A new calibration method for depth sensor

作者：Ye Qin^*; Gui Po Po

来源：Journal of Optoelectronics - Laser, 2015, 26(6): 1146-1151.

DOI：10.16136/j.joel.2015.06.0088

• Calibration method
• Central projection
• Co-ordinate system
• Constraint conditions
• Depth sensors
• Indirect adjustment
• Intrinsic parameters
• Projection equations

Depth sensor is a good choice for the same application, such as mobil robot navigation. It also should be calibrated before using. In order to enhance accuracy of factory intrinsic parameters, we attempt to present a new calibration method for depth sensor, for example Kinect 2.0 which consists of two cameras an RGB camera and a depth camera). For Kinect 2.0, we should calibrate its RGB camera and depth camera separately. In our method, the intrinsic parameters of depth camera will be used in RGB camera calibration, so we will calibrate the depth camera firstly. For depth camera, we assume a fixed-length line in its view, and we regard the distance between two line endpoints which own the same value as the constraint condition, and solve the intrinsic parameters of the depth camera with indirect adjustment. Transform depth image points to 3D space points (Kinect coordinate system) with intrinsic parameters and associate them with RGB image points, and then calibrate RGB camera based on central projection equation. Experimental results show that when depth image points are transformed to 3D space points, the precision of the lines is about 2.5 mm, while when depth image points are transformed to RGB image points, the precision of points is about 1 pixel, which is better than that of Microsoft Kinect SDK. If you want to get high-precision Kinect intrinsic parameters in the same applications, our method may be a good choice.

• 出版日期2015
• 单位同济大学