With the development of human activities in the oceans, underwater autonomous robots have become important equipments for marine surveying and scientific research. In this paper, an avoidance system of autonomous underwater vehicle (AUV) is designed based on the image recognition technology, implementing underwater autonomous cruise, avoiding obstacles, floating and diving, real-time data transmission. The system uses a CMOS analog camera and LM1881 video separation chip to extract the analog video signal. The analog video signal is converted to digital signal by the MC9S12XS128 single chip external interruption function and internal AD transducer. Then the edge detection algorithm is used to recognize obstacles through the digital video signal. The digital signal and video streaming will be transmitted to the upper computer by nRF24L01 wireless data module. After the data is received, the program running on the host computer will display the data and picture to the screen, in order to monitor the underwater situation in real-time. At the same time, the video module of wireless video collection card uses OpenCV base to secondary development and realize more complex algorithm of upper PC. The results show that system realizes functions of obstacle avoidance and underwater surveying, thus indicating higher cost performance and practical significance.