Reliability-based multi-layer depth feature target tracking method

• Main Authors: ZHOU WENJUAN, BEI SHAOYI, ZHOU LINWEI, YIN MINGFENG, HUA XU, CHEN CHANGKAI, YOU LIPING, JIN SHENGXIN
• Format: Patent
• Language: Chinese; English
• Published: 06.07.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses a reliability-based multi-layer depth feature target tracking method. The characterization capability of different-layer features is measured by calculating the reliability of a channel by utilizing different discrimination capabilities of different-layer features in a tracking scene, so that the positioning information of the different-layer features is fused, and more accurate target position information is obtained. In the tracking process, a scale pool technology is adopted to update the size of a target in real time, and features extracted from a previous frame are fused with original template features to obtain a tracking model with higher robustness. According to the method, the representation capability of the model is improved, and more accurate target positioning is realized; the generalization ability of the model is improved; the ability of the tracking model to cope with the target scale change in a complex scene is improved; the robustness of the tracking model for dealin