CN110659559A - Multi-target tracking method and system for monitoring scene - Google Patents
Multi-target tracking method and system for monitoring scene Download PDFInfo
- Publication number
- CN110659559A CN110659559A CN201910713029.7A CN201910713029A CN110659559A CN 110659559 A CN110659559 A CN 110659559A CN 201910713029 A CN201910713029 A CN 201910713029A CN 110659559 A CN110659559 A CN 110659559A
- Authority
- CN
- China
- Prior art keywords
- target
- hypothesis
- detection
- frames
- frame
- 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
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000012544 monitoring process Methods 0.000 title claims abstract description 20
- 238000001514 detection method Methods 0.000 claims abstract description 68
- 238000013138 pruning Methods 0.000 claims abstract description 25
- 238000005457 optimization Methods 0.000 claims abstract description 19
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000000717 retained effect Effects 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 9
- 238000000605 extraction Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 2
- 244000141353 Prunus domestica Species 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
Abstract
The invention discloses a multi-target tracking method and a system for a monitored scene, which are based on multi-hypothesis tracking after pruning optimization, and the method comprises the following steps: receiving a target detection frame of a current frame; generating hypotheses of possible positions of each target based on the target detection frame of the current frame; solving the maximum independent subset and the problem based on all the hypotheses of a plurality of recent frames to obtain a mutually non-conflicting hypothesis path which can be selected by each target, and outputting the position of each target before the plurality of frames; and (4) pruning optimization is carried out by combining the maximum independent subset and the solution of the problem and the Hungarian algorithm, so as to obtain a multi-target tracking result. The invention provides a novel pruning optimizing method, which greatly reduces the complexity of a multi-hypothesis tracking algorithm. For multi-target tracking of a monitoring scene, CPU resources can be fully utilized, GPU resources are not occupied, and multi-targets are tracked in real time; and the adjustment of algorithm complexity and precision balance can be realized by changing the time for integrating information.
Description
Technical Field
The invention relates to the technical field of computers, in particular to a real-time tracking method and a real-time tracking system for multiple targets in a monitoring scene.
Background
In the process of intelligently analyzing the monitoring video, the detection frame of the target is extracted, and a specific identification code is given to each target by a multi-target tracking method for distinguishing different targets. The multi-target tracking method is a very classical problem in the field of computer research. Typically, an algorithm will convert the tracking problem into a graph model to solve. But it is often difficult to flexibly adjust the accuracy and speed balance of the algorithm. In recent years, the improvement of the multi-target tracking method is mostly a feature extraction method based on deep learning. A depth feature extraction network is introduced into the multi-target tracking method to obtain the depth features of the target, so that the tracking accuracy in a complex scene is improved, and the search space can be pruned to a certain extent. But depth feature based approaches are limited in speed by GPU resources.
In a scene of intelligent processing of an actual monitoring video, GPU resources are often very tight, and CPU resources are generally loose. The speed of the tracking algorithm of the depth feature based network depends heavily on the performance of the GPU. When the number of targets in a scene is large, or other algorithm modules occupy GPU resources, the depth feature extraction process can severely slow down the real-time performance of the multi-target tracking method.
In addition, different application scenes may have different requirements on the balance of the complexity and the precision of the algorithm, and most of the existing formed algorithms do not have any adjusting space and can only perform calculation under the complexity of a single algorithm.
The multi-hypothesis tracking which is not based on the depth feature extraction can effectively utilize CPU resources, and can realize the balance adjustment of complexity and precision by adjusting the time for integrating information. However, in the original multi-hypothesis tracking algorithm, although the maximum independent subset sum algorithm (MWIS) prunes the hypothesis tree, the number of hypotheses per target still increases geometrically with time before reaching the pruning time point, and it is difficult to integrate the hypothesis information for a long time to make a judgment.
Disclosure of Invention
The purpose of the invention is realized by the following technical scheme.
According to a first aspect of the invention, a multi-target tracking method for a monitored scene is provided, which is based on multi-hypothesis tracking after pruning optimization, and comprises the following steps:
receiving a target detection frame of a current frame;
generating hypotheses of possible positions of each target based on the target detection frame of the current frame;
solving the maximum independent subset and the problem based on all the hypotheses of a plurality of recent frames to obtain a mutually non-conflicting hypothesis path which can be selected by each target, and outputting the position of each target before the plurality of frames;
and (4) pruning optimization is carried out by combining the maximum independent subset and the solution of the problem and the Hungarian algorithm, so as to obtain a multi-target tracking result.
