The work aims to develop a dynamic interactive assembly method for complex high-end equipment based on digital twin technology and an improved dynamic target detection algorithm to address the issues of low efficiency and poor precision in the assembly of complex equipment parts. A human-machine interaction framework towards digital twins was constructed and the target detection algorithm was optimized to achieve real-time monitoring of the complex equipment assembly process. Operators could interact dynamically with the system through a visual interface, thereby achieving real-time optimization of the assembly process. Initially, real-time data of the equipment was collected through multi-source sensors. Subsequently, a virtual model of the equipment was constructed based on real-time data, generating videos highly consistent with the actual assembly scene, and precise detection and classification of industrial parts were achieved through the optimized YOLOv8n algorithm. Then, the assembly process was dynamically optimized through the human-machine interaction process, and the detection results were displayed in real-time through the designed interactive interface, providing intelligent support for equipment assembly. The method proposed in this study could achieve precise detection and real-time dynamic response during the assembly process of complex high-end equipment. The improved dynamic target detection results not only maintained high precision but also increased the detection speed by 24.9% compared with the unimproved version. This study not only provides an innovative solution for human-machine interactive assembly of complex high-end equipment but also offers strong support for the further application and development of digital twin technology in the industrial field.