GM-APD lidar reconstruction algorithm with few frames based on sample function

Under the conditions of few frames and low signal-to-background ratio (SBR), when the Geiger-mode avalanche photodiode (GM-APD) lidar reconstructs the range image, the statistical form of its echo data shows sparsity and discreteness. The probability distribution histogram of photon triggering is prone to losing the target peak, making it difficult to estimate the position of the target peak accurately, which affects the detection of small and weak targets. To address this problem, the concept of peak prominence of the echo signal was defined, and an algorithm based on the sample function was proposed. This algorithm first recovers the lost target peak, filters the target peak to remove small spikes, and calculates the peak prominence of the local peaks. The local peak with the largest peak prominence is the target peak, thus accurately estimating its position of the target peak. The experimental results on the self-built GM-APD laser radar detection imaging platform show that in a real scene with SBR = 0.0263, the three evaluation indicators of the proposed algorithm are superior to SPIRAL. The target restoration degree (K) is 43 % higher than SPIRAL, the peak signal-to-noise ratio (PSNR) is improved by 9.5 dB, and the mean absolute error (MAE) is reduced by 95bins. This confirms the importance of this algorithm in the field of few-frame and low signal-to-noise ratio GM-APD detection imaging and has broad application prospects.