Registration Of Array Laser Three Dimension Imaging Point Cloud

The three-dimensional lidar system is composed of a laser pulse or a series of imaging beams emitted by a lidar. The two-dimensional plane information of the echo signal returned from the target is used to obtain the 3D data from the laser radar ranging. Lidar has the advantages of directivity, coherence, monochromaticity and brightness, so that the anti-jamming capability of LIDAR is stronger than that of ordinary microwave radar, and it is more helpful for target detection and recognition, such as detection Submarines, stealth aircraft, geological exploration, missile tracking and other fields. The array laser radar technology with three-dimensional imaging by arrayed laser has the advantages of high imaging speed and high reliability, which makes it widely used in military field, national economic construction field and agro-forestry ecological field. However, due to the complexity of the 3D laser radar environment, if we want to improve the detection capability and imaging accuracy of laser radar, in addition to improving the performance of laser radar(LIDAR), it is very important to improve the performance of laser radar. We also need to find a more effective method of echo signal processing and laser three-dimensional imaging construction, so this paper focuses on this key issue, starting from the basis of the registration processing method, to carry out the array research on registration technology of laser imaging point cloud.According to the imaging characteristics of array 3D laser imaging system, a new method based on local distance is proposed, which is based on the three-dimensional imaging system. In this paper, the initial registration method of point cloud for feature description histogram. At the same time, by analyzing the influence of the system performance index on the registration result, an adaptive threshold registration method is proposed. The simulation results of array laser 3D cloud point and scanning laser imaging point cloud verify the effectiveness of the algorithm. Finally, the registration method proposed in this paper is applied to the target pose estimation. Through a large number of experiments and analysis, the above methods can achieve accurate and efficient point cloud registration, and can register the point cloud data from different imaging perspectives to the same coordinate system. The three-dimensional image can be obtained with more information. The main contents of this paper include the following aspects:(1) Theoretical analysisThrough analysis of array laser imaging system imaging principle, the simulation model of array laser three-dimensional imaging system is established. The hardware performance indexes of the imaging system are fully considered in the model, including the system timing capability, range resolution and lateral resolution. The simulation process can truly reflect the performance of imaging results. By comparing the 3D registration data of different types of targets and different imaging scenes, the performance of several typical registration methods are compared and analyzed from the aspects of registration accuracy and registration efficiency.In this paper, a new point cloud registration method based on local range feature description histogram(LRFH) is proposed for the specific working mode of array laser three-dimensional imaging system. In the registration process, the optimal sampling uniformity(OSAC) method is proposed. For accurate registration method, an accurate registration method based on adaptive threshold(ADT-ICP) for 3D registration of laser point cloud is proposed by theoretically analyzing the effect of ranging resolution and imaging lateral resolution on the imaging effect in imaging system. Finally, by analyzing the basic principle of target attitude measurement, a three-dimensional attitude measurement method based on point cloud model matching is proposed. The influence of system imaging parameters on attitude measurement can be used to guide the rational design of the system in practice.(2) Experimental simulation analysis part:For verification, the data from the experiment are the point cloud data obtained by array laser 3D simulation of the target model in the NASA database, and the scanning laser imaging point cloud data in the 3D database of Stanford University. For the above two types of imaging points The results of registration are measured by registration accuracy and surface penetration. The validity and accuracy of the proposed algorithm are verified. The applicability of the proposed method is verified by the imaging data with distance error. Finally, the registration accuracy of the proposed method is verified. Array laser imaging point cloud data, the validity of the proposed attitude estimation method is verified.