1385 results about "Multi target tracking" patented technology

The invention relates to the technical field of image processing, and discloses an intelligent machine vision detection method and system based on image processing, and a storage medium. The method comprises the steps of performing candidate region generation and geometric constraint optimization processing on fusion features through a cascade bidirectional region suggestion network to obtain a target candidate region, and performing instance segmentation processing on the target candidate region through a graph convolutional network and a boundary dynamic adjustment mechanism to obtain a segmentation result; performing feature extraction processing on the segmentation result through a space-time joint attention network to obtain space-time context features; performing multi-target tracking processing on the spatio-temporal context features through depth correlation measurement and motion mode matching to obtain a trajectory prediction result; and performing abnormal mode modeling processing on the trajectory prediction result through hierarchical contrast learning and multi-modal feature fusion to obtain abnormal features, and generating a visual inspection report according to the abnormal features. According to the invention, the accuracy of intelligent machine vision detection based on image processing is improved.

The invention relates to the technical field of multi-target tracking, in particular to a dynamic multi-target tracking and trajectory prediction method. Comprising the following steps: acquiring a color image and a depth image, and preprocessing; constructing a dynamic threshold double-flow neural network, extracting features from the preprocessed color image and depth image to obtain a visual feature set and a depth feature set, fusing the visual feature set and the depth feature set through an adaptive attention mechanism to obtain a fused feature set, and combining a dynamic threshold to obtain an observation information list of a target; updating the observation information list through a self-motion decoupling mechanism to obtain a real information list; establishing space association and time association between the target and the track based on the real information list to obtain a newest track set; and analyzing target characteristics, environmental constraints, social behaviors and historical trajectories, generating a prediction trajectory of each target and a corresponding multi-factor score, generating a time-varying influence graph, and performing planning and decision making by adopting a double-layer decision-making mechanism.

The invention provides a multi-target tracking detection method and system based on cascade matching and trajectory confirmation, and relates to the technical field of digital video monitoring and tracking, and the method comprises the steps: obtaining a personnel construction video, and decomposing a video frame into a plurality of images; identifying a target in the current frame of image based on an improved YOLOv5 network, obtaining a target detection frame, and obtaining a prediction trajectory frame of the current frame through a Kalman filtering algorithm based on the trajectory frame of the previous frame; performing multi-target tracking on the target detection frame and the prediction trajectory frame of the current frame by using a DeepSort algorithm, respectively obtaining motion features and appearance features of the target detection frame and the prediction trajectory frame, performing cascade matching on the motion features and the appearance features of the target detection frame and the prediction trajectory frame by using a Hungary algorithm, and performing matching optimization by using an IOU matching algorithm; and carrying out fusion updating on the successfully matched target detection frame and prediction trajectory frame by using a Kalman filtering algorithm, and replacing the original prediction trajectory frame to obtain a tracking result.

The invention discloses an event and RGB image fused SNN target tracking method and system, and the method comprises the steps: converting an event flow from an event camera into an event frame image through a representation method based on time stacking, and enabling the event frame image to be aligned with an RGB frame image from an RGB camera in time and space; performing target detection on the event frame image # imgabs0 # and the RGB frame image # imgabs1 # after alignment by using a spiking neural network SNN, wherein the spiking neural network SNN comprises a feature extraction module, a feature fusion module and a detection head; and learning the similarity of the targets by using the anchor frame extracted by the spiking neural network SNN and the fusion feature map # imgabs2 # to perform multi-target tracking. The invention aims to realize fusion of the fusion event and the RGB image so as to ensure high precision and high efficiency of multi-target tracking, and keep stable detection performance and tracking performance in static and high-motion scenes.

The invention discloses an intelligent power distribution room safety monitoring management system, which comprises the following steps: checking states of a sensor, a gateway and a camera, and collecting environment data through the sensor; preprocessing the collected data, identifying a smoke area by using a YOLOv5 model, and starting a linkage safety mechanism according to abnormal environmental parameters; performing personnel detection by using a YOLOv5s model, capturing and analyzing a behavior track through a SimAM mechanism, and realizing multi-target tracking and abnormal behavior detection; grading the abnormal events based on a rule classifier, and triggering equipment linkage control; and storing the data in an NoSQL database, analyzing a trend to predict potential dangers, and optimizing a YOLOv5 model through historical data. The system can monitor the environment and personnel safety conditions in the power distribution room in real time, has efficient response capability and a trend prediction function, and can be widely applied to the field of safety monitoring of intelligent power grids and industrial environments.

