Recognition and drop-off detection of insulator based on aerial image

In order to improve the accuracy of recognition insulators and effectively reduce the influence caused by the texture and illumination of the background, a novel insulator recognition method merged the shape, color and texture of insulator is presented. Firstly, the parallel line features were perceived from different directions in the inspection images as candidates of insulator regions, and then the insulator candidates region are extended to neighboring regions for local neighborhood significance analysis based on local binary pattern to identify and main color components analysis based method to compensate the insulator region. The identified insulator region is adaptively divided into blocks according to the average distance between the insulator sheets, and drop-off defect of insulator is detected based on analysis of texture features changing in those blocks. Experimental results show that the method can effectively identify the glass insulator, synthetic insulator and diagnosis the glass insulator's off-chip defect through recognition of the UAV.