Research on Infrared Image Segmentation of Power Equipment Using Niblack Optimized by Bat Algorithm

Abstract

Infrared image segmentation of power equipment is the basis for intelligent diagnosis of power equipment faults. In order to reduce the influence of non-uniform background on the infrared image segmentation of power equipment and improve the accuracy and efficiency of image segmentation, an improved Niblack image segmentation method based on bat algorithm is proposed. This method uses the variance between classes as the fitness function of the bat algorithm to automatically search for the optimal segmentation threshold of the non-overlapping rectangular neighborhood in the Niblack method, and uses it for binarization of the current neighborhood to extract target area from the infrared image. Experimental results show that compared with the traditional Otsu method, Niblack method and other algorithms, the segmentation algorithm reduces the ME by at least 34% to 84%. Compared with the BA+Otsu method, the average time consumption is reduced by 70%, effectively improving the accuracy and efficiency of the infrared image segmentation detection of the device.