Research On 3D Reconstruction Method Of Case Scene Based On Point Cloud Data

In the current public security work,3D reconstruction technology,as an important means of collecting and preserving evidence on the scene,is facing problems such as low model reconstruction rate and lack of detailed surface texture information.The registration of point cloud data is a key step in the 3D reconstruction process,and the quality of the results directly affects the quality of the reconstruction model.In order to effectively solve the above problems,improve the reconstruction accuracy of the scene objects and scenes,speed up the reconstruction,and improve the work efficiency of the police,this paper studies an improvement based on FPFH features.The point cloud registration algorithm optimizes the preprocessing and registration process of the point cloud to improve the accuracy and speed of point cloud registration.The main work and innovations of this thesis are as follows:Firstly,PCL and its dependent libraries are compiled and installed.The configuration and function test of the PCL point cloud library running environment are completed,and the point cloud processing function is called normally in Windows and Linux system.Secondly,the concept of a spatial point cloud hybrid filter is proposed for the limitation that a filter can only filter one type of noise.The voxel filter is used to streamline the redundant point cloud,and the invalid outlier point cloud data is removed by statistical filtering,which improves the overall quality of the spatial point cloud.Thirdly,through the analysis of the inadequacies of different registration algorithms,an improved point cloud registration algorithm based on FPFH features is proposed.Based on the denoising and de-interference of the point cloud by the hybrid filter,the KD tree is used to establish the spatial index of the point cloud data,and the FPFH histogram and the corresponding point estimation algorithm are innovatively combined and used.The corresponding point rejection principle eliminates the error matching,finds the best feature point pairs,improves the effective matching rate of the feature point pairs,and achieves accurate and fast registration of the two point clouds.Finally,the improved point cloud registration algorithm based on FPFH feature is used to achieve the registration of point clouds with different angles and the registration of multiple point clouds.The registration accuracy is high and the surface texture details are clear.