Visual signal processing is a crucial aspect of artificial intelligence. Although deep neural network provides very high accuracy to process visual information, the massive model causes extremely high computing power consumption and is difficult to implement in mobile devices. This project aims to solve this crucial problem by integrating biological principle and neuromorphic architecture with deep neural network to develop a neuromorphic intelligent visual system-on-chip for mobile devices. The project includes four aspects: 1) low-power intelligent visual image sensing chip, 2) neuromorphic obstacle avoidance algorithm for analog sensing of insect vision and spatial recognition, 3) realization of a low-latency and low-power deep learning chip based on neuromorphic intelligence, 4)an analog-digital mixed neuromorphic intelligent chip architecture. By integrating these four subprojects, we aim to design, develop, and integrate a low-voltage, low-power, real-time mobile neuromorphic intelligent visual system that can perform tasks of recognition and track prediction.