Results and Discussions When the noise intensity is 20000 counts/s, the signals after 200 counts are subjected to noise reduction processing. Compared with original data, the noise intensity and variance of preprocessed data decrease, and the noise distribution is transformed from a binomial distribution to a Gaussian distribution (Fig. 2). After noise reduction processing, the Gaussian filter distribution is determined using the Gaussian filter parameters. As the Gaussian filter width increases, its distribution becomes closer to the Gaussian distribution (Fig. 3). We further verify the effect of noise reduction on the detection performance using the Monte Carlo experiment, the results show that the detection probability of proposed method is 4-6 times higher than that of bin-grouping constant false alarm detection method [Fig. 6(a)]. Compared with current constant false alarm detection methods, the detection probability of proposed method increased by approximately seven times [Fig. 6(b)]. Based on actual experimental point cloud data, the noise reduction process can filter out many interference points (Fig. 8). Therefore, the weak target detection results with linear array radar data can be improved compared with the conventional method (Fig. 9). The detection results show that the highest detection probability increases by 7.4 times compared with D-CFAR [Fig. 10(a)]; moreover, the rate of change in the false alarm probability decreases [Fig. 10(b)].

Since the development of single-photon radars, humans can detect more distant targets. However, the echo signals weaken, and the target detection technology requires urgent improvement. The target detection theory of single-photon LiDAR has been adopted from classical microwave radar. Many detection theories in classical radar have been transferred to quantum LiDAR and achieved good results, but research on single-photon LiDAR detection theory is not as advanced as that of classical radar. Current quantum laser detection methods still have limitations, such as the inability to precisely detect targets when the cumulative number of times is few; thus, it is difficult to detect echo targets with low signal-to-noise ratios. In addition, the current methods used to improve the detection probability are applied through signal-noise reduction, which in turn, complicates the detection process. Therefore, it is essential to improve the target detection performance at different accumulation times and noise intensities to overcome the shortcomings of the current detection methods.

This study focused on the target detection of photon-counting LiDAR in a noisy background. First, according to the echo characteristics of the single-photon LiDAR, noise reduction was processed using numerical comparison and a Gaussian filter, which significantly decreased the noise intensity while maintaining the echo target information. Next, based on the processed signal which has obeyed a Gaussian distribution, we applied CFAR to determine the target. In CFAR process, we proposed an algorithm to estimate parameters of target and noise more accurately, thus further improving the detection ability. Finally, the detection performance was verified through experiments and simulation comparisons.

This study demonstrates that the target can still be detected with a high detection probability in the case of strong noise or weak echo signal. Compared with the conventional methods, the proposed method improves the average detection probability of the target by 40% in the same signal-to-noise ratio environment, and the noise intensity is decreased.