<p>The extraction of fault regions is crucial for the automatic fault diagnosis of substation equipment. The segmentation accuracy of substation equipment is easy subject to the features of infrared images, which may affect subsequent fault diagnosis. To address this issue, we propose a fault region extraction method combining improved superpixel pre-segmentation with enhanced fuzzy clustering. Firstly, based on the correspondence between temperature information and grayscale information in infrared images, density-based spatial clustering of applications with noise (DBSCAN) superpixel segmentation is employed to merge pixels in regions with similar temperature characteristics. By calculating the pixel mean of each region, the computational complexity is reduced. Secondly, in order to further extract the high temperature part of the equipment accurately, fuzzy c-means (FCM) is utilized to cluster different temperature regions. In view of the lack of image spatial information and limited robustness of FCM, objective function is reconstructed by combining image spatial information and domain pixel influence. The spatial distance is introduced to control the influence of neighborhood pixels adaptively and maintain the balance between noise and detail in the image. Experiments are conducted on a dataset of substation equipment infrared images. Compared with the classical clustering segmentation algorithms, the region extracted by the proposed method is clearer and more complete, which effectively improves the segmentation accuracy and ensures the real-time performance.</p>