To solve the problem of fuzzy edge information,weak contrast,image noise and difficult to extract thermal infrared image edge by the traditional methods,an edge detection method is proposed by combining multi-level trapezoidal fuzzy enhancement,fuzzy C-means clustering and Sobel operator.Firstly,multi-level trapezoidal fuzzy enhancement is applied in thermal infrared image,then the interesting temperature zone of thermal infrared is classified by using fuzzy C-means clustering to extract the object,finally the edge of the target is detected by Sobel operator.Based on MATLAB,the result shows that the method has higher detection accuracy.It can detect edge information of fuzzy image,and can extract detail information which other methods can’t extract.It obtains single pixel edge and has a small amount of calculation.The method has an ideal effect.