Nothing Special   »   [go: up one dir, main page]

CN113658222A - Vehicle detection tracking method and device - Google Patents

Vehicle detection tracking method and device Download PDF

Info

Publication number
CN113658222A
CN113658222A CN202110882079.5A CN202110882079A CN113658222A CN 113658222 A CN113658222 A CN 113658222A CN 202110882079 A CN202110882079 A CN 202110882079A CN 113658222 A CN113658222 A CN 113658222A
Authority
CN
China
Prior art keywords
vehicle
target detector
iou
training
yolo
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
Application number
CN202110882079.5A
Other languages
Chinese (zh)
Inventor
吉长江
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Yingpu Technology Co Ltd
Original Assignee
Shanghai Yingpu Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai Yingpu Technology Co Ltd filed Critical Shanghai Yingpu Technology Co Ltd
Priority to CN202110882079.5A priority Critical patent/CN113658222A/en
Publication of CN113658222A publication Critical patent/CN113658222A/en
Pending legal-status Critical Current

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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • 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
    • G06T2207/30241—Trajectory

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Multimedia (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Evolutionary Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)

Abstract

The application discloses a vehicle detection tracking method and device. The method comprises the following steps: training a Yolo target detector; detecting the type and the position of a vehicle in each frame of the video by using a trained Yolo target detector; the output of the Yolo target detector is matched with the IOU in the historic vehicle using the IOU tracker and assigned a unique ID. The device comprises: a training module configured to train a Yolo target detector; a detection module configured to detect the type and location of the vehicle in each frame of the video using a trained Yolo target detector; and a matching module configured to match the output of the Yolo target detector with the IOU in the historical vehicle using the IOU tracker and assign a unique ID.

