Multi-target vehicle tracking device based on detector and tracker data association

Abstract

The invention discloses a multi-target vehicle tracking device based on detector and tracker data association. The multi-target vehicle tracking device comprises a multi-stage target detector, a correlation filtering tracker, a data association module and a human-computer interaction interface. The multi-stage target detector is used for periodically detecting small vehicle targets in an input video frame in a multi-scale and actual aerial photography scene and outputting all small vehicle target frames of the frame; the correlation filtering tracker uses the positions and sizes of targets output by the multi-stage target detector to initialize the targets in parallel, and then tracks the targets frame by frame. The data association module periodically performs data association and matching on a detection result and a tracker prediction result, and is used for maintaining the id of each target and correcting the target box. The man-machine interaction interface is used for a user to select videos and algorithm models, check evaluation indexes and interact with a tracked target in the operation process of the device.

Description

technical field [0001] The invention belongs to the technical field of video target detection and tracking, and in particular relates to a multi-target vehicle tracking device based on data association between a detector and a tracker. Background technique [0002] The video target detection task identifies the target type and determines the target position by analyzing and processing the video frame by frame; using the detection result, initializes the target position in the first frame of the video, and then uses the tracking algorithm to predict the position of the target frame by frame; in the tracking process, Use the data association algorithm to associate and analyze the detection results and tracking results, and maintain the id information of the target in consecutive frames. [0003] Vehicles are one of the key objects in video analysis scenarios. Vehicle multi-target detection and tracking in video sequences has always been an important research direction in this ...

AI Technical Summary

Problems solved by technology

SORT Based on IoU Intersection-Union Ratio and Allocation Strategy [2] The algorithm provides an idea for the data association method, which uses the detection results to correct the tracker prediction results, but when this method tracks multiple targets, it is prone to target ID exchange, and when the target reappears after a short tracking drift, The original ID cannot be maintained, and with the appearance of new targets and the disappearance of old targets, the existing method has the phenomenon that the tracker cannot be created and deleted in time

Images