Multi-Bernoulli video multi-target detection and tracking method based on yolov3

Abstract

The invention discloses a multi-target detection and tracking method based on YOLOv3 multi-Bernoulli video, which belongs to the fields of machine vision and intelligent information processing. The present invention introduces YOLOv3 detection technology under the multi-Bernoulli filter framework, uses anti-interference convolution features to describe targets, and interactively fuses detection results and tracking results to realize accurate estimation of unknown and time-varying video multi-target states; During the tracking process, the matching detection frame is combined with the target trajectory and target template to perform real-time target new judgment and occluded target re-identification, while considering the detected target and estimated target identification information to realize target identification and track tracking , which can effectively improve the tracking accuracy of occluded targets and reduce trajectory fragments. Experiments show that the invention has good tracking effect and robustness, and can widely meet the actual design requirements of systems such as intelligent video surveillance, human-computer interaction, and intelligent traffic control.

Description

technical field [0001] The invention relates to a multi-target detection and tracking method based on YOLOv3 multi-Bernoulli video, belonging to the fields of machine vision and intelligent information processing. Background technique [0002] In the field of video multi-target tracking applications in complex environments, in addition to problems such as illumination changes, target deformation, and target occlusion, there are also complex problems such as unknown number of targets, uncertain new targets, crossing or close movement of targets, disappearance of targets, and clutter interference. Situation has always been a difficult and challenging problem in the field of multi-target tracking. [0003] Aiming at the problem of video multi-target tracking, in the early days, the target detection and tracking method based on data association was mainly used. First, the target detector was used to detect multiple targets in the video sequence, and then the video multi-target t...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem that existing existing target tracking methods cannot detect new targets and when multiple targets occlude and interfere with each other, the tracking accuracy decreases, and even the target is missed. Nourishing video multi-target detection and tracking method, in the detection and tracking process of the method, the k and k+1 frame video sequences of video are detected by YOLOv3 technology; the number of detection frames at k moment is n , the detection frame state set is The number of detection frames at time k+1 is m , and its detection frame state set is in, Represents the i-th detection frame state vector, parameter respectively represent the abscissa and ordinate of the upper left corner of the i-th detection frame at time k, and the width, height and label of the detection frame;

Images