Three-dimensional Reconstruction Method And Application Based On Laser Point Cloud Image Fusion

At present,many 3D scene reconstruction methods have appeared in the academic circle,but due to factors such as the cost of sensors and the size of the application range,it is difficult to convert academic achievements into productivity and apply them to projects.Lidar can scan and collect laser point cloud data that can accurately describe the scene in any complex environment.However,the environment described by the laser point cloud only has distance and intensity information,without color texture information.The image data can be collected by a sports camera,which has rich image texture information,but cannot describe depth information.Therefore,the use of laser scanners equipped with lidar and sports cameras to collect data and perform data fusion has become a research focus in the field of 3D reconstruction.(1)A synchronous data acquisition system for sports camera and lidar is constructed.The system is composed of multiple sensors such as lidar,sports camera,satellite guide and IMU.The sampling frequency and time error of different sensors are significantly different.The system integrates the algorithm of the laptop computer to control the motion camera to collect data through WiFi into the LOAM algorithm,so that when the lidar starts to collect data,the motion camera also simultaneously collects image data.Then the collected laser point cloud and image data are optimized after the algorithm is processed to achieve data fusion.(2)A laser point cloud data segmentation method combining manual and automatic is designed.First,collect the point cloud data of the calibration board at different positions by moving the calibration board,and realize the point cloud filtering based on the guided filtering point cloud algorithm,and then realize the point cloud registration operation based on the ICP algorithm.Then the manual method is used to segment the point cloud of the calibration plate,and then the Kmeans clustering algorithm is used to automatically and accurately segment the point cloud of the calibration plate.Finally,the point cloud center of the calibration board and the coordinates of each corner point are solved by the iterative algorithm of the point cloud center.(3)A joint calibration method of sports camera and lidar based on LM algorithm is designed.First,the distortion correction of the image data collected by the sports camera is realized through the OpenCV optimization algorithm,and the corner coordinates of the image are identified and solved based on the Hairrs algorithm.Then multiple sets of point cloud corner coordinates of the calibration board and the corresponding image corner pixel coordinates are calculated by direct linear transformation to calculate the initial calibration value between the two sensors,and the minimum difference between the point cloud reprojection coordinates and the image pixel coordinates is constructed.Two multiplication function.At the same time,the LM algorithm is improved and optimized by introducing the damping factor and reducing the reference value.The improved LM algorithm is used to perform nonlinear optimization of the least square function to find the minimum value,and then the optimized joint calibration result is solved.Finally,under the data fusion system based on the principle of collinear equations,the calibration results are used to calculate the pixel coordinates of the image corresponding to the laser point cloud coordinates through the image pixel and point cloud homography calculation method,and the image RGB value is assigned to the corresponding laser Point cloud,finally realize the three-dimensional reconstruction of point cloud image fusion.Experiments show that compared with the initial value of the calibration result,the reprojection error is reduced by 35%,and the 3D reconstruction effect has been significantly improved.