The invention belongs to the field of outdoor AGV obstacle tracking, and particularly relates to an AGV obstacle tracking method based on multi-modal data and interactive Kalman filtering. Comprising the following steps: calibrating a monocular camera and a laser radar, and establishing point cloud data and image data with synchronous timestamps; performing preprocessing and target detection operation on the point cloud data and the target detection frame to obtain preprocessed point cloud data and a generated target detection frame so as to obtain obstacle category information; acquiring an obstacle observation position according to the preprocessed point cloud data and the target detection frame; meanwhile, an improved SORT multi-target tracking algorithm is adopted, and a historical track of the obstacle is established; through a plurality of Kalman filtering fusion methods, a trajectory prediction result is obtained through prediction; finally, the AGV communicates with a dispatching center, global information is obtained, and the track trend is corrected in a targeted mode in combination with obstacle category information; and the historical track of the obstacle and the corrected prediction track are sent to the AGV for obstacle avoidance path planning.

The invention discloses an automatic driving path optimization control method and device fusing environmental perception and prediction. The method comprises the following steps: acquiring environmental perception data of an automatic driving vehicle through a radar and a camera; the initial posture, the target posture and the environment perception data of the automatic driving vehicle serve as road constraint conditions, and a preliminary driving path of the automatic driving vehicle from the current position to the target position is planned through a path planning algorithm; state estimation and prediction are carried out on surrounding vehicles and obstacles through a multi-target tracking algorithm; according to future motion trails of surrounding vehicles and obstacles, road constraint conditions and physical restrictions of the autonomous vehicle, performing path optimization on the initial path to obtain an optimized driving path; and converting the optimized driving path into a control instruction of the automatic driving vehicle, and driving the automatic driving vehicle to stably drive along the optimized driving path through the control instruction. According to the invention, the safety and the driving efficiency of the automatic driving vehicle are further improved.

The invention discloses a construction site unsafe behavior detection method and system based on image recognition, and relates to the technical field of intelligent visual safety monitoring, and the method comprises the steps: collecting a construction image containing a constructor, carrying out the rectangular frame labeling of unsafe behaviors in the construction image, and generating a structured labeling data set; inputting the structured annotation data set into a YOLOv5 target detection model, performing training through Mosaic data enhancement and an adaptive anchor box generation algorithm, and outputting the trained YOLOv5 target detection model; based on the trained YOLOv5 target detection model, performing 128-dimensional feature vector extraction and track association matching on a detected target through a DeepSORT multi-target tracking algorithm, and outputting a track data table; and the recognition stability and behavior analysis depth of the system in a complex environment are enhanced, so that automation and intelligentization of safety management of a construction site are effectively supported.

The invention discloses a cross-border tracking traffic early warning identification method based on multi-array cameras, and the method comprises the following steps: S1, deploying the multi-array cameras, a laser radar and an auxiliary sensor, and collecting original image data; s2, preprocessing the collected original image data and the point cloud data collected by the laser radar to generate a multi-modal data set; s3, constructing an improved Deformable DETR model, performing feature extraction on the multi-modal data set by using the improved Deformable DETR model, identifying the category and the position of a target, and generating a detection frame and a confidence score; s4, according to a target detection result, constructing a multi-target tracking framework, and extracting a motion trail of the target; and S5, analyzing the movement track of the target, extracting the movement characteristics of the target, and detecting an abnormal traffic event. According to the method, the improved Deformable DETR model is combined with multi-modal data fusion, accurate detection and intelligent tracking of the cross-border target are achieved, and the method has the advantages of being high in detection accuracy, high in tracking stability and accurate in anomaly recognition.

Various examples, systems, and methods are disclosed relating to dynamic novel view reconstruction based at least in part on flow rematching. A first computing system can update a graph neural network based at least on video data representing a plurality of first objects and a plurality of first labels corresponding to the plurality of first objects. The first computing system can cause the graph neural network to generate a plurality of second labels of a first example video and update the graph neural network based at least on the plurality of second labels and the first example video. The first computing system can cause the graph neural network to generate a plurality of third labels of a second example video. The first computing system can output a request for a modification to the at least one third label responsive to the uncertainty score satisfying an annotation criterion.

