With the gradual increase in the proportion of the global elderly population, the number of bathroom accidents also increases yearly. Due to the humid environment in bathrooms, the floors are mostly made of tiles, which are relatively prone to slips and falls. However, the existing pose recognition methods mostly use color images, and color images have concerns about personal privacy and are susceptible to lighting changes and complex spatial interference. Therefore, this paper proposes an abnormal behavior recognition system based on thermal image AI edge technology. Use a dangerous posture when using the toilet to avoid accidents in the bathroom, such as using the toilet being discovered too late, increasing the risk.