Abstract: Infrared image segmentation of electrical equipment is a main method of electric power fault diagnosis. Aimed at the shortages in traditional electrical equipment image segmentation methods, such as the high computation cost, the poor segmentation accuracy and the low segmentation speed, an improved butterfly optimizing Kapur entropy algorithm is proposed to construct multi-thresholds segmentation. Levy flight strategy was introduced to improve the position update way in improved butterfly optimization algorithm (IBOA), which could promote the global optimizing ability. The Gauss chaos mutation mechanism was applied to disturbance the elite individual and enhance the population diversity, which could avoid the convergence of the algorithm to the local optimum. The optimization performance of IBOA was tested by benchmark functions. Kapur entropy was used as the fitness function of IBOA, and an image segmentation method based on IBOA and Kapur entropy maximization was designed. The image segmentation performance was verified by using three classic Berkeley images and one magnetic resonance image. The improved algorithm was applied on the infrared image segmentation of electrical equipment. It is proved that under the condition of the non-uniform backgrounds and noise interference, the improved algorithm can also effectively promote the segmentation accuracy and efficiency of infrared image so as to ensure the success rate of fault diagnosis of power equipment.