Abstract:As a traditional bulk aquatic product in China， the Chinese mitten crab （Eriocheir sinensis） possesses a complex physiological structure， resulting in different shell-meat separation methods for various body parts. Prior to mechanical shell-meat separation， intelligent recognition and targeted preprocessing of its characteristic parts are required. This study investigated the distribution of shell and meat in the crab， established an effective identification model for its characteristic parts based on YOLO v11， and examined the impact of cooking degree on shell-meat separation. The results showed that crab meat was relatively dispersed within the body， and the meat content in each part was significantly lower than that of the shell. The YOLO v11-based intelligent recognition model performed well， with F1 scores exceeding 0.9 under most confidence thresholds，and can accurately identify the crab′s characteristic parts. The optimal steaming durations for the legs， claws，and body were 20 s， 20 s， and 60 s， respectively. Slight cooking can greatly reduce processing time and energy consumption while maintaining the original physicochemical properties of the crab meat.