Feature Matching technology of Visible and Infrared Fusion Image based on VIFB

Abstract

In view of the needs of all-time navigation tasks, visual navigation technology based on visible and infrared images has received extensive attention. At present, visible and infrared fusion technology is mostly used in remote sensing detection, target recognition and other fields, but it has not been widely used in the field of navigation and lacks a representative technology evaluation system. Therefore, a feature matching performance evaluation framework for visible and infrared fusion image based on the VIFB framework is proposed in this paper. A variety of typical indoor and outdoor target scenes are designed to test 20 visible and infrared image fusion algorithms. Then the ORB feature matching method is used to test the feature matching effect of image pairs before and after fusion. A variety of qualitative and quantitative evaluation indexes are designed to evaluate the VIFB framework. The feature matching performance of the images before and after processing is comprehensively evaluated. The experiments show that the fusion image can effectively deal with poor lighting and smoke occlusion. In various test scenarios, NSCT_SR, HMSD_GF and LatLRR algorithms show better feature matching performance, and CNN, MST_SR algorithms better integrate visible details and infrared salient structural information. The images fused by ADF, DLF, FPDE algorithms have good grayscale equalization effect.