Research On Target Detection Algorithm Of Gm-APD Array Lidar

As an active imaging technology,lidar has become a research hotspot in the field of target detection.Gm-APD(Geiger-Mode of Avalanche Photodiodes)array lidar,as a lidar with single photon sensitivity and sub-picosecond time resolution,has great application value in long-distance detection.This paper applies this detector to Obtain the intensity direction and distance direction of the long-distance detection target,the target detection algorithm of DSP platform is studied,and the algorithm is simulated..This article first introduces the working principle of Gm-APD array lidar,and at the same time establishes a Poisson probability model of Gm-APD array lidar based on the Poisson properties of photons.According to the actual project requirements,the detection target features are long-distance and small targets,and only a small number of pixels are occupied in the 64 64 detection array.Based on this feature,this paper proposes to use the sparseness of the target,the photoelectron number and avalanche obtained through cumulative measurement For Poisson probability,a sparse Poisson reconstruction algorithm is used to inversely derive the actual intensity information of the detected target.According to the actual detection environment,this paper has carried out simulation research.The Frobenius norm(F norm)of the error matrix between the estimated reflectance distribution and the reflectance distribution of the real target is 3.08,which is proved by comparison with median filtering and accumulation approximation.The algorithm has smaller errors and higher accuracy.Secondly,due to the sparseness of the detected target,this paper proposes to use the orthogonal matching pursuit(OMP)algorithm to calculate the sparse estimation solution of the target distance,through the range of the solution,after filtering out incorrect distance values,through the sparse Poisson reconstruction algorithm solves the distance value of the detected target.By changing the size of the regularization parameter and analyzing the relationship between the reconstructed distance value noise and accuracy,it is concluded that as the regularization parameter increases,the distance value accuracy increases,but the noise increases significantly.Finally,using DSP which has the advantages of an independent hardware multiplication unit and the ability to operate on floating-point numbers.In this paper,the designed algorithm is simulated and verified on the DSP platform,and the simulation results are analyzed.