Sonar Image Feature Detection and Matching for Acoustic Structure from Motion

This project mainly focuses on the prerequisite of feature-based ASfM algorithm.

This project mainly focuses on the prerequisite of feature-based ASfM algorithm. To overcome the challenges brought by imaging sonars, firstly, the idea of constant false alarm rate (CFAR) algorithm is adapted from radar image processing for object segmentation and denoising for sonar images. Secondly, a self-defined loss is designed and introduced to remove the outlier matches of two sets of feature points from A-KAZE detector.