An automatic parts detection system based on computer vision

This paper proposes a new approach for parts detecting based on the Gaussian mixture model. In this work, firstly, an industrial camera is mounted over the moving conveyor belt to acquire the images of the parts which will be sent to the industrial personal computer (IPC). Secondly, we extract the features (such as area, perimeter and so on) of the parts from the acquired images by using a series of algorithms. Then, through comparing the area of the extracted features with the standard value, the unqualified parts will be confirmed. At last, we start the robot arm to push out the unqualified parts in real-time. To verify the effect of this proposed approach, an experiment has been done to detect and push out the unqualified parts. The experimental results show that the unqualified parts can be detected and pushed out effectively, and the success rate is more than 90%.