Aiming at the needs of high-altitude mountain scene matching navigation task and solving the problems of scarcity of man-made objects and non-significant texture in mountainous areas, we propose a comprehensive image feature combining multiscale gray level co-occurrence matrix, correlation surface features and local/global statistical features, and take the matching probability as the regression target, and construct a matching probability prediction model by using the regression of the support vector machine to realize suitability analysis of the area. The experimental results show that a small number of training samples can be used to predict the matching probability of mountainous scenes, which has a higher performance improvement compared with the traditional single-scale texture features, and can effectively provide data support for the reference map database preparation, route planning of scene matching navigation.