The optimal ladder type edge detection algorithm is researched. Gauss filter is used to smooth images and the finite difference of the first order partial derivative is used to calculate the gradient magnitude and the direction of each pixel. Non maximum value suppression is performed to gradient amplitude and double threshold algorithm is used to detect edges, which avoids double borders or unclosed area problems from other operators. Sinusoidal encoding method is used to convert infrared gray images to pseudo color images, and then the temperature jump area from the fuselage front to the engine is obtained. And the center of the area is taken as a tracking point to finish missile terminal guidance.