Research On Methods Of Vehicles Detection At Urban Road Intersections Scenes Based On Video

The target detection and counting of urban road traffic scenarios is one of the important research content of intelligent traffic monitoring system.Accurate real-time the detection and counting of vehicle at urban road intersection is a prerequisite for reducing traffic congestion,intelligent transportation management and smart urban construction.Therefore,Researchs on methods of vehicles detection at urban road intersections scenes based on video play an important role in the building of urban intelligent transportation systems.In order to obtain the accurate detection and counting of vehicle at complex urban traffic intersection scene,this dissertation focuses on background modeling method under the condition of slow-moving or temporarily stopped vehicles and sudden or gradual illumination changes scenes and the way of accurate vehicle counting with Foreground Time-Spatial Image.This paper did in-depth research on these contents and the innovative works are summaried as follows:(1)Aiming to efficiently resolve the problem that the subtraction background model is easily contaminated by slow-moving or temporarily stopped vehicles,the Gaussian Mixture Model with Confidence Measurement(GMMCM)is proposed for vehicle detection in complex urban traffic intersections scenes based on Gaussian Mixture Model(GMM).According to the current traffic state,each pixel of background model is set a confidence measurement.Whether to update the background model and the corresponding adaptive learning rate depends on if the current pixel point is in confidence period.Using the real-world urban traffic videos and CDnet2014 public databases with intersections scenes,the qualitivative experiments are conducted by GMMCM,compared with three commonly used models including Gaussian Mixture Model(GMM),Self-adaptive Gaussian Mixture Model(SAGMM)and Local Parameter Learning Gaussian Mixture Model(LPLGMM).The experimental results showed that GMMCM excels others latest improved GMM methods in keeping the background model being unpolluted from slow-moving or temporarily stopped vehicles.All experimental results including qualitative and quantitative evaluations show the effectiveness and superiority of GMMCM in slow-moving or temporarily stopped vehicles detection of the complex urban traffic scenes,compared with SDC(Sigma-Delta with Confidence)、ViBe(Visual Background extractor)、GMM、SAGMM and LPLGMM.(2)Aiming to efficiently resolve the problem of real-time of the vehicle detection and the subtraction background model is easily contaminated by slow-moving or temporarily stopped vehicles,The Pixel-Based Adaptive Segmenter with Confidence Measurement(PBASCM)is proposed for vehicle detection in complex urban traffic intersections scenes based on Pixel-Based Adaptive Segmenter(PBAS).The background of PBASCM is modeled based on the history of recently observed pixel values and each pixel in the background model is assigned a confidence measurement based on the current traffic state.The foreground decision depends on an adaptive threshold,whereas the background model is updated based on the current traffic state and whether the corresponding pixel point is in the confidence period.Using real-world urban traffic intersections videos,the overall results of real-time and detection accuracy analyses demonstrated that PBASCM achieved better performance in both qualitative and quantitative evaluations,compared with CB(Codebook)、GMM、ALW(Adaptive Light-Weight)、SDC、ViBe and PBAS.Thus,our experimental results demonstrate that PBASCM is effective and suitable for real-time vehicle detection in complex urban traffic scenes.(3)An Adaptive Local Mean binary Pattern Background Model(ALMPBM)is proposed to resolve the deficiency in current background models,which are easily contaminated by the sudden or gradual illumination changes in complex urban traffic intersections scenes.According to the Weber’s law in discriminating the intensity of illumination changes,we first develop Adaptive Local Mean Binary(ALMB)pattern textures,and then the Adaptive Local Mean binary Pattern Background Model(ALMPBM)is modeled based on a robust consensus method.The experimental results including qualitative and quantitative evaluations on real-world urban traffic intersections videos show that the proposed ALMPBM offer the best performance compared to GMM、ViBe、LBPBS(Local Binary Pattern Background Subtraction)、LTPBS(Local Ternary Pattern Background Subtraction)、SLPBS(Scale Invariant Local Patterns Background Subtraction)and LBSPBS(Local Binary Similarity Patterns Background Subtraction).(4)A Foreground Time-Spatial Image(FTSI)is proposed for counting vehicles in complex urban traffic intersections scenes to resolve deficiencies of traditional counting methods,which are highly computationally expensive and become unsuccessful with increasing complexity in urban traffic scenarios.A self-adaptive sample consensus background model,which can deal with slow-moving or temporarily stopped vehicles and the sudden or gradual illumination changes for each pixel,is constructed on the Virtual Detection Line(VDL)in the frames of a video.The Foreground of the Virtual Detection Line(FVDL)is then collected over time to form a FTSI and counting the number of connected components in the FTSI reveals the number of vehicles.Based on real-world urban traffic intersections videos,the experimental results including qualitative and quantitative evaluations on real-world urban traffic intersections videos show that the proposed FTSI method offer the best performance in counting vehicles compared to TSI(Time-Spatial Images)、ETSI(Edge Time-Spatial Images)、ACC(Area Change Counting)and TVA(Tracking Vehicle Area).