Kinect V2 Edge base Registration with RANSAC-ICP

This article proposed a method to increase the precision of RGB-D registration applying RGB and depth obtained from Kinect V2. Procedure consists of 2 major components. The first part is to convert the image depth into the resolution and field of view (FOV) of color images and find both edges. Second part is to use both edge images from the first step to adjust registration result together with RANSAC (Random sample consensus) and ICP (iterative closed point) to reduce error between both edges. The experimental results showed that the proposed method improved the accuracy of RGB-D registration.