Abstract: Periodic magnetic particle testing of the welds of pressure equipment such as spherical tanks is an important measure to ensure its safety. The development of magnetic particle testing system based on wall climbing robot is an important step towards the automatic inspection of spherical tanks. The wall climbing inspection robot carries a visual sensor to collect magnetic particle testing images in real time. Therefore, the rapid automatic analysis of magnetic particle testing images is an important research content of the wall climbing inspection robot. In this study, the image pixel analysis method was used to automatically identify the real-time images collected by the visual sensor. The images were quickly and automatically picked into three identical folders, named defect-free folder, defective folder and risk categories folder. The system is to assist the magnetic particle tester to identify weld defects quickly.