Research On The Method Of Three-dimensional Reconstruction At Case Scene Based On The Kinect

With the rapid development of society and improvement of Science and Technology,three dimensional reconstruction at case crime scene has become an important means of criminal investigation.The traditional three dimensional reconstruction based on photos has some limitations,especially in the reconstruction precision and reconstruction time;therefore it cannot meet the requirements of public security practice.However the acquired range image based on Kinect can directly reflect the distance information.On this basis,the 3D model at the crime scene can be reconstructed with the help of range image,so it can provide strong technical support for the public security organs to solve cases and improve the case handling efficiency.This paper focused on the 3D reconstruction method based on Kinect,on the basis of descriptions on the OpenNI frame and PCL point cloud library,the author made deep research on the filtering method of depth information based on Kinect and registration method based on3 D point cloud.Finally,the 3D model with color information was studied.Main procedures were as follows:First of all,Kinect related hardware and basic theory were introduced in detail.The installation of the OpenNI somatosensory frame was completed then the PCL point cloud Library was successfully configured.Then the depth information and color information in real time were acquired.Secondly,the depth image was filtered by different filtering methods,and the experimental results were analyzed with histogram.On this basis,an improved hybrid filtering method was proposed.The experiment proved that the improved filtering method could obtain better filtering effect,and it laid good foundation for after 3D reconstruction.Then,coordinate transformation principle was studied,and 3D point cloud data based on depth information was acquired,then improved ICP registration algorithm was proposed.The registration experiments of point cloud data in different coordinate systems were carried out,and then the experimental results were compared with those based on the traditional ICP registration algorithm.Finally the experimental results showed that the improved ICP registration algorithm could get more accurate registration results.Finally,point cloud rendering was carried out based on the mobile cube algorithm,and then the 3D model with Mesh structure was generated.Compared with the model based on Microsoft's own SDK,the color information was added on this basis.The 3D point cloud withcolor information was gained,and finally the 3D reconstruction of the small range was realized.