Description

Vehicle detection tracking method and device
Technical Field
The application relates to the field of vehicle motion behavior recognition in videos, in particular to track recognition in vehicle motion behavior recognition.
Background
Trajectory recognition is the basis for vehicle motion behavior recognition. Through the vehicle trajectory, the road manager can acquire information about abnormal vehicle behavior, such as illegal lane change, abnormal parking, and the like. Trajectory recognition can be described simply as a fusion of object detection and object tracking. The main tasks in vehicle trajectory identification can be described as follows: -initiating and detecting a vehicle trajectory in motion, -tracking the vehicle trajectory using the coordinates and estimation of motion parameters, -identifying the category of the observed vehicle. With the rapid development of convolutional neural network technology and target detection technology, the mainstream research on track identification focuses on a method of tracking according to detection, and these track identification algorithms track vehicles by using the output of target detection, thereby achieving the balance between speed and accuracy. Object detection aims at locating instances of semantic objects of a particular class in a given image. Before the deep learning era, its typical detection process can be divided into three steps: proposal generation, feature vectorization and classification. Deep convolutional neural network with deep learning development
The success of the network (DCNN), object detection, has also introduced CNN as its backbone network. Today, object detection based on deep learning can be generally divided into two series: a two-stage detector and a one-stage detector, the most representative of the two-stage detector being fast R-CNN; the first detector is most typically YOLO and SSD. The two-stage detector first generates and extracts features from a proposed set using an architecture called the Region Proposal Network (RPN), and then applies a region classifier to predict the proposed classes. The framework of the two-stage detector is intended to achieve high localization and target recognition accuracy. The primary detector skips the region proposal step and detects directly on a dense sample of possible locations.
There are two methods of visual object tracking: a generative model method and a discriminative model method. The model generation method models a target region in a current frame and finds a region most similar to the model in the next frame, typical algorithms comprise a Kalman filter, a particle filter and mean shift, the classical methods cannot process and adapt to complex tracking change, and the robustness and the accuracy of the classical methods are surpassed by the current algorithm. The discriminant model method is also called "detection tracking" method, and unlike object detection, the output of object detection is a bounding box with its class, with an ID also associated to each box by the detection tracking method.
The method has the advantages that the prediction accuracy is low, the method is sensitive to the fluctuation of the vehicle track, the problem that the prediction error difference at different time points is large cannot be solved when a certain model is used alone, the model construction process is too complex, the time cost is relatively high, the method is sensitive to the change of noise data, the prediction error is increased continuously along with the increase of noise, and the method is approximately linear.
Disclosure of Invention
It is an object of the present application to overcome the above problems or to at least partially solve or mitigate the above problems.
According to one aspect of the present application, there is provided a vehicle detection tracking method, including:
training a Yolo target detector;
detecting the type and the position of a vehicle in each frame of the video by using a trained Yolo target detector;
the output of the Yolo target detector is matched with the IOU in the historic vehicle using the IOU tracker and assigned a unique ID.
Optionally, the training of the Yolo target detector comprises:
data enhancement and focus loss training are performed on the Yolo target detector.
Optionally, before the training of the Yolo target detector, the method further comprises:
the original training data set is augmented by random geometric transformations and random color dithering.
Optionally, the focus loss is used as the object confidence in the process of training the Yolo target detector and detecting the type and position of the vehicle in each frame of the video by using the trained Yolo target detector.
Optionally, during the process of training the Yolo target detector, the Yolo batch processing of each GPU is synchronized.
Optionally, the matching the output of the Yolo target detector and the IOU in the historical vehicle and assigning a unique ID with the IOU tracker includes:
for each frame of the current frame, respectively calculating an F-IOU between the frame and each frame in the previous frame, wherein the F-IOU is a combined upper intersection between the two frames, and recording vehicles in the two frames corresponding to the largest F-IOU as the same vehicle to form a track of the vehicle;
calculating the F-IOU of each vehicle in the current frame and the last frames of each track;
when the F-IOU between a vehicle and a trajectory is at a maximum and exceeds a threshold, the vehicle is added to the trajectory, otherwise a new trajectory is created with the vehicle and added to the set of trajectories.
According to another aspect of the present application, there is provided a vehicle detection tracking apparatus including:
a training module configured to train a Yolo target detector;
a detection module configured to detect the type and location of the vehicle in each frame of the video using a trained Yolo target detector; and
a matching module configured to match the output of the Yolo target detector with the IOU in the historical vehicle using the IOU tracker and assign a unique ID.
Optionally, in the training module, training the Yolo target detector includes:
data enhancement and focus loss training are performed on the Yolo target detector.
Optionally, in the training module and the detection module, the focus loss is used as an object confidence.
Optionally, in the training module, the Yolo batch processing of each GPU is synchronized.
According to the vehicle detection and tracking method and device, the Yolo target detector and the IOU tracker are integrated, so that the speed and the precision of detection and tracking can be improved, the prediction error is reduced, and stable detection and tracking can be performed on the vehicle.
The above and other objects, advantages and features of the present application will become more apparent to those skilled in the art from the following detailed description of specific embodiments thereof, taken in conjunction with the accompanying drawings.
Drawings
Some specific embodiments of the present application will be described in detail hereinafter by way of illustration and not limitation with reference to the accompanying drawings. The same reference numbers in the drawings identify the same or similar elements or components. Those skilled in the art will appreciate that the drawings are not necessarily drawn to scale. In the drawings:
FIG. 1 is a schematic flow chart diagram of a vehicle detection and tracking method according to one embodiment of the present application;
FIG. 2 is a diagram of computing an F-IOU using two frames, where t represents a time instant, according to one embodiment of the present application;
FIG. 3 is a schematic diagram of a vehicle detection and tracking device according to an embodiment of the present application;
FIG. 4 is a schematic block diagram of a computing device according to one embodiment of the present application;
fig. 5 is a schematic structural diagram of a computer-readable storage medium according to an embodiment of the present application.
Detailed Description
FIG. 1 is a schematic flow chart diagram of a vehicle detection and tracking method according to one embodiment of the present application. The vehicle detection tracking method may generally include the following steps S1 to S3.
And step S1, training the Yolo target detector.
The present embodiment uses Yolo for vehicle detection. Yolo applies a single neural network to the complete image, divides the image into 13 x 13 regions (when the input size is 416 x 416), and predicts the bounding box and probability for each region, which are weighted by the prediction probability. The unique architecture makes Yolo extremely Fast, more than 100 times faster than two-level detectors such as Fast R-CNN, so the present embodiment selects Yolo as the detector in the vehicle trajectory recognition algorithm, trains the Yolo object detector using the COCO data set and MIO-TCD data set to detect the type and position of the vehicle in each frame of the video, and improves the accuracy of Yolo from 3 aspects: data expansion, focus loss, and synchronous batch normalization.
