With the rapid development of UAV technology, the use of unmanned aerial vehicles instead of artificial power line inspection line is the industry development trend. Although UAV power patrols are more efficient and safer than manual inspection, UAV inspection continues to face problems such as aviation restrictions, airborne weight, flight life, obstacle avoidance technology and other factors. Unmanned aerial vehicle obstacle avoidance technology needs to be improved, in the inspection process, no chance will appear with the surrounding obstacles collision situation, and especially the small wires on the UAV caused a huge threat. These lines are often not obvious targets, radar, ultrasound and other technologies to get less echo, resulting in obstacle avoidance difficulties. The UAV obstacle avoidance technology based on binocular vision system has been widely studied and paid attention to in the field of unmanned aerial vehicles. This paper designs a realtime UAV power line detection system based on binocular vision system. In this paper, a matching cost measurement method based on TCensus (morphological Tophat transform and MiniCensus) is proposed to enhance the weak target in the original image, and the support region based on the cross structure is used to improve the matching accuracy. The experimental results show that the binocular vision system can effectively detect the distance between the UAV and the power line, and the detection error is 5%. The stereo matching algorithm proposed in this paper can obtain a better depth map, and the TCensus algorithm can effectively preserve the original image in the weak target line information.