According to a second aspect of the present invention, a monitored scene multi-target tracking system is provided, which includes the following modules:
the detection frame receiving module is used for receiving a target detection frame of the current frame;
an assumed position generating module for generating an assumption of a possible position of each target based on the target detection frame of the current frame;
the target position calculation module is used for solving the maximum independent subset and the problem based on all the assumptions of the latest frames to obtain the mutually non-conflicting assumed path which can be selected by each target, and outputting the position of each target before the frames;
and the pruning optimization module is used for carrying out pruning optimization by combining the maximum independent subset and the solution of the problems and the Hungarian algorithm to obtain a multi-target tracking result.
The invention has the advantages that: compared with the prior art, the multi-target tracking of the real-time monitoring scene disclosed by the invention has the following beneficial effects: the method greatly reduces the complexity of the multi-hypothesis tracking algorithm by providing a novel pruning optimizing method on the basis of the traditional multi-hypothesis tracking algorithm. For multi-target tracking of a monitoring scene, CPU resources can be fully utilized, GPU resources are not occupied, and multi-targets are tracked in real time; and the adjustment of algorithm complexity and precision balance can be realized by changing the time for integrating information.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:
FIG. 1 is a general flow diagram of a method for monitoring scene multi-target tracking according to an embodiment of the invention;
FIG. 2 illustrates a flow diagram for reading in a detection box according to an embodiment of the invention;
FIG. 3 is a diagram illustrating a matching process assuming detection boxes according to an embodiment of the present invention;
FIG. 4 illustrates an exemplary diagram of a conversion to a maximum independent subset and a problem according to an embodiment of the invention;
FIG. 5 illustrates an effect diagram of pruning optimization according to an embodiment of the present invention;
FIG. 6 illustrates a detailed flow diagram of multi-hypothesis tracking according to an embodiment of the invention;
FIG. 7 is a diagram illustrating an exemplary effect of a backtracking length n on algorithm accuracy and speed according to an embodiment of the present invention;
FIG. 8 is a diagram illustrating a multi-target tracking system for a monitored scene according to an embodiment of the invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
The assumption in the present invention is that the multi-hypothesis tracking algorithm maintains an assumption of the possible positions of the targets for each target because of the inaccuracy that detection algorithms usually bring. Hypothetical dependencies between different frames constitute a hypothetical tree. The actual position of the target in a certain frame needs the information of a plurality of frames after synthesis to be determined.
As shown in fig. 1, the invention discloses a multi-target tracking method in a monitoring scene, comprising the following steps:
A1. reading in a target detection frame from a preamble module: reading a detection frame in a current video frame from detectors deployed on different server sides or the same server side (as shown in FIG. 2);
A2. generating new hypotheses based on the detection boxes: for each hypothesis, the detection boxes in the proximity range generate a new hypothesis for matching (as shown in FIG. 3);
A3. determining the position of the target n frames before: by converting the hypothesis tree formed by all the hypotheses from the t-n frame to the t frame into the Maximum Independent subset and problem (MWIS), the selectable path on the hypothesis tree which can be selected by each target and does not conflict with each other can be solved. The hypotheses not on the paths before the t-n frames are deleted, and each target only keeps one hypothesis, namely the position of the target output by the algorithm when the target is in the t-n frames. (as shown in FIGS. 4 and 6)
A4. Pruning optimization: pruning optimization is carried out by combining the MWIS solution and the Hungarian algorithm (as shown in figures 5 and 6), and a multi-target tracking result is obtained.
If the algorithm precision is not enough or the complexity is too high, the time length of the algorithm integration information can be adjusted by adjusting the size of n, and the proper balance between the speed and the precision is obtained. (as shown in FIG. 7)
Preferably, in the multi-target tracking method, the step of reading in the target detection frame includes the steps as shown in fig. 2:
C1. reading and decoding of the monitoring video: and reading and decoding the required monitoring into a single-frame picture through a disk or a network. Such as the nth frame of video.
