Human-oriented image recognition for industrial inspection system

Abstract

An algorithm was proposed intending for external view image processing of FRP panel in visual inspection process. In the proposed algorithm, three kinds of inspection information such as location, size and level of defects are treated for clustering image data. The proposed algorithm enables one to process the dynamic sequential images with small facility and to obtain a reasonable efficiency of visual inspection. When the algorithm proposed by the authors was included, instead of a visual inspection process by a human being, it is expected that the productivity will be further improved and the human being, machine and computer will be cooperated respectively. Some numerical studies for visual inspection of FRP panel shows that the performance of the proposed algorithm is similar to that of expert inspector.