Research On Smoke Vehicle Detection Technology Based On Video Image

Video-based smoky vehicle detection technology has a very important significance for environmental protection,and has broad application background in intelligent transportation.The smoky vehicle intelligent detection system can work for a long time and maintain the accuracy of the detect results,and is suitable for large-scale deployment of urban roads.Based on the monitoring video collected from traffic bayonet,this paper studies the vehicle detection and tracking,vehicle shadow elimination and the design of smoky vehicle classifier,mainly including the following aspects:(1)Research on vehicle detection and tracking algorithms.To overcome the shortcoming of ViBe algorithm that only uses the gray information of image and produces ghost region easily,an improved method of ViBe target detection is designed.Then the morphological operation is used to optimize the vehicle detection effect.The vehicle area is extracted by contour extraction method.Finally,a fast vehicle tracking method is designed for the application scenario.(2)Research on vehicle shadow elimination algorithm.In view of the fact that the vehicle detection algorithm can not suppress the defects of vehicle shadow,and the existence of shadow area will affect the characteristics of the vehicle,thus affecting the subsequent smoky vehicle determination,a shadow detection method combining RGB color information and uniform LBP texture information is designed.And the shadow elimination rule designed in this paper is used to eliminate the detected shadow areas.(3)Research on classification methods of smoky vehicles.Through the analysis of LeNet-5 network structure,an improved LeNet-5 network is designed,mainly through introducing Inception module to redesign the network structure and re-select the network activation function and classification function to improve the classification ability of the network.And the construction method of the smoky car sample set and training method are discussed.(4)Software design and implementation.Based on the above algorithm,a smoky vehicle detection software is designed and implemented,which realizes the functions of moving target detection,shadow elimination,vehicle detection and tracking,smoky vehicle detection and so on.Finally,the software system was tested for functional and performance through the video of the actual traffic bayonet.