Heuristic three-dimensional human body grid reconstruction method based on attitude prior constraint

• Main Authors: YANG JIANTAO, LIU SHENG, PAN SONGQI, FENG YUAN, TIAN XIAOPENG, ZHANG YINENG
• Format: Patent
• Language: Chinese; English
• Published: 09.01.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention belongs to the technical field of monocular three-dimensional human body reconstruction, and discloses a heuristic three-dimensional human body grid reconstruction method based on attitude priori constraint, which extracts a comprehensive priori constraint vector from a feature map, focuses self attention on the body part of a human body in an image, and avoids feature redundancy. In addition, skeleton information is merged into a reciprocal mapping layer, so that the human body grid posture correction capability of the model is enhanced, and meanwhile, the semantic consistency of input and output is ensured. In addition, according to the extracted region prior constraint, the local vertex is finely adjusted by pointing cross attention to key points around the vertex. And then, in the fine tuning process, determining the optimal offset of the current query vertex by utilizing a heuristic fine tuning strategy. Compared with a traditional method based on Transform, the heuristic Transform provided