| Traffic safety is closely related to weather environment.In the case of fog,the visibility of the environment is reduced,and the driver’s vision is affected.It is difficult for auxiliary driving vehicle detection equipment to give warning in time,the traffic accident happened.Fog leads to the decline of image quality,contrast and visual effect.In view of this situation,this thesis proposes a fog road image enhancement algorithm based on improved SSR,vehicle detection is performed on the enhanced image by using YOLOv3 model.Compared with the fog image,the enhancement result of the proposed algorithm has higher recall rate and accuracy.The research contents of this thesis are as follows:1.Considering the real-time requirement of the application scenario,this thesis chooses to improve the SSR algorithm,using Bilateral filtering function instead of Gaussian filtering function.The principle of the algorithm is as follows:convert foggy image to HSI space,take a linear stretch for the S component,enhancing I component with improved SSR algorithm,convert back to RGB space,using sigmoid function to restore color of image,get enhanced image.The experimental results show that the performance of this algorithm is better than other traditional algorithms in visual effect and objective indicators.2.The YOLOv3 network model is established,and the appropriate training parameters are selected for the network model.The automatic driving data set BDD100K is used to train and test the network model.Vehicle detection experiments are carried out on foggy images and enhanced images.The experimental results show that:compared with the detection results of fog image,the detection results of the enhanced image have lower miss detection rate and higher confidence.The average accuracy of the network model can reach 85.55%,and the frame frequency can reach 30fps.The network model has good accuracy and real-time performance,and the detection effect is better than the direct detection of fog image. |
PDF Full Text Request