Nallasivam et al. | Moving human target detection and tracking in video frames | |
| CN108830204B (en) | Method for detecting abnormality in target-oriented surveillance video | |
| Augustin et al. | Motion and feature based person tracking in surveillance videos | |
| CN109754409A (en) | A kind of monitor video pedestrian target matched jamming System and method for | |
| CN110717934A (en) | Anti-occlusion target tracking method based on STRCF | |
| CN114926859A (en) | Pedestrian multi-target tracking method in dense scene combined with head tracking | |
| Zheng et al. | Adaptive edge-based mean shift for drastic change gray target tracking | |
| He et al. | Fast online multi-pedestrian tracking via integrating motion model and deep appearance model | |
| Wei et al. | Learning spatio-temporal information for multi-object tracking | |
| CN110349184A (en) | The more pedestrian tracting methods differentiated based on iterative filtering and observation | |
| KR101690050B1 (en) | Intelligent video security system | |
| CN107729811B (en) | Night flame detection method based on scene modeling | |
| Chen et al. | Intrusion detection of specific area based on video | |
| CN113012193A (en) | Multi-pedestrian tracking method based on deep learning | |
| CN116342645A (en) | Multi-target tracking method for natatorium scene |
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 |