(1) Data expansion
Data augmentation can significantly improve the diversity of training data without actually collecting new data. The data expansion used in the present embodiment includes random geometric transformations, random color dithering. For the first method, it can be implemented by performing random translation, random rotation, random scaling and random reflection on the original training data. For the second approach, this may be achieved by processing the raw training data using random HSV saturation values and random HSV intensity values.
(2) Loss of focus
The present embodiment takes advantage of the loss of focus to solve the problem of unbalanced data sets. If the number of samples for one category in the dataset is much larger than the number of samples for the other categories, this dataset will be considered as an unbalanced dataset, and unbalanced data will lead to inefficient training. In Yolo, the imbalance occurs in the object confidence. In calculating the object confidence loss, (13 × 13+26 + 52) × 3 ═ 10647 object boxes will be presented, with most of the boxes not matching the real object. Thus, changing the loss of object confidence from cross entropy to focus loss will suppress the imbalance. Conventional cross entropy loss CE (p)t) Expressed as: CE (p)t)＝-log(pt) Wherein p istIs the predicted probability that the sample is in the true category. Negative samples have a greater impact on the loss, resulting in a network bias towards the background category. And the focal point loss FL (p)t) Is defined as: FL (p)t)＝-(1-pt)γlog(pt) Where γ is the focusing parameter, when ptApproaching 1 or 0 means that the network is very deterministic to the prediction, where the focus loss is much lower than the cross entropy loss; otherwise, the focus loss will be more than ptThe focus loss is slightly lower near 1 or 0.
(3) Synchronous batch normalization
Batch normalization allows for the use of higher learning rates with less concern for weight initialization. When multiple devices (typically GPUs) are used to train the model, the standard implementation of batch normalization in a common framework is asynchronous, with data being normalized individually within each GPU, rather than globally. This problem does not plague a simple network that occupies little RAM, each GPU can carry a large batch size, but the Yolo batch size of each GPU is small (about 16 per GPU), at which point, synchronous batch normalization is very important. In the embodiment, a special function transformation method for the numerical value is used, that is, assuming that some original numerical value is x, a function with normalization function is sleeved on the original numerical value, and the numerical value x before normalization is converted to form a normalized numerical value.
And step S2, detecting the type and the position of the vehicle in each frame of the video by using the trained Yolo target detector.
Step S3, matching the output of the Yolo target detector and the IOU in the historical vehicle by the IOU tracker and assigning a unique ID.
The associative above Intersection (IOU) is the ratio between the intersection and union of the prediction box and the ground truth box, expressed as
Figure BDA0003192740290000051
Where area (a) represents a prediction box and area (b) represents a ground true box, the IOU between two frames is calculated instead of the true and prediction boxes, and is referred to as the F-IOU. For each frame in the current frame, calculating the F-IOU between the frame and all frames in the previous frame, regarding the two frames corresponding to the largest F-IOU in the calculation result as the same object, and allocating the same ID to the current frame and the previous frame. For the current frame, the IOU tracker will calculate the F-IOU for each vehicle (as shown in FIG. 2) and the last few boxes for each trajectory if vehicle v and trajectory t1Is maximum and exceeds a threshold, then vehicle v is added to trajectory t1Otherwise, a new trajectory t is created with the vehicle v2And join the trace set. Furthermore, the track set needs to be updated every frame. The trajectory without any vehicle added in the consecutive 5 frames will be deleted to eliminate the interference of the past trajectory. In addition, the trajectory set needs to be updated every frame, and the trajectory without any vehicle added in the consecutive 5 frames will be deleted to eliminate the interference of the historical trajectory.
In summary, the trajectory identification process of the vehicle detection and tracking method of the embodiment uses Yolo as a detector, and uses the IOU tracker to identify the vehicle and its trajectory. After obtaining the vehicle trajectory, the road administrator may further monitor and analyze the movement of the vehicle, including the following aspects:
(1) recognizing lane change, converting the monitoring image into a bird's-eye view through affine transformation, then detecting the change of the processed vehicle track, and recognizing the lane change from the monitoring video to provide legal basis for traffic monitoring and data support for analyzing microscopic traffic flow;
(2) the abnormal parking of the vehicle is identified, whether the current driving state of the vehicle is normal or not can be determined by analyzing the motion information of the vehicle track, and the reliable abnormal parking early warning information can be provided for monitoring personnel by identifying the abnormal parking of the vehicle;
(3) by combining the idea of re-identifying the vehicle, the same vehicle track under different cameras can be stitched, and then the vehicle running track under continuous and dense monitoring can be drawn.
The gradual detection and tracking method of the Yolo target detector and the IOU tracker in the embodiment, which are fused together, can improve the speed and precision of detection and tracking, reduce the prediction error, stably detect and track the vehicle and alleviate the problem of Yolo gradient disappearance.
Fig. 3 is a schematic structural diagram of a vehicle detection and tracking device according to an embodiment of the present application. The vehicle detection and tracking device may generally include:
a training module 1 configured to train a Yolo target detector;
a detection module 2 configured to detect the type and position of the vehicle in each frame of the video using a trained Yolo target detector; and
a matching module 3 configured to match the output of the Yolo target detector with the IOU in the historic vehicle using the IOU tracker and assign a unique ID.
As a preferred embodiment of the present application, in the training module 1, training the Yolo target detector includes:
data enhancement and focus loss training are performed on the Yolo target detector.
In the training module 1 and the detection module 2, the focus loss is used as the object confidence.
As a preferred embodiment of the present application, in the training module 1, the Yolo batch processing of each GPU is synchronized.
Embodiments also provide a computing device, referring to fig. 4, comprising a memory 1120, a processor 1110 and a computer program stored in said memory 1120 and executable by said processor 1110, the computer program being stored in a space 1130 for program code in the memory 1120, the computer program, when executed by the processor 1110, implementing the method steps 1131 for performing any of the methods according to the invention.
The embodiment of the application also provides a computer readable storage medium. Referring to fig. 5, the computer readable storage medium comprises a storage unit for program code provided with a program 1131' for performing the steps of the method according to the invention, which program is executed by a processor.
The embodiment of the application also provides a computer program product containing instructions. Which, when run on a computer, causes the computer to carry out the steps of the method according to the invention.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed by a computer, cause the computer to perform, in whole or in part, the procedures or functions described in accordance with the embodiments of the application. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
Those of skill would further appreciate that the various illustrative components and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
It will be understood by those skilled in the art that all or part of the steps in the method for implementing the above embodiments may be implemented by a program, and the program may be stored in a computer-readable storage medium, where the storage medium is a non-transitory medium, such as a random access memory, a read only memory, a flash memory, a hard disk, a solid state disk, a magnetic tape (magnetic tape), a floppy disk (floppy disk), an optical disk (optical disk), and any combination thereof.
The above description is only for the preferred embodiment of the present application, but the scope of the present application 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 application should be covered within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (10)

