A new vision monitoring method for resistance spot welding with high speed industrial camera is presented, and by this method the dynamic procedure of spot welding electrodes touching can be inspected. The pressure and current influence for resistance spot welding spatter can be distinguished by a serial of continuous images with millisecond time interval. The other spatter phenomena caused by welding pole wearing or fixing obliquely can also be distinguished by vision images. By a series of experiments, this machine vision system is be built by C++ language, which can detect the resistance spot welding spatter occurring and distinguish the spattering possible reasons. In conclusion, the machine vision monitoring system built in this paper can capture thousands of frames at one welding duty cycle, and then find the key frame with significantly welding spatter features, subsequently calculate the amount of spatter and distinguish the classification. The running testing of this system shows that this machine vision system run stably and can timely find the spot welding spatter problems in time.