In order to solve the problem of low recognition accuracy caused by fuzzy recognition of similar areas in indoor image recognition, a deep learning algorithm is introduced to achieve high-precision recognition, and a new design is made for the pixel relationship of the image. The specific calculation process includes three parts: concave-convex relationship analysis of indoor images based on deep learning, image fusion recognition based on deep learning, and indoor image classification and coding calculation based on deep learning features. Through the above process, the high-precision recognition and classification of images are realized. Experimental data comparison results show that the algorithm has high effectiveness, meets the requirements of application, and meets the standards of algorithm promotion.