The work aims to discuss the visual image features of assistive robot for AD from the perspective of prototype theory and carry out design practice, in order to meet the needs of cognitive stimulation training and daily care of elderly patients with Alzheimer disease at home. According to the selection principle of prototypes, the prototype samples were determined and the typical task scenarios were constructed. Image vocabularies were collected and key frames were added to intercept action parameters to obtain static features and dynamic features of the prototype respectively. The level of visual image feature variables of assistive robot for AD was constructed, and the prototype features were mapped to robot. The mapping results were used as a basis for visual image design of assistive robot for AD. Based on prototype theory and visual image features, this research constructs a design method of visual image features for assistive robot for AD. The mapping relationship between perceptual images and feature variables is adopted to carry out design practice and verification, thus providing an effective reference and a new direction for the design of household assistive robot for elderly AD patients.