Method for tracking multiple vehicles and multiple pedestrian targets in video

Abstract

The invention discloses a method and device for tracking multiple vehicles and multiple pedestrian targets in a video, and the method comprises the steps of building different feature extraction operators through employing a multi-feature appearance modeling method based on a target category, completing the accurate appearance description of the target, and achieving the accurate description of the target; using a hierarchical progressive feature extraction algorithm, and using a high-threshold matching algorithm to complete rapid dimensionality reduction of an incidence matrix, so that the number of times of depth feature extraction is reduced, and the timeliness of calculation is ensured; according to the invention, the relative orientation constraint relation between the targets is usedfor assisting the apparent characteristics to complete target matching, effective distinguishing of part of targets with similar appearances is achieved, and the mismatching rate is reduced; the recovery of the missing detection target is completed by using the target apparent characteristics and the motion prediction, and the recovery rate of the missing detection target is greatly improved. Themethod is high in accuracy, strong in adaptability and high in detection efficiency.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method and device for tracking multiple vehicles and multiple pedestrians in a video. Background technique [0002] Video multi-target tracking is a research hotspot in the field of computer vision, and it is an important part of many intelligent vision systems. The main tasks of video multi-object tracking include locating object locations and maintaining object IDs. The video sequence used for target tracking is obtained by projecting the 3D real world onto the 2D image plane by the video acquisition device, which will inevitably introduce information loss, and the illumination changes, scene changes, and noise during the imaging process will also affect the quality of the video sequence . In addition to facing the problem of blurred target appearance caused by video degradation, video multi-target tracking also faces challenges such as complex and changeable backgr...

AI Technical Summary

Problems solved by technology

Aiming at the low timeliness of the tracking algorithm caused by time-consuming depth feature extraction, the present invention proposes a layered and progressive feature extraction algorithm to reduce the number of depth feature extractions as much as possible on the premise of ensuring tracking accuracy

Images