Auto-relation model based detecting algorithm for industrial images

It makes sense that efficiently orients some objects by analyzing real-time industrial images about the case of mixed substance in industrial material. In this paper, we present a new algorithm based on auto-relation model to deal with the need for recognizing the materials mixed in salt carried on the conveyer belt. Generally, those mixed substance will be with their different textures and gray values, and bring about gradients changes. These changes have shown us clues to the problem in industrial inspection. In the approach, a rank-value median filter is implemented in pre-process in order to decrease image noise at the first, then we build up a transformation model based on the auto-relation of image function, and give the cost function concerning the determinant and trace of this matrix to analyze the distributions on image energy changes. Finally, a piloting set containing the points with high image energy changes will be constructed, and those possible positions of mixed materials are decided.