In order to meet the functional requirements for autofocusing in different scenarios, the advantages and disadvantages of various existing autofocus implementation schemes are evaluated and analyzed. A focusing scheme based on corner detection algorithm is proposed. The sharpness of the image can be judged by detecting the number of corners in the image. According to the performance of the FPGA image processing system of the thermal imager, the algorithm is improved and optimized. A real-time focusing method for multi-threshold corner detection is proposed. According to the focusing state of the current scene, the thermal imager can independently determine whether to perform the focusing operation to meet the needs of different detection scenarios, and finally realize the real-time automatic focusing function of the thermal imager in different scenarios.