基于压缩感知的激光复合成像超分辨三维重构方法  

Super-resolution 3D reconstruction method of laser composite imaging based on compressed sensing

作　　者：唐昌华[1] 吴琼[1] 时兵[1] 陈坚[1] 李洪亮[1] TANG Changhua;WU Qiong;SHI Bing;CHEN Jian;LI Hongliang(College of Humanities&Information Changchun University of Technology,Changchun 130122,China)

机构地区：[1]长春工业大学人文信息学院,长春130122

出　　处：《激光杂志》2023年第9期109-113,共5页Laser Journal

基　　金：吉林省教育厅科学技术研究项目(No.JJKH20221281KJ)。

摘　　要：三维重构的图像分辨率一般较差,为获得更精确的三维重构图像,基于压缩感知设计激光复合成像超分辨三维重构方法。计算随机目标下重建距离的条件分布概率,获得能量方程,对其进行正则化处理,同时得到列向量以及稀疏矩阵的解析解,构建区域相似度目标方程;基于压缩感知技术重建超分辨率图像像素,得到相应的测量矩阵,两个点集三维坐标之间的关系式,计算两个点集的最小目标误差函数,获取点集间对应的判定条件,设计激光复合图像三维重构算法。在选定的实验图上选择三个点位,对比这三个点位在该方法以及另外三种方法下的重构效果,实验图像显示,压缩感知方法的精度明显优于其他三种方法。在像素均方根误差的计算中,压缩感知算法在三个点位下的误差值分别为1.64、2.28、3.17,均小于其他方法,证明该方法的像素误差更小,精度更高。The image resolution of 3D reconstruction is generally poor.In order to obtain more accurate 3D recon-struction images,a super-resolution 3D reconstruction method of laser composite imaging based on compressed sensing is designed.The conditional distribution probability of reconstruction distance under random target is calculated,the energy equation is obtained,and the regularization process is obtained,and the analytical solution of column vector and sparse matrix is obtained to construct the regional similarity target equation.Based on the compressed sensing technology,the super-resolution image pixels are reconstructed,the corresponding measurement matrix and the rela-tionship between the 3D coordinates of the two point sets are obtained,the minimum objective error function of the two point sets is calculated,the corresponding judgment conditions between the point sets are obtained,and the laser com-posite image three-dimensional reconstruction algorithm is designed.Three points are selected on the selected experi-mental plot,and the reconstruction effect of these three points in this method and the other three methods was com-pared.The experimental images show that the accuracy of the compressed sensing method is significantly better than the other three methods.In the calculation of the root mean square error of pixels,the error values of the compressed sensing algorithm at three points are 1.64,2.28 and 3.17 respectively,which are smaller than other methods,proving that the pixel error of this method is smaller and the accuracy is higher.

关 键 词：压缩感知算法 激光复合成像技术 超分辨率 三维重构 均方根误差 

分 类 号：TN249[电子电信—物理电子学]