Research On Recognition Technology Of Rail Invaders Based On Binocular Stereo Vision

In the process of rapid development of China's railway industry,security is always the top priority.At present.the accidents caused by railway invaders occur frequently,which seriously threaten the safety of train operation.However,the existing monitoring methods have great limitations:manual inspection is costly and inefficient,grid monitoring is only applicable to contact objects,and sensor monitoring is difficult to be applied in a wide range.Aiming at video surveillance,this paper applies the binocular stereo vision technology to rail invader detection,solving the problems of false detection of suspended objects and inability to locate invaders by single camera.Besides,the result of three-dimensional reconstruction is used to estimate the volume of the invader.The research content and innovation points of the paper are as follows.(1)In consideration of the nationwide fog and haze,the dark channel prior is used in the image preprocessing to eliminate the fog and haze noise,enhancing the contrast of the image and increasing the image details.(2)An automatic rail detection algorithm is presented,which combines mathematical morphology processing and Progressive Probabilistic Hough Transform to effectively eliminate the impact of sleepers and gravels,accurately realizing the automatic extraction of the rail in the scene.Based on the result of the extraction,the intrusion limit is implemented.(3)An improved Visual Background Extractor(ViBe)algorithm is proposed for object detection.The improved algorithm eliminates the "Ghost" problem of the traditional ViBe algorithm when there is a moving object in the first video frame,getting accurate extraction of the foreground target.The result of object detection is used to initialize the Kernel Correlation Filter(KCF)to realize real-time tracking of the target.(4)A binocular stereo vision system was established.On the basis of camera calibration and image rectification,stereo matching and 3D reconstruction are applied to restore the real scene in stereoscopic world.And target tracking,distance measurement and volume estimation in the three-dimensional space are then realized by using the reconstruction result.(5)A novel method for dynamic invader detection in three-dimensional space is proposed.Comparing to the existing algorithms,the proposed method reduces the amount of calculation to a large degree,for it only needs to compare the basic coordinate values but not to calculate the space plane equations.The research of this paper is based on the real railway scene.The proposed algorithms fully consider the characteristics of the railway scene,and the parameters of other algorithms involved are also adjusted according to the actual conditions.In summary,the overall research has a strong practical application value.