C2. Selecting a detector: and selecting a detector for detecting the target, and extracting a detection frame of the target from the single-frame picture.
C3. And (3) transmission of the detection frame: if the detector and the tracker are deployed on different servers, the result of the detector needs to be transmitted to the server A, B where the tracker is located through a network data packet; if the detector and the tracker are deployed at the same server A, the detection frame can be transmitted through the process or the inter-thread passage, and the tracking result of the t-n frame is finally obtained.
Preferably, in the multi-target tracking method, the step of generating a new hypothesis based on the detection frame is shown in fig. 3 and includes:
D1. finding detection boxes whose records are supposed to be adjacent: traversing all hypotheses of the current frame maintained at present, and searching and recording each hypothesis and a detection frame adjacent to the hypothesis within a certain range;
D2. new assumptions were generated: for each hypothesis and detection box matching, a new hypothesis corresponding to the matched detection box is generated, i.e. each hypothesis may reach the position of the detection box in the adjacent range. If there are no detection boxes in the proximity, an empty hypothesis is generated, which does not match any detection box, and the state variables (position, velocity) of the empty hypothesis are the same as its parent hypothesis. The null hypothesis corresponds to a situation where the target position is not given by the detector and needs to be predicted by the tracker. If the detection frames are larger than a threshold value, then in addition to generating the hypothesis matching the detection frames, an empty hypothesis is generated to take into account the case that the frames of the detector are inaccurate or belong to other close targets.
Preferably, in the multi-target tracking method, as shown in fig. 4 (in the case of fig. 4, n is 2), the step of determining the position of the target before n frames includes:
E1. and according to the detection frame of the current frame, calculating the matching cost of a new hypothesis generated by matching the current hypothesis with the detection frame according to the matching distance, the direction of the matched target speed and the difference between the target size and the detection frame size.
E2. Go back to the hypothesis tree n frames ago. Hypothesis tree n frames ago in the last round of algorithm operation, the hypothesis of each target is derived from the same parent node. And calculating the sum of the matching cost of each path on the hypothesis tree with the depth of n, and putting the sum at a leaf node of the hypothesis tree as the cost of each path.
E3. Optionally, the cost of linear fitting of the velocity of each path in the hypothesis tree is calculated, corresponding to the matching mode that preferentially considers linear motion. The cost is added to the leaf nodes.
E4. The hypothesis tree is transformed into the largest Independent subset and the problem (MWIS) is solved. In the MWIS problem, each vertex is an assumed path, and an edge between the vertices corresponds to two vertices belonging to the same target, or within an n time period, at some moment, the two vertices share the same detection box, that is, the two vertices cannot coexist due to the conflict of the target or the detection box; the value of each vertex is the negative of the cost of each path; the MWIS problem is a proven NP-hard problem. By approximating the MWIS problem or solving it exactly, a set of selectable paths on the hypothesis tree can be obtained, so that the paths that can be selected by each target on the set of paths do not conflict with each other (do not match the same detection box at the same time), and each target has only one set of paths, and the cost of the set of paths is minimal.
E5. The hypotheses not on the set of paths before n frames are pruned out so that only one hypothesis is retained for each target before n frames. The assumption is that the target is at a position n frames before, and is used as an output result of the current moment of the algorithm.
Preferably, in the multi-target tracking method, as shown in fig. 5, the pruning optimization step includes (in the case of fig. 5, n is 3):
F1. deleting the new hypothesis, called the reserved child hypothesis, generated based on the detection box, except the MWIS solution, and the other hypotheses in the same layer as the parent hypothesis, called the parent hypothesis;
F2. optionally, for the remaining unmatched detection boxes in the proximity range, according to the matching distance, the direction of the speed of the matched target, and the difference between the size of the target and the size of the detection box, calculating the matching cost between the father hypothesis and the unmatched detection boxes, and performing matching by using a Hungarian algorithm.
F3. If there are still remaining detection boxes, then the parent hypothesis that is spatially closest to it is looked up and matched with it to generate a new child hypothesis. If any parent fails to match any of the detection boxes, an empty hypothesis is generated from the parent, and the state variables (position, velocity) of the empty hypothesis are the same as its parent.
