Aiming at the problems of poor accuracy and low robustness of traditional AGAST feature matching algorithm, an image matching algorithm based on bilateral filtering and AGAST-BEBLID is proposed. Firstly, bilateral filtering is used to de-noise and enhance the image edge detail effect. Secondly, BEBLID algorithm is used to create efficient binary descriptors in the feature extraction stage to generate better local feature descriptions. Then GMS algorithm combined with Hamming distance is used to filter the KNN matched images to achieve coarse feature matching. Finally, GC-RANSAC algorithm is used to fit the local optimal model in the mismatching elimination stage, and the image features are accurately matched. The experimental results show that the overall average accuracy of the improved algorithm in complex environments is 10.57% higher than AKAZE, BRISK 17.20% and SIFT 19.45% higher, respectively.