人本智造：人体行为识别关键技术分析与展望

With the continuous deep integration of new generation information technology and manufacturing technology, the human-centric smart manufacturing paradigm is reshaping traditional industrial production models. Human activity recognition technology, as a key enabling technology for implementing human-oriented smart manufacturing, primarily focuses on intelligent recognition and understanding of human activity semantics, which shows broad application prospects. A systematic exploration of the current development status, key challenges, and application prospects of human activity recognition technology in industrial scenarios helps promote theoretical development and innovative practices of human-oriented smart manufacturing. First, based on the developmental trajectory of human activity recognition technology, this study deeply analyzes the evolution process of core technologies such as human perception, activity modeling, and activity recognition, laying the technical foundation for industrial applications of human activity recognition technology; second, focusing on the special requirements of industrial scenarios, it emphasizes research on key technologies including robust multi-modal perception systems, multi-scale activity understanding frameworks, human-machine collaboration with integrated intention understanding, and optimized deployment in industrial scenarios; on this basis, it systematically analyzes and evaluates the quality of human activity datasets in industrial scenarios, and highlights the practical progress of human activity recognition technology in typical application scenarios such as production safety control, production scheduling optimization, process improvement, and activity enhancement; finally, combined with emerging technologies such as spatial intelligence, physiological-cognitive integration, and multi-modal large language models, it envisions future development directions for human activity recognition technology in industrial settings.