Pedestrian multi-target tracking method based on correlation clustering and space-time constraint

Abstract

The invention discloses a pedestrian multi-target tracking method based on correlation clustering and space-time constraint. The method comprises the steps: pedestrian visual feature extraction, correlation clustering based on visual features, pedestrian trajectory association under a single camera, and pedestrian trajectory matching under a cross-camera condition by using a space-time constraintmethod. Aiming at the problem that pedestrian tracking under a single camera is easy to interrupt, a space-time sliding window is introduced to solve the problem; meanwhile, a space-time constraint method is introduced to associate the same pedestrian in a cross-camera scene, so pedestrian multi-target tracking in a cross-camera scene is realized; by utilizing the method provided by the invention,tracking indexes such as MOTA, MOTP and recall rate in pedestrian tracking can be consistently improved.

Description

technical field [0001] The invention relates to the fields of computer vision and pattern recognition, in particular to a pedestrian multi-target tracking method based on correlation clustering and time-space constraints. Background technique [0002] Computer Vision (Computer Vision) is a science that studies how to make machines "see". To put it further, it refers to a computer technology that uses cameras and computers instead of human eyes to identify, track and measure targets. Computer vision aims to create an artificial intelligence system that can obtain "information" from images or multidimensional data, so it has been a hot and difficult research topic in the field of computer science in recent years. [0003] As the basis of high-level vision tasks, the multi-object tracking problem has become a crucial research problem in the field of computer vision. Multiple object tracking, that is, Multiple Object Tracking (MOT). Its main task is to find an image sequence, ...

AI Technical Summary

Problems solved by technology

However, cross-camera pedestrian tracking has some difficult problems in practical applications, which are mainly reflected in two aspects: on the one hand, many current tracking algorithms judge the overlap area between adjacent pedestrian detection frames by calculating Whether it belongs to the same pedestrian target, but due to the large number of objects under the surveillance video, the calculation process of pedestrian tracking will be interrupted due to the occlusion of the object

On the other hand, for pedestrian tracking under cross-camera, most of the methods currently used are similar to the pedestrian retrieval method in pedestrian re-identification. This method is mostly implemented based on the characteristics of pedestrian targets, and the spatiotemporal information in the data set is not used.

Images