A Feature Matching Method Based on Rolling Guided Filter and Collinear Triangular Matrix Optimal Transport

Unmanned aerial vehicle(UAV) cross-view object matching is widely used in applications such as field, disaster detection, and express delvery. The feature-matching method is a key technology for UAV cross-view object localization. This paper proposed a feature-matching method based on collinear triangular matrix optimal transport. First, a rolling-guided filter is introduced to calculate the keypoint response, which addresses the problem of scale difference between the two images; Then, the response of keypoints is employed to calculate the network loss, which addresses the problem of viewing angle differences between images. Finally, collinear triangular matrix optimal transport is proposed to advance the accuracy and speed of the algorithm. The experimental results show that the proposed method have at least 4% higher AR than the existed method.