Application of adaptive convolution masking to the automation of visual inspection

An approach is presented for the automation of important aspects of human visual inspection in quality control. Pattern recognition and digital image processing are used to detect and classify defects in full gray-scale images of complex mechanical assemblies. The method simulates the processes of adaptation, fixation, and feature extraction in the human visual system. It applies an algorithm for optimizing convolution masks to distinguish between acceptable and unacceptable images. As a numerical example, the technique is used to detect a number of defects in X-ray images of complex mechanical assemblies.<
>