With the improvement of economic level and medical technology, surgical treatment has become more and more popular. On the other hand, the current surgical imaging system has some problems to be solved urgently. For example, the camera at the top is often hidden by the doctor's head. So we propose to deploy the camera system between doctors and patients. As the operation progresses, the doctor's hands and various surgical instruments will frequently change positions, which leads to new occlusion effects. The traditional photography system using a single camera always has a fixed angle of view, so this problem will inevitably occur. To against it, the system we proposed uses multiple cameras to record surgical videos from various angles. In this way, if an individual camera has a poor field of view, the system will automatically switch another camera that captures the target region well. Through the introduction of a deep learning algorithm (Yolov4), the surgical instrument and the surgical area are recognized, and then the necessary parameters (the recognition rate of the surgical area or scalpel) are extracted as the basis for switching viewpoints. This system captures high-quality surgical videos using multi-view cameras, and then imports them into a high-performance computer. The system splices the video sequences, image processing, camera calibration and other operations, and finally feeds the processed video to the monitor or panel for surgeons.