Study On The Geometric Calibration Of Thermal-infrared For The Reconstruction Of Three-dimensional Temperature Field

With the development of the computer technology, the three-dimensionalreconstruction technology has been more and more widely used. Moreover, as theinfrared thermal imaging technology improved continuously, thethree-dimensional temperature field reconstruction based on the infrared thermalimages has shown its dominant advantage in all kinds of field. Thus, during thethree-dimensional temperature field reconstruction, the calibration of the infraredthermal image is the first step, which is also the most important step.In this paper, we accomplish the calibration of the binocular infrared thermalimage with the computer vision library OpenCV, using the binocular visualsystem to execute the three-dimensional temperature field reconstruction. Firstly,we analyze the camera model and the thought of calibration, in which we focuson the camera calibration method based on the Zhengyou Zhang plane calibrationmethod, to propose a new camera calibration algorithm combining the Zhangcalibration method and the angular point-based sub-pixel feature extractionalgorithm. We perform the experiment with the self-made calibration boardapplying to the infrared thermal image and obtain the calibration results of thecamera, including the internal parameter of the infrared thermal image,deformation coefficient, spin matrix and translation vector etc, to accomplish thecalibration of the infrared thermal image. We also testify the accuracy of thecalibration results with the actual distance measurement and the theoreticalarithmetic. Secondly, we perform the feature extraction and the stereo matching basedon the feature points, which are the basis and key to the three-dimensionalreconstruction. By analyzing and contrasting the Harris, SUSAN and SIFTangular point detection algorithm and the actual experiment, we select the SIFTalgorithm to find the feature point in the corresponding image and solve theproblem of the infrared image texture scarce, and then accomplish the stereomatching experiment of the infrared image. In order to accurate the matching, theexperimental image needs adjusting before the matching to keep the matchingpair in the same epipolar line.Finally, after obtaining the correct matching point, we need to accomplishgetting the three-dimensional information of the points in the space with thetriangulation measurement principle, and give each point the correspondingtemperature value. Meanwhile, we denoise the obtained three-dimensionaltemperature field image with the denoising algorithm based on the k-d tree, toaccomplish the experiment of the three-dimensional temperature fieldreconstruction.