Facing the lunar surface survey of the Lunar Exploring Engineering, the paper summarizes the environment sensing technology based on vision image. For the image matching is the most important step in the process of the lunar exploring images, the accuracy and speed of the matching method is the key problem of the lunar exploring, which play an important role in the rover auto navigating and tele-operating. To conquer difficult problem that there are significant illumination variation of the imaging, lack of image texture, and non-uniform distribution of the image texture, the huge change of the disparity for the prominent target in the scene, in the image process Engineering, the image matching method is proposed which divided the whole image into M×N regions, and each region employs the Forstner algorithm to extract features， by which the semi-uniform distribution features of whole image and avoiding of the features gathering is achieved. According to the semi- uniform distribution features, the Sift and Least Square Matching method are used to realize accurate image matching. Guided by the matched features of the first step, the locale plane is detected to restrict dense image registering. The matching experiments show that the method is effective to deal with the image captured by the lunar exploring rover, that has large variation of illumination and lacking of image texture. The robustness and high accuracy of the method is also proved. The method satisfied the request of the lunar surface exploring.