Research On Gm-APD Laser Radar Signal Extraction Algorithm

Gm-APD laser imaging radar has become the current research focus because of its sensitive response,high precision and easy integration.This paper mainly research the Gm-APD laser radar signal extraction algorithm for low signal-to-noise ratio data with high background light.When signal is weak,the signal extraction becoming a difficult point of Gm-APD ladar development,and the peak picking has method signal extraction effect is not well.Therefore,three algorithms for improving the signal extraction quality are proposed.This paper research the low SNR echo data signal extraction algorithms:Firstly,the research status of Gm-APD and the research status of its reconstruction algorithm are investigated.The main research contents of this paper are determined through the analysis of the research status at home and abroad.At the same time,the GmAPD firing principle,the principle of imaging radar system,Poisson distribution firing model,traditional Gm-APD signal extraction algorithm and objective evaluation index of reconstructing 3D image are researched.Secondly,according to the shortcomings of the peak method and the characteristics of low SNR data,a first-order weighted Gaussian matching filtering algorithm is proposed based on the related idea,then the false alarm rate expression of the detection rate is derived.When the target is located at the middle of the gate,compared with the peak picking method,the first-order weighted Gaussian matching filtering algorithm has a higher detection probability and a lower false alarm probability;When the noise which is in front of the gate that firing counts is too high and dense,the first-order weighted Gaussian matching filtering algorithm can not get the right signal.According to the Poisson distribution model,the maximum likelihood estimation algorithm can be obtained.The estimated signal value and noise value that affect the performance of the algorithm are analyzed.The results show that there is the best combination value;when there is a little difference between the data and the theoretical.At the same time,the bump hunting search algorithm is obtained according to the weak signal convex feature,and the influence of the variance value on the signal extraction is analyzed.The result shows that there is the best combination value.Since threshold method can not denoise the low signal-to-noise ratio echo data well,a cross-shape neighbor denoising algorithm is proposed to effectively improve the denoising and retaining target capability.Finally,the signal extraction performance of the four algorithms is compared.Firstly,in order to figure out the applicability of the three algorithms proposed in this paper,the data which PSNR is up to 69.50 is used under fixed frames;20 frames data in two different scenarios.The three new algorithms can improve the target reduction of the peak target by ?18%,it has applicability.Secondly,use the data which PSNR is up to 10.17.The reconstruction results of 500 frames show that the bump hunting search algorithm has the best target reduction after denoising,which can be higher than the peak method ?17%.Finally,use the data which PSNR is up to 1.50.The reconstruction results of 1000 frames show that the bump hunting search algorithm has the best target reduction after denoising,which can be higher than the peak method of 30%.In order to analyze the results of different frames more comprehensively,the low signal-to-noise ratio echo data in two scenes is used to compare the reconstruction results of different frames,and the signal extraction effect of the bump hunting search algorithm is best.