The invention relates to an adaptive electrical equipment fault detection method based on Meanshift. The method comprises the steps that step 1, the infrared image of fault electric equipment is acquired, wherein the power fault area is the area of high brightness in the infrared image; step 2, a weight factor based on the neighborhood grayscale is introduced to extend a mean shift algorithm so as to obtain an extended mean shift model; and step 3, a descending clustering threshold segmentation mechanism is utilized so that the extended mean shift model is enabled to quickly cluster the fault area and effective extraction of the fault area can be realized. According to the extended mean shift method assigning sample weight, the area is clustered according to the grayscale characteristics by using a self-iteration clustering strategy and finally the fault area can be completely extracted, and the descending threshold segmentation mechanism according to the temperature difference is put forward so that online inspection, data input and analysis and other operation can be facilitated for the operation and maintenance personnel.