F4. And taking the reserved sub-hypothesis and the newly generated sub-hypothesis as the hypothesis finally generated by the algorithm at the moment. It can be shown that after pruning optimization, the number of hypotheses is less than or equal to the number of hypotheses not subjected to pruning optimization, and the optimal solution can be found in the space after pruning all the time.
FIG. 8 is a block diagram illustrating a monitored scene multi-target tracking system 100, according to an embodiment of the invention. As shown in fig. 8, the following modules are included:
a detection frame receiving module 101, configured to receive a target detection frame of a current frame;
a hypothesis position generating module 102, configured to generate hypotheses about possible positions of each target based on the target detection frame of the current frame;
the target position calculation module 103 is used for solving the maximum independent subset and the problem based on all the hypotheses of the latest frames to obtain a mutually non-conflicting hypothesis path which can be selected by each target, and outputting the position of each target before the frames;
and the pruning optimization module 104 is used for carrying out pruning optimization by combining the maximum independent subset and the solution of the problem and the Hungarian algorithm to obtain a multi-target tracking result.
By the method and the system, a multi-hypothesis tracking algorithm with greatly reduced complexity is obtained, and the method and the system can be applied to various tracking scenes. Specifically, by introducing the Hungarian algorithm into the pruning result of the MWIS, the assumed tree is further pruned, so that the number of assumptions is greatly reduced, and the optimal solution cannot be lost in the pruning process. The algorithm has the advantages of higher accuracy, no occupation of GPU resources and capability of adjusting the balance between accuracy and complexity in the practical application of monitoring scenes.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.
Claims (10)
1. A multi-target tracking method for a monitored scene is based on multi-hypothesis tracking after pruning optimization, and is characterized by comprising the following steps:
receiving a target detection frame of a current frame;
generating hypotheses of possible positions of each target based on the target detection frame of the current frame;
solving the maximum independent subset and the problem based on all the hypotheses of a plurality of recent frames to obtain a mutually non-conflicting hypothesis path which can be selected by each target, and outputting the position of each target before the plurality of frames;
and (4) pruning optimization is carried out by combining the maximum independent subset and the solution of the problem and the Hungarian algorithm, so as to obtain a multi-target tracking result.
2. The multi-target tracking method for monitoring scenes as claimed in claim 1, wherein the step of receiving the target detection frame of the current frame comprises:
reading and decoding a monitoring video, and reading and decoding a required monitor into a single-frame picture through a disk or a network;
selecting a detector for detecting a target and a detection frame for extracting the target from a single-frame picture;
and transmitting the detection frame of the target to a tracker.
3. The multi-target tracking method for monitoring scenes as claimed in claim 2,
if the detector and the tracker are deployed at different server ends, transmitting the result of the detector to the server end where the tracker is located through a network data packet; if the detector and the tracker are deployed at the same server, the detection frame is transmitted through the process or the inter-thread traffic.
4. The multi-target tracking method for monitoring scenes as claimed in claim 1 or 2, wherein the generating of the hypothesis of the possible position of each target based on the detection frame of the current frame comprises:
finding detection boxes whose records are supposed to be adjacent: traversing all hypotheses of the current frame maintained at present, and searching and recording a detection frame adjacent to each hypothesis within a certain range;
new assumptions were generated: for each hypothesis and detection box match, a new hypothesis is generated corresponding to the matched detection box.
5. The multi-target tracking method for monitoring scenes as claimed in claim 4,
if there are no detection boxes in the proximity range, then generating an empty hypothesis, which does not match any detection box, and whose state variable is the same as its parent hypothesis;
if there are detection boxes within the proximity but all are above a certain threshold, then in addition to generating a hypothesis that matches the detection box, a null hypothesis is also generated.
6. The multi-target tracking method for monitoring scenes as claimed in any one of claims 1 to 5,
solving the maximum independent subset and the problem based on all the hypotheses of the latest frames to obtain a mutually non-conflicting hypothesis path which can be selected by each target, and outputting the position of each target before the frames, wherein the method comprises the following steps of:
calculating the cost of a new hypothesis of the current frame generated before according to the detection frame of the current frame;
calculating the cost of each path in the hypothesis tree in the plurality of frames;
converting the hypothesis tree into a maximum independent subset and solving a problem to obtain a group of selectable paths on the hypothesis tree which can be selected corresponding to each target and are not mutually conflicted;
for each target, only the hypothesis selected from the maximum independent subset and the solution to the problem is retained, and the remaining hypotheses are deleted, wherein the retained hypotheses correspond to positions of the target before the plurality of frames.