The invention discloses a multi-target tracking method based on weak clue and trajectory prediction, and belongs to the technical field of deep learning multi-target tracking. The method comprises the following steps: firstly, setting a target detection model, then setting a target tracking algorithm, predicting a future trajectory of a target by using a position velocity-LSTM model when shielding occurs, and calculating the movement direction cost; if not, track prediction is carried out by adopting Kalman filtering; when the position cost is calculated, appearance features are extracted by using an appearance feature model on the basis of a designed mixed height intersection-to-union ratio using a weak clue, namely a height state, and after each clue is obtained, trajectory distribution is carried out on a target by using strong and weak clues through a Hungary algorithm. In order to balance and utilize each cost, multi-target tracking is realized through three times of association based on a configured cost allocation strategy. According to the method, the problems of shielding, similar objects, appearance degradation and the like are solved to a certain extent, and the model generalization ability and the target tracking ability to complex scenes are enhanced.

The invention discloses a bird flight path intelligent tracking and monitoring method and system based on big data analysis. The method comprises the following steps: S1, generating an original bird observation data set; s2, outputting a standardized bird observation data set; s3, calling a multi-target tracking algorithm based on the standardized bird observation data set to obtain a bird trajectory fragment data set; s4, constructing a lion group optimization search model, and inputting the bird trajectory fragment data set into the lion group optimization search model to form a candidate trajectory data set; s5, inputting the candidate trajectory data set into a space-time Transform model configured with rotation position embedding, and generating a continuous flight path data set; and S6, outputting a cross-region bird migration monitoring result set. According to the method, bird migration continuous tracking requirements under different time periods, multiple geographical zones and heterogeneous equipment acquisition conditions can be covered, application scenes of ecological protection, migration research and airspace management and control are effectively supported, and the method has remarkable engineering popularization and ecological protection values.

The invention discloses a self-adaptive illegal parking detection method, a detection system and a storage medium. The self-adaptive illegal parking detection method comprises the following steps: initializing the system; each video frame in an input video stream is processed according to the following steps: vehicle detection; performing regional filtration; performing multi-target tracking; vehicle state calculation: traversing each tracked vehicle, and updating the information of the tracked vehicle in the vehicle information object; multi-dimensional illegal parking judgment and alarm: according to a calculation result in the parking duration calculation step, executing the following judgment logics: traversing all static vehicles, determining a scene, obtaining a dynamic threshold value, triggering condition judgment and triggering an action; result visualization and output are carried out; and cleaning resources. By introducing the multi-dimensional dynamic judgment logic, the technical scheme of the invention can identify illegal parking behaviors more intelligently and more accurately, and the practicability and reliability of the system are significantly improved.

The invention provides a multi-target tracking algorithm of a wide-view-angle feature fusion memory network, and belongs to the technical field of computer vision and artificial intelligence. The technical problem that the tracking performance is insufficient when a long-distance small-size target and a fast moving target are processed under the view angle of the unmanned aerial vehicle is solved. According to the technical scheme, the method comprises the following steps: S1, video frame preprocessing and target detection; s2, carrying out deep fusion on the picture feature information in a wide-view fusion memory network; s3, track confidence modeling is introduced in the filtering process of adaptive motion filtering AMF; and S4, in the first step of association, performing first association through a motion feature extraction model MFEM. The method has the advantages that the target position is adjusted and optimized step by step through dynamic updating of a tracking algorithm, and finally information of all tracks of the target is obtained.

The invention provides a lottery store violation detection method and system based on multi-modal data fusion, and relates to the technical field of intelligent monitoring, and the method comprises the steps: detecting a current abnormal event, carrying out time sequence perception target detection on monitoring video data, and recognizing violation electric equipment in a video image frame based on adaptive feature enhancement and multi-target tracking. The method comprises the following steps: acquiring time sequence data of current abnormity and illegal electric equipment detection, constructing a time sequence incidence relation of current abnormity, equipment detection and an electric state based on a dynamic causal network, calculating a time-varying weight and instantaneous causal intensity by utilizing conditional entropy increment, acquiring optimal time lag in combination with eigenvector conversion and a dynamic programming algorithm, and determining the current abnormity and illegal electric equipment detection according to the optimal time lag. And weighting the integral of the instantaneous causal intensity and the exponential function of the optimal time delay to obtain an event matching score, and distinguishing a temporary power utilization event and a continuous illegal power utilization event based on the event matching score.

