Under low-brightness conditions,it is difficult to obtain high-quality survey data by traditional accident survey methods.This paper proposes a low-brightness accident scene reconstruction method based on UAV-borne LiDAR. First,the methodological framework of UAV-borne LiDAR survey is established.Then the noise is removed using a statistical filtering algorithm with Gaussian distribution,and the moving objects around the scene are filtered out by judging the occupancy status of the spatially divided voxels.Finally,the sensor's position data is used to registration the point cloud data to obtain a 3D point cloud model of the accident scene.This paper also explores how UAV flight altitude and LiDAR side overlap rate affect the modeling accuracy.An empirical study is conducted at a simulated accident scene at night,and it is found that when the UAV flight height is 15 m and the LiDAR side overlap rate is 50%,the accuracy and processing time can reach a better state.Comparison and analysis with aerial photography modeling and traditional manual survey methods,the root mean square error(RMSE)of airborne LiDAR modeling is 0.04636,which is lower than that of aerial photography modeling,which can be better applied to the survey of traffic accidents at the scene of low illumination.