1. A vehicle detection tracking method, comprising:
training a Yolo target detector;
detecting the type and the position of a vehicle in each frame of the video by using a trained Yolo target detector;
the output of the Yolo target detector is matched with the IOU in the historic vehicle using the IOU tracker and assigned a unique ID.
2. The method of claim 1, wherein the training a Yolo target detector comprises:
data enhancement and focus loss training are performed on the Yolo target detector.
3. The method of claim 2, wherein prior to the training of the Yolo target detector, the method further comprises:
the original training data set is augmented by random geometric transformations and random color dithering.
4. The method according to any one of claims 1 to 3, wherein the focus loss is taken as an object confidence in the process of training the Yolo target detector and detecting the type and position of the vehicle in each frame of the video by using the trained Yolo target detector.
5. The method according to any one of claims 1-4, wherein during the training of the Yolo target detector, the Yolo batch processing of each GPU is synchronized.
6. The method of any one of claims 1-5, wherein matching and assigning unique IDs to outputs of a Yolo target detector and IOUs in historical vehicles using an IOU tracker comprises:
for each frame of the current frame, respectively calculating an F-IOU between the frame and each frame in the previous frame, wherein the F-IOU is a combined upper intersection between the two frames, and recording vehicles in the two frames corresponding to the largest F-IOU as the same vehicle to form a track of the vehicle;
calculating the F-IOU of each vehicle in the current frame and the last frames of each track;
when the F-IOU between a vehicle and a trajectory is at a maximum and exceeds a threshold, the vehicle is added to the trajectory, otherwise a new trajectory is created with the vehicle and added to the set of trajectories.
7. A vehicle detection tracking apparatus, comprising:
a training module configured to train a Yolo target detector;
a detection module configured to detect the type and location of the vehicle in each frame of the video using a trained Yolo target detector; and
a matching module configured to match the output of the Yolo target detector with the IOU in the historical vehicle using the IOU tracker and assign a unique ID.
8. The apparatus of claim 7, wherein the training module to train the Yolo target detector comprises:
data enhancement and focus loss training are performed on the Yolo target detector.
9. The apparatus of claim 7 or 8, wherein the training module and the detection module use focus loss as object confidence.
10. The apparatus of any of claims 7 to 9, wherein the Yolo batch processing of each GPU is synchronized in the training module.
CN202110882079.5A 2021-08-02 2021-08-02 Vehicle detection tracking method and device Pending CN113658222A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110882079.5A CN113658222A (en) 2021-08-02 2021-08-02 Vehicle detection tracking method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110882079.5A CN113658222A (en) 2021-08-02 2021-08-02 Vehicle detection tracking method and device

