Cross-view gait recognition method, system and equipment based on multi-feature aggregation

• Main Authors: WANG ZHONGYUAN, CUI LAIZHONG, ZOU QIN, LIANG CHAO, ZOU HUA, CHEN JIANYU
• Format: Patent
• Language: Chinese; English
• Published: 18.10.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention discloses a cross-view gait recognition method, system and equipment based on multi-feature aggregation. The method comprises the following steps: firstly, preprocessing an input video to obtain a gait sequence; and inputting the processed gait sequence into a cross-view gait recognition network (GaitAMR) based on multi-feature aggregation to obtain a recognition result. Specifically, the invention provides a spatial feature recombination evaluation strategy (SFRA), and occlusion existing in a spatial domain is eliminated by reconstructing high-quality appearance features of a subject. Secondly, a global and partial time aggregation strategy (HPTA) is proposed, which extracts body motion descriptors from both global and local aspects to capture unique motion patterns of the human body. And finally, providing a view projection prediction learning strategy (VPPL), and embedding the features under the optimal view as supplementary information into a feature recognition part so as to enhance the vie