Research On The Method Of Accuracy Improvement Of Time-of-Flight Camera By Combination Of A High-Resolution Camera

Time-of-flight(ToF)camera is a compact depth camera that can obtain three-dimensional information.It can obtain 3D information of scene at a high frame rate by emitting near-infrared light signals and calculating the time(phase)difference between the emitted and returned light of each pixel.ToF camera can get both amplitude and depth images at the same time,and it does not depend on ambient light.Therefore,ToF camera has broad application prospects in navigation,medical technology,automation and other fields.However,due to the limitations of its own imaging technology and characteristics,ToF camera has the problems of error and low accuracy in imaging.This article mainly involves research on the difficulties and shortcomings of ToF technology,based on the imaging principles of ToF camera,depth errors caused by multiple factors,and pinhole camera model,focusing on the distance-related error and the low resolution of ToF images.Experiments have been carried out to verify the proposed method.The main tasks completed are as the following:1.A method for depth error correction of each pixel is proposed: the amplitude image of ToF camera is employed as the calibration image and the homography based camera calibration method is used for intrinsic parameters calibration and distortion correction;the ToF camera is used to capture images at different distances,we establish a Gaussian noise model of the depth image to optimize the depth image and use the least square method to fit the depth map taken at each position as a reference plane,so we obtain distance-error data at different distances;the distance-error model for prediction is fitted by polynomial function to correct the actual depth image.2.Aiming at the problem of low resolution of ToF images: The intrinsic and extrinsic parameters and relative pose are obtained by joint calibration of combined a traditional high-resolution camera with fine and rich texture features and ToF cameras.ToF amplitude image and depth image are used for coarse scene features point extraction and depth determination,respectively.The sub-pixel feature points of the ToF images and the fine feature points extracted from the high-resolution image are corresponding by the relative transform matrix.A method of pixel distance weighting is proposed for feature point depth correction and the final corrected feature points are back-projected onto ToF images.