Infrared images have been widely used in fault detection of insulators. However

it is difficult to deal with large volumes of images collected by Unmanned Aerial Vehicle（UAV） just using the traditional manual diagnosis way. Consequently

we proposed a novel method to automatically recognize and detect insulators in infrared images collected by UAV. First

the image edges are extracted from infrared image using Laplace operator

then

a histogram of edge density along the lines crossing the image is constructed

periodic textual features are extracted

and the insulators are finally recognized by a two-step clustering method. Infrared images with different backgrounds（plant

farmland

tower） are selected as the test data

and the recognition rate by our method is over 85%. Experiments show that our method can automatically recognize insulators from the UAV infrared images with a complex background

and can be further applied for intelligent infrared diagnosis using UAV.