A Photon Counting Lidar Point Cloud Data Processing Method

Lidar has been widely used in military and civilian fields due to its advantages of high angular resolution and high measuring accuracy.Unlike traditional lidar,the photon-counting lidar uses Geiger-mode avalanche photo diode(Gm-APD)as the core photoelectric detection device to detect long-range targets and weak echo signals.It is an important method to realize the 3D information perception of the target.The traditional point cloud data processing algorithms can't effectively complete the 3D image reconstruction task under the environment of strong background noise and low signal-to-noise ratio.In order to make up for the above shortcomings,this thesis deeply studies the working principle of photon-counting lidar,conducts researches on point cloud data processing methods based on the theoretical model of photon-counting lidar.Firstly,in view of the problem that traditional algorithms can't accurately estimate the depth of the target and the reconstructed depth image has a great error,a photon-counting lidar depth image denoising algorithm is proposed.According to the different distribution characteristics of signal photons and noise photons in time domain,the echo signal photons are extracted from the original point cloud data without increasing the complexity of the system.By solving the convex optimization function related to time-of-flight of photons,the algorithm reduces the depth estimation error and the reconstruction accuracy of depth image is improved.Secondly,in order to solve the problem that prior parameters reduce the robustness of the traditional algorithms in depth image reconstruction,a photon-counting lidar depth image reconstruction algorithm based on convolutional neural network is proposed.Based on the flexibility of convolutional neural network in the task of denoising and reconstruction,the point cloud data set of photon-counting lidar is created,and the depth image reconstruction convolutional neural network is trained with the simulation of lidar detection process.The proposed method improves the practicability of photon-counting lidar and the depth image reconstruction of different targets in different situation is realized.Finally,based on the time-correlated single-photon counting technology and the principle of pulse laser ranging,a prototype of the 3D information sensing system is redesigned and built.Through point cloud data processing,the depth information perception of buildings at 562 m and 8.4km is realized.