Publications (1)

Publication Number Publication Date
CN113658222A true CN113658222A (en) 2021-11-16

Family

ID=78490271

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110882079.5A Pending CN113658222A (en) 2021-08-02 2021-08-02 Vehicle detection tracking method and device

Country Status (1)

Country Link
CN (1) CN113658222A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114501061A (en) * 2022-01-25 2022-05-13 上海影谱科技有限公司 Video frame alignment method and system based on object detection
CN114972410A (en) * 2022-06-16 2022-08-30 上海影谱科技有限公司 Multi-level matching video racing car tracking method and system

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108470353A (en) * 2018-03-01 2018-08-31 腾讯科技（深圳）有限公司 A kind of method for tracking target, device and storage medium
US20190130189A1 (en) * 2017-10-30 2019-05-02 Qualcomm Incorporated Suppressing duplicated bounding boxes from object detection in a video analytics system
CN110378210A (en) * 2019-06-11 2019-10-25 江苏大学 A kind of vehicle and car plate detection based on lightweight YOLOv3 and long short focus merge distance measuring method
CN111445501A (en) * 2020-03-25 2020-07-24 苏州科达科技股份有限公司 Multi-target tracking method, device and storage medium
CN111508002A (en) * 2020-04-20 2020-08-07 北京理工大学 Small-sized low-flying target visual detection tracking system and method thereof
CN111932580A (en) * 2020-07-03 2020-11-13 江苏大学 Road 3D vehicle tracking method and system based on Kalman filtering and Hungary algorithm
US20210065384A1 (en) * 2019-08-29 2021-03-04 Boe Technology Group Co., Ltd. Target tracking method, device, system and non-transitory computer readable storage medium
CN112884816A (en) * 2021-03-23 2021-06-01 武汉理工大学 Vehicle feature deep learning recognition track tracking method based on image system
CN113129336A (en) * 2021-03-31 2021-07-16 同济大学 End-to-end multi-vehicle tracking method, system and computer readable medium

