Research on Area-Matching Algorithm Based on Feature-Matching Constraints

According to the characters of the menology and combining with the advantages of two matching methods, a new method of feature-constrained area matching is proposed. In the preprocessing part, the original images are processed by histogram equalization algorithm in order to increase the contrast, enhance the texture character and satisfy the needs of feature-based matching. In the feature-based matching part, the corners which are detected by Harris algorithm are matched with the monotonicity and uniqueness constraints. With the bidirection constraint the disparity map is processed to eliminate the mismatching, increase the accuracy, and lay a foundation for area-based matching. In the area-based matching part, with the constraint of the results for feature matching, the normalized covariance operator is used for area-based matching. Lastly, the disparity is post processed to eliminate the mismatching and fill in the point which is not matched with kinds of constraints, and then the accurate and dense disparity map is obtained. The experiment results indicate that the algorithm performs well and satisfies the needs of automatic navigation in the Lunar Rover experimental lab.