Deep learning-based cross-camera pedestrian multi-target tracking method and device

Abstract

The invention relates to a deep learning-based cross-camera pedestrian multi-target tracking method and device. The method comprises the steps of obtaining video data of multiple data sources; inputting the pedestrian detection frame into a trained encoder decoder deep network, simultaneously carrying out pedestrian detection and pedestrian apparent characteristic extraction, obtaining a pedestrian detection frame taking the pedestrian as a central point for detection, and outputting pedestrian apparent characteristics according to the central point; performing association matching on all pedestrian detection boxes based on all pedestrian apparent characteristics to obtain tracking trajectories of all pedestrians; for the tracking trajectory of each pedestrian in the plurality of videos, selecting a high-quality frame; respectively extracting pedestrian face features and gait features by using the trained deep neural network; carrying out feature fusion or matching on various features;and finally, performing cross-camera trajectory association clustering. The method and the device can be directly applied to an old monitoring system, have good robustness, and can realize accurate and efficient cross-camera pedestrian tracking.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method and device for cross-camera pedestrian multi-target tracking based on deep learning. Background technique [0002] With the advancement of computer performance and the development of deep learning, many computer vision technologies are widely used in people's real life. For the traditional video surveillance system, it needs to rely on a lot of human resources to know whether the specified target appears, where and when it appears, and it mainly plays the role of evidence collection after the event. Research developers in related fields integrate computer vision, image processing, pattern recognition, artificial intelligence and other technologies to develop an intelligent video surveillance system, which can automatically identify and track pedestrians in video sequences with the help of powerful computer data processing capabilities . [0003] In existing ped...

AI Technical Summary

Problems solved by technology

[0006] In practical applications, enterprises or units that need intelligent monitoring systems directly replace the old monitoring systems with new intelligent monitoring equipment, but the replacement of new intelligent monitoring equipment requires a lot of additional funds and causes great inconvenience to the elimination of old monitoring equipment. Necessary waste of resources

However, the video frames in the old surveillance video have problems such as low resolution, obvious lighting changes, and insufficient occlusion. How to alleviate and reduce the impact of these deficiencies and factors is the key to solving the problem of pedestrian tracking.

Images