Research On The Construction Method Of Infrared Three-Dimensional Temperature Distribution Model Based On Point Cloud Processing

Thermal infrared images are widely used in many fields,such as medical treatment,agriculture and military affairs.Compared with visible light images,thermal infrared images are notsensitive to light and are suitable for shooting at night.Due to the many advantages of thermal infrared images,combining two-dimensional temperature images with three-dimensional reconstruction technology to construct an infrared three-dimensional temperature distribution model can obtain more accurate temperature distribution information.In this paper,the method of constructing the infrared three-dimensional temperature distribution model is deeply studied,and through the method of homography transformation,point cloud processing,etc,the information fusion of infrared image and depth image is realized.The specific work content is as follows.First,in order to obtain the accurate homography matrix,the process of infrared and visible light registration is studied.Based on Super Point algorithm,Super Glue algorithm and improved RANSAC algorithm,the extraction of image feature points,matching of feature points and generation of homography matrix are realized.In view of the feature that Super Glue can filter out outliers while matching feature points,the RANSAC algorithm is improved.The improvement is made to add a loop when calculating the number of interior points,so as to improve the accuracy of image matching.Through comparative analysis through experiments,it is verified that this method can improve the accuracy of the homography matrix.Secondly,in order to construct the infrared three-dimensional temperature model,the fusion method of depth image and infrared image is studied.First,the depth image and the infrared image are processed separately.The depth image is converted into a point cloud.According to the relationship between the gray value of the pixel point and the temperature value of the pixel point,the color corresponding to the temperature value is given to form a pseudo-color image.The infrared image undergoes a homography transformation.Then,the processed infrared image is fused with the point cloud,and the phenomenon of partial color loss in the point cloud is processed.Finally,a preliminary infrared three-dimensional temperature model is formed.At last,after analysis,the preliminarily generated infrared 3D temperature model is more dependent on the imaging effect of the depth camera,and the generated 3D model has shortcomings such as voids and fractures.Therefore,a point cloud processing module is added on the basis of the generated infrared 3D temperature model,and the point cloud processing is studied in depth.In this module,a series of processing of point cloud is performed on outlier point removal,cluster segmentation and completion reconstruction.A point cloud completion method based on projection interpolation is given,and finally a better3 D temperature model is obtained.The model can not only accurately reflect the temperature distribution of the target object,but also present the three-dimensional shape and spatial position information of the object.This is very meaningful for many application fields.