Abstract: In order to solve the problemswith traditional image matching algorithms, such as too many false matching point pairs, and poor robustness in complex environment, an image matching algorithm is proposed based on improved FAST to extract feature points, combing with opposite color features.First, the corner points extracted by the improved FAST algorithm are used as feature points, and the feature points are described by combining with the improved Opponent SIFT algorithm. Then, the extracted feature point pairs are roughly matched based on the character location algorithm, to reduce the risk of false matchingof the feature point pairs in the whole matching process. Finally, in order to avoid the problem that the correct point pairs are mistakenly removed caused by the RANSAC algorithm easily falling into the local optimal solution, the vector field consistency is used instead of RANSAC for purification of matched point pairs to reduce the false matching rate. The comparison experiments show that the accuracy of the improved algorithm is greater than 91%, and it has good robustness and adaptability for the variation of differences.

Key words: feature point matching, improved FAST algorithm, opposite color feature, Opponent SIFT algorithm, vector field consistency