To explore the design of the virtual reality (VR) augmented reality (AR) mobile platform and game decision model based on deep learning (DL), the gesture-based interaction of VR games based on Leap Motion is researched. Based on the interactive features of gestures, a set of general gesture interaction rules in VR games is established. In the meantime, according to the theoretical basis and the characteristics of VR, a set of general models of VR gesture interaction is designed, the factors affecting the efficiency of VR gesture interaction are studied, and reasonable interaction feedback is designed. By using the computer vision and image processing technology, gesture-based interaction can collect natural gestures, extract gesture features, recognize gesture indications, and respond to the user demands. Also, it can extract the basic gestures from gesture-based interaction in VR, analyze the basic features of gestures and gesture-based interaction in VR games, and describes the gesture features by mathematical vectors and sets. The research results show that the application of gesture feature design method in the game can analyze the factors affecting the interaction efficiency. Also, the usability of the gesture-based interaction designed by the gesture design method is verified by tests. Therefore, the AR&DL platform of “AR+DL” establishes a learning platform supported by DL and AR technology. The game decision model is used to describe the process of gesture-based interaction in the game, and the factors affecting the interaction efficiency are reduced, which has certain reference and guidance for VR applications using gesture-based interaction.

The use of computer technology for three-dimensional (3 D) reconstruction is one of the important development directions of social production. The purpose is to find a new method that can be used in traditional handicraft design, and to explore the application of 3 D reconstruction technology in it. Based on the description and analysis of 3 D reconstruction technology, the 3 D reconstruction algorithm based on Poisson equation is analyzed, and the key steps and problems of the method are clarified. Then, by introducing the shielding design constraint, a 3 D reconstruction algorithm based on shielded Poisson equation is proposed. Finally, the performance of two algorithms is compared by reconstructing the 3 D image of rabbit. The results show that: when the depth value of the algorithm is 11, the surface of the rabbit image obtained by the proposed optimization algorithm is smoother, and the details are more delicate and fluent; under different depth values, with the increase of the depth value, the number of vertices and faces of the two algorithms increase, and the optimal depth values of 3 D reconstruction are more than 8. However, the proposed optimization algorithm has more vertices, and performs better in the reconstruction process; the larger the depth value is, the more time and memory are consumed in 3 D reconstruction, so it is necessary to select the appropriate depth value; the shielding parameters of the algorithm have a great impact on the fineness of the reconstruction model. The larger the parameter is, the higher the fineness is. In a word, the proposed 3 D reconstruction algorithm based on shielded Poisson equation has better practicability and superiority.

No abstract