7. The monitored scene multi-target tracking method as claimed in claim 6, further comprising:
and calculating the cost of linear fitting of the speed of each path on the hypothesis tree, corresponding to the matching mode of linear motion, and adding the cost to the leaf nodes of the hypothesis tree.
8. The multi-target tracking method for monitoring scenes of claim 7,
the pruning optimization is carried out by combining the maximum independent subset and the solution of the problem and the Hungarian algorithm to obtain a multi-target tracking result, and the method comprises the following steps:
removing assumptions previously generated based on the detection box other than the maximum independent subset and the solution to the problem;
and for the remaining detection boxes which are not matched in the adjacent range, the Hungarian algorithm and the previous assumption are adopted for matching.
9. The monitored scene multi-target tracking method as claimed in claim 8, further comprising:
if the residual detection frames still exist, searching the parent hypothesis which is closest to the detection frames in space, and matching the parent hypothesis with the parent hypothesis to generate a new child hypothesis; if any parent hypothesis fails to match any detection box, generating a null hypothesis by the parent hypothesis, wherein the state variable of the null hypothesis is the same as the parent hypothesis;
the retained sub-hypotheses and the newly generated sub-hypotheses are used as final generated hypotheses.
10. A multi-target tracking system for a monitored scene comprises the following modules:
the detection frame receiving module is used for receiving a target detection frame of the current frame;
an assumed position generating module for generating an assumption of a possible position of each target based on the target detection frame of the current frame;
the target position calculation module is used for solving the maximum independent subset and the problem based on all the assumptions of the latest frames to obtain the mutually non-conflicting assumed path which can be selected by each target, and outputting the position of each target before the frames;
and the pruning optimization module is used for carrying out pruning optimization by combining the maximum independent subset and the solution of the problems and the Hungarian algorithm to obtain a multi-target tracking result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910713029.7A CN110659559B (en) | 2019-08-02 | 2019-08-02 | Multi-target tracking method and system for monitoring scene |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910713029.7A CN110659559B (en) | 2019-08-02 | 2019-08-02 | Multi-target tracking method and system for monitoring scene |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110659559A true CN110659559A (en) | 2020-01-07 |
| CN110659559B CN110659559B (en) | 2022-05-06 |
Family
ID=69036886
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910713029.7A Active CN110659559B (en) | 2019-08-02 | 2019-08-02 | Multi-target tracking method and system for monitoring scene |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110659559B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111462174A (en) * | 2020-03-06 | 2020-07-28 | 北京百度网讯科技有限公司 | Multi-target tracking method and device and electronic equipment |
| CN111812638A (en) * | 2020-07-17 | 2020-10-23 | 北京理工大学 | IMM-MHT multi-target tracking method |
| CN113158926A (en) * | 2021-04-26 | 2021-07-23 | 广东博智林机器人有限公司 | High-altitude parabolic track distinguishing method, device, equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130234882A1 (en) * | 2012-03-09 | 2013-09-12 | Raytheon Company | Multiple hypothesis tracking using maximum weight independent set |
| CN108303684A (en) * | 2018-01-31 | 2018-07-20 | 长沙深之瞳信息科技有限公司 | Ground surveillance radar multi-object tracking method based on radial velocity information |
-
2019
- 2019-08-02 CN CN201910713029.7A patent/CN110659559B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130234882A1 (en) * | 2012-03-09 | 2013-09-12 | Raytheon Company | Multiple hypothesis tracking using maximum weight independent set |
| CN108303684A (en) * | 2018-01-31 | 2018-07-20 | 长沙深之瞳信息科技有限公司 | Ground surveillance radar multi-object tracking method based on radial velocity information |
Non-Patent Citations (3)
| Title |
|---|
| WILLIAM BRENDEL: "Multiobject Tracking as Maximum Weight Independent Set", 《2011 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| 陈杭等: "关联深度自适应的多假设跟踪研究", 《系统工程与电子技术》 * |