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190130189A1 (en) * 2017-10-30 2019-05-02 Qualcomm Incorporated Suppressing duplicated bounding boxes from object detection in a video analytics system
CN108470353A (en) * 2018-03-01 2018-08-31 腾讯科技（深圳）有限公司 A kind of method for tracking target, device and storage medium
CN110378210A (en) * 2019-06-11 2019-10-25 江苏大学 A kind of vehicle and car plate detection based on lightweight YOLOv3 and long short focus merge distance measuring method
US20210065384A1 (en) * 2019-08-29 2021-03-04 Boe Technology Group Co., Ltd. Target tracking method, device, system and non-transitory computer readable storage medium
CN111445501A (en) * 2020-03-25 2020-07-24 苏州科达科技股份有限公司 Multi-target tracking method, device and storage medium
CN111508002A (en) * 2020-04-20 2020-08-07 北京理工大学 Small-sized low-flying target visual detection tracking system and method thereof
CN111932580A (en) * 2020-07-03 2020-11-13 江苏大学 Road 3D vehicle tracking method and system based on Kalman filtering and Hungary algorithm
CN112884816A (en) * 2021-03-23 2021-06-01 武汉理工大学 Vehicle feature deep learning recognition track tracking method based on image system
CN113129336A (en) * 2021-03-31 2021-07-16 同济大学 End-to-end multi-vehicle tracking method, system and computer readable medium

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114501061A (en) * 2022-01-25 2022-05-13 上海影谱科技有限公司 Video frame alignment method and system based on object detection
CN114501061B (en) * 2022-01-25 2024-03-15 上海影谱科技有限公司 Video frame alignment method and system based on object detection
CN114972410A (en) * 2022-06-16 2022-08-30 上海影谱科技有限公司 Multi-level matching video racing car tracking method and system

Similar Documents

Publication Publication Date Title
CN108470332B (en) Multi-target tracking method and device
Jana et al. YOLO based Detection and Classification of Objects in video records
CN112016464B (en) Method and device for detecting face shielding, electronic equipment and storage medium
AU2019343959B2 (en) Region proposal with tracker feedback
US11106903B1 (en) Object detection in image data
US20180114071A1 (en) Method for analysing media content
Li et al. Multi-object recognition method based on improved yolov2 model
CN113658222A (en) Vehicle detection tracking method and device
Zhang et al. A comparative analysis of object detection algorithms in naturalistic driving videos
Xu et al. Jointly modeling association and motion cues for robust infrared UAV tracking
Yuan et al. Multiple object detection and tracking from drone videos based on GM-YOLO and multi-tracker
Rishika et al. Real-time vehicle detection and tracking using YOLO-based deep sort model: A computer vision application for traffic surveillance
AFFES et al. Comparison of YOLOV5, YOLOV6, YOLOV7 and YOLOV8 for Intelligent Video Surveillance.
Nauman et al. Identification of Anomalous Behavioral Patterns in Crowd Scenes.
CN112133100B (en) Vehicle detection method based on R-CNN
Vashishtha et al. Vehicle Detection System using YOLOv4
Kim et al. Development of a real-time automatic passenger counting system using head detection based on deep learning
Meena et al. Hybrid neural network architecture for multi-label object recognition using feature fusion
Geng et al. Pedestrian Fall Event Detection in Complex Scenes Based on Attention‐Guided Neural Network
CN111985378A (en) Road target detection method, device and equipment and vehicle
US11875526B1 (en) Method for training object detector capable of predicting center of mass of object projected onto ground, method for recognizing identical object in specific space captured from a plurality of cameras having different viewing frustums using trained object detector, and learning device and object recognizing device using the same
Shen et al. Vehicle detection based on improved YOLOv5s using coordinate attention and decoupled head
US11955272B1 (en) Method for generating object detector based on deep learning capable of detecting extended object class and detector generation device using the same
CN112446300B (en) Method, system and computer storage medium for traffic density analysis
Wang et al. Vehicle trajectory recognition based on video object detection

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