The invention discloses a radar and vision multi-mode fused three-dimensional multi-target tracking method and related equipment. The method comprises the following steps: establishing a three-dimensional constant turning rate and speed motion model for detecting a target vehicle; constructing a multi-modal measurement model, and obtaining a fusion target measurement result according to the point cloud data and the image data collected by the 4D millimeter wave radar; performing two-stage front and back frame matching and filtering according to a fusion target measurement result to obtain a matching target of the track; track life cycle management: newly building, confirming, keeping or deleting the track according to the matching result and the time step; according to the rule after track re-tracking, under the condition that the target is temporarily lost or shielded, a virtual track in the shielding period is constructed, state correction is conducted on the virtual track, the corrected state serves as the initial state of the current moment, and tracking continues. According to the invention, through fusion of complementary information of the 4D millimeter wave radar and the visual sensor, accurate sensing and tracking of three-dimensional multiple targets in a complex automatic driving environment are realized.

The invention discloses a personnel behavior analysis management system and method based on AI vision, relates to the technical field of vision analysis, and solves the problem of imperfect definition of abnormal behaviors by introducing a dynamic behavior feature verification module, calculating an image verification factor Bcz and evaluating and extracting potential abnormal behavior features. A multi-target online analysis module is adopted, a convolutional neural network and a space-time diagram convolutional network are fused, a behavior abnormal factor Ax and a target tracking factor Tg are extracted through an attention mechanism, and Kalman filtering and an SORT algorithm are combined, so that accurate tracking of multi-target behaviors is realized, and the problem of difficulty in multi-target tracking is solved; the problem that the platform data fusion degree is low is solved through the self-adaptive comprehensive evaluation module; the abnormal behavior recognition, target tracking and multi-source data fusion analysis capabilities are effectively improved, and accurate monitoring and intelligent intervention of abnormal behaviors in a complex scene are realized.

A three-dimensional target detection and tracking method fusing multi-sensor information belongs to the technical field of computer vision, and can make up for the defect of a single mode by collecting camera and radar image information and fusing camera images and laser radar point clouds. Multi-target tracking is carried out by using a multi-data coupling method of 2D + 3D data, a remote object is identified through a camera, when the object enters an induction range of the laser radar, the laser radar can carry out more accurate positioning, and position estimation of the target in a 2D image is corrected by using a more accurate 3D position provided by the laser radar. The method is used for solving the problems of inaccurate target identification, target shielding and the like under complex road conditions; a traditional single-mode three-dimensional target detection algorithm cannot effectively fuse the advantages of multi-sensor data and is difficult to accurately identify objects in a distance or under a complex road condition.

The invention discloses a multi-target tracking method based on track guidance, and the method comprises the following steps: carrying out the detection of a video frame through a target detector, obtaining the appearance information and position information of a target person, keeping a high-confidence detection frame and a low-confidence detection frame, and initializing a track; predicting the position of the target person in the next frame according to the track position information, wherein the corrected state vector and covariance matrix are used in the Kalman filter prediction stage; performing first matching on the high-confidence detection frame and all target candidate frames according to the motion similarity, the appearance similarity and the speed direction calculation cost matrix, and then performing second matching between the unmatched target candidate frames and the low-confidence detection frame by using the motion similarity and the speed direction calculation cost matrix; and performing unified management on matching results, and outputting target trajectory information. According to the invention, an end-to-end multi-target tracking algorithm can be realized, and the real-time rate and accuracy of tracking are improved.

The invention relates to the technical field of information, and discloses a parking lot panoramic safety monitoring method and system, and the method comprises the steps: obtaining a real-time video stream of a parking lot, and carrying out the denoising preprocessing of each frame of image, and obtaining a clear image; adjusting the brightness and the contrast ratio of the clear image by adopting a self-adaptive illumination compensation algorithm to obtain an illumination compensation image; according to the angle and distance change of the vehicle, correcting the vehicle image visual angle in the enhanced image by adopting a perspective transformation algorithm to obtain a perspective transformation image; identifying a vehicle position and a bounding box in the perspective transformation image by adopting a vehicle detection model based on deep learning, and determining a specific position of the vehicle in the image; and in the vehicle dense area, a multi-target tracking algorithm is combined with the moving tracks and the feature information of the vehicles, the shielding relation between the vehicles is judged, and the positions of the shielded vehicles are determined. The method can realize panoramic safety monitoring of the parking lot.

