The application of augmented reality technology is gradually becoming popular, but there are still issues with insufficient interactivity and presentation effects in the integration of interior design. To address these issues, this article proposes a method that combines augmented reality technology with computer algorithms to optimize the interior design process. Firstly, the indoor environment is reconstructed in 3D using the ORB (Oriented FAST and Rotating BRIEF) feature extraction algorithm to obtain an accurate model of the real scene; next, using Visual Inertial Odometry (VIO) and Simultaneous Localization and Mapping (SLAM) technology, virtual design elements are seamlessly integrated with the actual scene to achieve real-time tracking and localization of augmented reality scenes; then, a Convolutional Neural Network (CNN) is used to learn and analyze user preferences, in order to optimize the recommendation of design solutions and ensure personalized and intelligent design. The experimental results show that the average error of the VIO+SLAM joint positioning method is only 1.6 cm, demonstrating the best fusion accuracy and viewpoint stability. In addition, the user acceptance rate of the personalized recommendation system reaches 82.5%. In the above data conclusions, the method proposed in this article has significant advantages in improving the positioning accuracy and personalized recommendation effect of virtual design elements, and can effectively solve the shortcomings of existing augmented reality systems in interior design.