| 陈莉等: "基于最大权重独立集的行人检测研究", 《宁波工程学院学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111462174A (en) * | 2020-03-06 | 2020-07-28 | 北京百度网讯科技有限公司 | Multi-target tracking method and device and electronic equipment |
| CN111462174B (en) * | 2020-03-06 | 2023-10-31 | 北京百度网讯科技有限公司 | Multi-target tracking method and device and electronic equipment |
| CN111812638A (en) * | 2020-07-17 | 2020-10-23 | 北京理工大学 | IMM-MHT multi-target tracking method |
| CN111812638B (en) * | 2020-07-17 | 2022-07-29 | 北京理工大学 | IMM-MHT multi-target tracking method |
| CN113158926A (en) * | 2021-04-26 | 2021-07-23 | 广东博智林机器人有限公司 | High-altitude parabolic track distinguishing method, device, equipment and storage medium |
| CN113158926B (en) * | 2021-04-26 | 2022-04-15 | 广东博智林机器人有限公司 | High-altitude parabolic track distinguishing method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110659559B (en) | 2022-05-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110659559B (en) | Multi-target tracking method and system for monitoring scene | |
| US11294955B2 (en) | System and method for optimization of audio fingerprint search | |
| Chen et al. | An equalized global graph model-based approach for multicamera object tracking | |
| Manen et al. | Pathtrack: Fast trajectory annotation with path supervision | |
| US10521672B2 (en) | Identifying and categorizing contextual data for media | |
| JP2018022475A (en) | Method and apparatus for updating background model | |
| CN114697128B (en) | Big data denoising method and big data acquisition system through artificial intelligence decision | |
| US20210158071A1 (en) | Match determination device, match determination method, storage medium | |
| US20210049517A1 (en) | Method and apparatus for generating a combined isolation forest model for detecting anomalies in data | |
| Chong et al. | Efficient multiple hypothesis tracking by track segment graph | |
| CN115330837A (en) | Robust target tracking method and system based on graph attention Transformer network | |
| CN111738042A (en) | Identification method, device and storage medium | |
| CN115510045A (en) | AI decision-based big data acquisition configuration method and intelligent scene system | |
| CN117456204A (en) | Target tracking method, device, video processing system, storage medium and terminal | |
| CN116545871A (en) | Multi-mode network traffic prediction method, device and medium | |
| Singh et al. | A greedy data association technique for multiple object tracking | |
| CN115187884A (en) | High-altitude parabolic identification method and device, electronic equipment and storage medium | |
| CN114077681B (en) | Image data processing method and device, computer equipment and storage medium | |
| CN115186781B (en) | Real-time fusion method for multi-source roadside traffic observation data | |
| Hu et al. | Modified multi-hypothesis tracking algorithm based on radar target appearance features | |
| Chen et al. | A Real-time Distributed Multi-camera Multi-object Tracking System | |
| CN115880615B (en) | Online multi-target tracking method based on fine-grained appearance representation | |
| CN107015253B (en) | Search detection method capable of dynamically predicting threat in navigation receiver | |
| Bouaziz | Visual SLAM with automatic map update in dynamic environments | |
| Luo et al. | Video sequence modeling by dynamic Bayesian networks: A systematic approach from coarse-to-fine grains |
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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: Room 101, building 1, block C, Qianjiang Century Park, ningwei street, Xiaoshan District, Hangzhou City, Zhejiang Province Applicant after: Hangzhou Weiming Information Technology Co.,Ltd. Applicant after: Institute of Information Technology, Zhejiang Peking University Address before: Room 288-1, 857 Xinbei Road, Ningwei Town, Xiaoshan District, Hangzhou City, Zhejiang Province Applicant before: Institute of Information Technology, Zhejiang Peking University Applicant before: Hangzhou Weiming Information Technology Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20200107 Assignee: Zhejiang smart video security Innovation Center Co.,Ltd. Assignor: Institute of Information Technology, Zhejiang Peking University Contract record no.: X2022330000930 Denomination of invention: A method and system of multi-target tracking in monitoring scene Granted publication date: 20220506 License type: Common License Record date: 20221229 |