Signal extraction algorithm of Gm-APD lidar with low SNR return

Abstract

Geiger mode Avalanche Photo Diode(Gm-APD) laser radar in the case of high background noise during the daytime causes the peak of data distribution is firing by the noise but not the signal, this phenomenon leads to difficulty in extracting the peak characteristics of the signal, lost target pixels of three-dimensional range image and the low image signal-to-noise ratio by using the traditional non-parametric estimation signal extraction algorithm. Aiming at this problem, the bump hunting combination algorithm(BHC) is proposed, which extracts echo signal position under low SNR echo data, removes noise pixels and provides favorable conditions for target recognition by combining the convex features of echo signals and pixelwise cross correlation. According to the experimental results, when using 500 frames of echo data, BHC improved 15.10 % of the target reduction degree and 21 times of the image signal-to-noise ratio compared peak-picking method, increased 3.83 % of the target reduction degree and 15 times of the image signal-to-noise ratio compared the maximum likelihood estimation(MLE) algorithm.