It is of great significance to obtain environmental information along railway lines using vehicle mounted LiDAR.However,the point cloud data collected by LiDAR is affected by equipment and environmental factors,leading to a lot of noise interference.These noises seriously affect the subsequent perception and monitoring tasks.To this end,an efficient denoising method for large scale point cloud of railway scene is proposed in this paper and a novel grid projection strategy is designed to efficiently reduce dimension and sampling for large scale railway point cloud data.Then,an improved clustering algorithm based on GPU is designed to quickly identify and separate outlier noise data.Finally,a multi strategy fusion method is designed to remove noise data effectively.The proposed method makes full use of the characteristics of forward movement of railway scenes to perform grid based spatial temporal compression on point cloud data,and further uses the matrix operation of GPU accelerated clustering algorithm to realize real time and efficient large scale point cloud denoising algorithm of railway scenes.The experimental results show that the proposed method not only improves the performance of denoising,but also greatly enhances the processing efficiency.