Development of system for crossarm reuse judgment on the basis of classification of rust images using support vector machine

We attempt to develop a crossarm reuse judgment system based on rust images that uses machine learning techniques. The system consists of a digital camera and a standard note book personal computer (PC). We estimate the degree of accuracy of the judgment of various pattern classification methods without special image processing techniques such as the extraction of features. The results show that a support vector machine is the most suitable instrument for this judgment system. We obtain the high degree of accuracy by compressing the image data in order to decrease the number of features.