The invention discloses a home decoration construction site monitoring system and method based on artificial intelligence, and relates to the technical field of video processing. A video data set, an environment parameter set and an equipment state set in a historical construction process are collected in advance, an image semantic segmentation model is constructed based on the video data set, and construction scene division is performed; constructing a target detection model based on the video data set, generating a target spatial-temporal trajectory based on a detection result of the target detection model by using a multi-target tracking algorithm, and constructing a construction behavior recognition model based on the video data set and the target spatial-temporal trajectory; based on the recognition result of the construction behavior recognition model, the environmental parameter set and the equipment state set, constructing a risk early warning model; the safety condition of a construction site is evaluated in real time, and dynamic risk early warning is provided.

The invention provides an online multi-target multi-camera vehicle tracking method based on deep learning. The method mainly comprises the following steps: capturing vehicle video images from a plurality of cameras; identifying and positioning a vehicle instance by using a YOLOv11 algorithm; 2048-dimensional appearance features of the vehicle are extracted through a ResNet101IBN convolutional neural network; applying a single-camera multi-target tracking algorithm to generate a target track under each camera; and through a hierarchical clustering algorithm, in combination with feature cosine distance and Dunn index optimization, clustering is performed on a target trajectory, and cross-camera target matching is completed. The method iteratively executes the steps, and meets the multi-target vehicle tracking requirement in a continuous video stream. By integrating a lightweight vehicle detector, efficient appearance feature extraction, accurate single-camera tracking and an advanced cross-camera association strategy, the real-time performance, accuracy and robustness of vehicle tracking are remarkably enhanced, and the method is suitable for complex multi-camera monitoring environments such as urban monitoring, traffic management and automatic driving assistance systems.

The invention provides a multi-target tracking method based on EKF-ANA and multi-distance trajectory matching, relates to the technical field of multi-target tracking, and provides an adaptive extended Kalman filter algorithm for a trajectory prediction stage and a multi-distance trajectory matching method for a trajectory matching stage based on a two-stage multi-target tracking framework. And a multi-target tracking function under a complex background is realized. According to the method, a two-stage detection-based multi-target tracking mode is used, firstly, the position, the category and the confidence coefficient of a detected target are obtained through a detection algorithm, and then a track is predicted through a self-adaptive extended Kalman filtering algorithm. And finally, performing association operation between the target and the track by using a Hungary matching algorithm based on a multi-distance weighted data association cost matrix. According to the method, the accuracy of target tracking can be improved, the problem of partial ID switching caused by shielding is reduced, and certain real-time performance is met while a multi-target tracking task is completed.

The invention discloses a multi-target tracking method based on threshold separation and feature enhancement, and the method achieves the association between a target and a track in a current frame through a data association algorithm based on a target detection result through the motion features and appearance features of the target. The method comprises the following steps: dividing a high-score detection frame and a low-score detection frame according to a result of a detector and a detection confidence score threshold value; extracting motion features and appearance features of the high-score detection frame, and predicting positions of all tracks in the track set in the current frame; the first association occurs between a high-score detection frame and a track set, motion features and appearance features are fused and used, the second association occurs between a low-score detection frame and a track which is not matched, and only the motion features are used for calculating a cost matrix; trajectory management includes initializing new trajectories, deleting expired trajectories, and removing repeated trajectories. According to the method, a feature extraction mode and a feature fusion mode are improved, and the high-order accuracy of multi-target tracking is improved.

The invention relates to the technical field of self-service, and particularly discloses a non-inductive payment method and device, and the method comprises the steps: collecting goods taking pictures of a plurality of clients through a binocular camera and an infrared array which are disposed at the top and side surfaces of a cabinet body, extracting hand key points through posture estimation, and distinguishing the identity through multi-target tracking; a gravity sensor under a shelf board detects the weight change of each goods allocation, and combines with a visual track to construct the space-time correlation of goods movement, so as to realize the cross validation of goods taking behaviors; the convolutional neural network performs abnormal behavior detection on the multi-modal data and generates a risk score; creating a virtual shopping basket containing a space-time mark for each customer, and judging commodity attribution by adopting a decision tree algorithm; and executing distributed bill generation and multi-channel fee deduction according to the risk level, and completing real-time settlement in a multi-client concurrent scene. According to the method, the commodity affiliation accuracy and the payment security are remarkably improved, and efficient and stable unmanned retail operation can be realized.

