A SVM based extrinsic calibration method for RGB-D camera

Abstract

Due to the popularity of target detection, 3D reconstruction and SLAM, RGB-D cameras are used more and more widely. Accurate calibration is the prerequisite for subsequent applications for RGB-D sensors, which install the color camera and depth camera on the same device. There are many methods to calibrate RGB devices and the results perform well. When it comes to depth camera calibration things got worse. It often requires specific calibration objects and complicated processes. In this paper, a convenient calibration method for depth camera is proposed, which applies the calibration algorithm of RGB camera to depth camera, and mainly solves the difficulty of finding corner points for depth camera. SVM classifier is used to fit the chessboard line according to the edge points and locate corners. Our experiment result shows that this method can calibrate camera extrinsic parameters accurately.