Visible-Infrared Images Matching Based on Deep Learning

This paper discusses the importance of visible-infrared imaging sensor in UAV cluster system and its application in post-disaster rescue. The technology is characterized by rapid deployment and flexible scheduling, which can quickly respond to disaster events and provide support. However, due to the differences between the two images, it is still a challenge to achieve their feature matching. Deep learning-based methods have become the mainstream research direction to solve this problem. This paper introduces a method based on local feature matching and Transformer to realize feature point matching of visible and infrared images. In this paper, ResNet-FPN structure is used to extract local features of the image, which can effectively obtain rich information and simplify the feature map. The extracted features will be sent to transformer to add location information and attention mechanism. And then feature point matching results can be obtained through coarse-grained and fine-grained matching. The effectiveness and performance of the proposed method are verified by experiments, and good results are obtained on multiple data sets.