The invention discloses a roadside fusion tracking method and system based on Leiye frame fusion, and relates to the technical field of intelligent traffic. The method comprises the steps that when a certain MEC server triggers a fusion instruction and a timer of the MEC server reaches trigger time, image sensing data and radar sensing data in current cache data are acquired; performing frame fusion on the image sensing data and the radar sensing data to generate a thunder fusion frame; and carrying out target continuous tracking based on the Reighting fusion frame and the historical fusion data to obtain a tracking result, updating the historical fusion data, calculating to obtain a complete credible interval and a basic credible interval of the MEC server, and outputting the Reighting fusion frame and the tracking result at the same time. According to the method, the perception results of radar vision are fused, then multi-target tracking is carried out based on the fusion result, the continuous tracking effect of the target is good, the perception precision is high, the universality is high, the detection effect is good under the conditions of a static state and a long distance, and the problems of ghosting, splitting, position jumping and the like are effectively solved.

The invention relates to the field of tunnel fire prediction, in particular to a new energy vehicle fire early warning method and system in a tunnel. The method comprises the following steps: identifying a new energy vehicle entering a tunnel by using a YOLOv8 model, and obtaining a visible light image, a thermal imaging image and millimeter wave radar point cloud data of the new energy vehicle; a multi-target tracking algorithm is introduced, target tracking is conducted on the new energy vehicles, the battery temperature of each new energy vehicle is calculated in combination with tunnel environment data, space-time synchronous calibration is conducted on collected data, and then the motion trails of the new energy vehicles are calculated; time synchronization calibration is conducted on the movement track and the battery temperature, and the battery temperature gradient of the new energy vehicle is calculated according to the result obtained after time synchronization calibration; and predicting and alarming the tunnel fire according to the battery temperature gradient. The problem that in the prior art, fire early warning for the new energy vehicle in the tunnel is lacked is solved, and tunnel operation safety is improved.

The invention provides a mobile robot figure tracking comprehensive framework based on multi-target tracking. Firstly, the framework combines an improved OC-SORT tracking algorithm with a gesture detection technology, and realizes dynamic selection, accurate recognition and position estimation of a plurality of objects. Secondly, accurate path tracking of the omni-directional wheeled mobile robot can be achieved by planning a Dubins path and providing transverse and longitudinal control models based on the omni-directional wheeled mobile robot. According to the improved OC-SORT tracking algorithm, evaluation is carried out on public data sets such as DanceTrack and Person Tracking, the tracking precision and the positioning precision reach 60.5 HOTA and 98.1% respectively, and the tracking precision and the positioning precision exceed existing multi-target tracking OC-SORT and single-target tracking ODFS. The robot personnel tracking framework provided by the invention can be operated on an embedded system equipped with NVIDIA TM # imgabs0 # Jetson Xavier NX. The performance of the system on customized personnel tracking data sets and the successful application of the system in real-world scenes prove the effectiveness of the system in real-time tracking and following of individuals.

The invention discloses a binocular vision-millimeter wave radar multi-element joint detection method, which comprises the following steps: constructing a vision monitoring component, carrying out target detection in a region by utilizing a learning network, and carrying out vision target tracking based on a multi-target tracking algorithm to obtain a vision target trajectory; the target echo signal is processed by a radar signal processing algorithm to generate a target plot, and based on a joint probability data association algorithm and a Kalman filtering algorithm, tracking and prediction of a radar target are realized to obtain a radar target trajectory; and according to the visual target trajectory and the radar target trajectory, establishing a multi-level constraint association algorithm, and constructing a feature-level deep fusion framework. According to the method, deep fusion of radar and binocular vision target data is realized, and the sensing coverage range and the information stability of a radar-binocular vision fusion system are enhanced.

The invention discloses a multi-target tracking method combining camera motion compensation and pseudo depth estimation, belongs to the field of computer vision, and is suitable for a complex automatic driving road environment. The method comprises the following steps: constructing a training set and a test set; detecting the image by using a deep learning detector and extracting features; a Kalman filter is adopted to correct a motion modeling state vector, and the target position and size prediction precision is improved; solving a homography matrix through feature point matching, performing global camera motion compensation, and reducing camera jitter and displacement interference; target pseudo depth information is calculated, hierarchical cascade matching is carried out, and association performance in dense and shielding scenes is optimized; a three-level cascade strategy is adopted to complete high confidence degree, low confidence degree and residual target matching in sequence; and finally, outputting a tracking result with a detection frame and identity information to obtain a trained model. According to the invention, accurate detection and stable tracking of multi-category targets can be realized in a complex environment, and identity switching is effectively reduced.