| Single photon lidar is a new interdisciplinary technology which integrates single photon detection technology,pulse laser technology,electronic engineering and signal processing.With the characteristics of high ranging accuracy,strong anti-jamming ability and long range,single photon lidar has been widely used in remote sensing detection,weapon guidance,terrain mapping,urban modeling and other military and civil fields.The detection range and imaging quality of single photon lidar have been challenged in various applications with the development of science and technology.When the background noise of remote imaging is large,the current traditional imaging methods need to integrate a single pixel for a long time in order to accurately estimate the depth information and obtain high-quality images.However,in practical applications,the integration time of system imaging is often limited,resulting in the number of signal photons is too small and submerged by noise,which seriously affects the depth estimation and imaging quality.To solve this problem,the depth estimation and image reconstruction methods of single photon lidar remote imaging are studied in this paper.The principle of remote scanning single photon lidar is constructed to realize the longdistance 3D imaging.In order to figure out the method to limit the backscattering,an imaging experiment is carried out by using the detector gating.The imaging difference between the free running mode and the gating mode is also analyzed quantitatively.The experiment shows that the signal strength of gating mode is 3.3 times higher than that of free running mode.To meet the needs of high noise background and fast imaging,a time-dependent Kalman depth estimation method is proposed based on the data collected from the prototype.The photon counting set is made by the time-of-flight correlation of photons,and the adaptive Kalman filter is used for depth estimation to provide a depth map for subsequent image reconstruction.Experimental results show that compared with the traditional maximum likelihood estimation,the root mean square error of depth estimation is improved by about 40%.For some pixels with missing information and large information difference in the above depth estimation image,a method of detect these pixels is proposed by calculating the degree of neighborhood discreteness,and then the super pixels in the depth image will be constructed based on the gradient of the intensity image to reset the depth information.Finally,based on the Poisson distribution of lidar photon counting,the prior convex optimization problem of intensity information is constructed and solved to complete the image reconstruction.The experiment shows that the reconstructed image greatly reduces the image noise,smoothes the depth image while retaining the edge information,and successfully reconstructs part of the damaged image details. |
