The article develops a vision based auto-recharging system for mobile robots, and programs a new docking processing to enhance successful rate. The system contains a docking station and a mobile robot. The docking station contains a docking structure, a control device, a charger and a detection device and a wireless RF interface. The mobile robot contains a power detection module (voltage and current), an auto-switch, a wireless RF interface, a control system and a camera. The docking structure is designed with one active degree of freedom and two passive degrees of freedom. The active degree of freedom can move forward to contact the recharging connect points that are arranged in the mobile robot. The two passive degrees of freedom can rotation in the Z-axis and use compression spring moving on various docking condition. In image processing, the mobile robot uses a webcam to capture the real-time image; and transmits the image signal to the computer via USB interface, and uses Otsu algorithm to recognize the position of the docking station. In the experiment results, the system had been successfully guided the mobile robot moving to the docking station using the proposed method.