Research On Traffic Sign Detection Algorithm Based On Multi-feature Fusion In Haze Weather

Traffic sign detection,as an important branch of intelligent traffic system research,has already become one of the hotspots of research at home and abroad.Traffic sign detection usually combines traffic signs' color characteristics and shape features,extracts the traffic signs accurately from natural scenes,and identifies the detected traffic signs through traffic sign recognition system.Therefore,the detection of traffic signs is of great significance in the aspects of automatic driving,maintenance of traffic signs,standard traffic behavior,and safe driving.However,the real traffic scene is complicated and changeable,because the traffic signs are usually in complex environment outdoors,the process of traffic signs detection is easily affected by illumination conditions,weather conditions and the rotation of directions,especially in recent years,with the accelerated process of industrialization,the problem of air pollution is becoming more and more serious,the images extracted from haze weather feature low contrast and poor visibility,causing serious problems for the detection of traffic signs.In view of the above problems,this paper starts from two aspects of image restoration and traffic sign detection in haze weather,and makes in-depth research on traffic sign detection in haze weather.First,this paper studies the problem of image restoration in haze weather,and proposes an improved image dehazing algorithm which based on dark channel prior.Based on the study and research of image dehazing algorithm in atmospheric scattering model and others based on atmospheric scattering model,combines the features of images which under detection in haze weather,to propose a fast image dehazing algorithm on the transmissivity of refinement,both ensure image clarity not affected by the haze weather,improve the detection accuracy of the traffic signs,and meet the real-time requirements in intelligent transportation systems.Secondly,the traffic sign detection algorithm based on color feature,shape feature and direction gradient histogram in haze weather has been respectively studied in depth in this paper.Though traffic sign detection algorithm based on color in daze weather has been preprocessed by image dehazing,the accuracy compared with the traditional traffic sign detection algorithm has some improvement,but due to be influenced by the light conditions,caused a certain degree of interference on the stability of the intelligent transportation system.The traffic sign detection algorithm based on shape feature can avoid the influence of illumination,but for traffic signs with deformation or severely covered,the accuracy will be more reduced.In addition,the complexity of the algorithm is relatively high,which can not meet the high real-time requirements of the intelligent transportation system.The traffic sign detection algorithm based on hog feature of directional gradient histogram in daze weather not only reduces the restriction of light conditions,improve the accuracy of traffic detection,but also reduces the complexity of traffic sign detection algorithm in a certain extent,but it still has some certain restriction for highly efficient intelligent transportation system.Finally,the traffic sign detection algorithm based on multi-feature fusion in daze weather proposed in this paper,preprocess the image through image dehazing,respectively extracts HOG feature of directional gradient histogram and local binary pattern LBP feature,do the fusion and then output test results.After Multi-feature fusion,the traffic sign detection algorithm can not only resist the effects of light conditions and daze weather,improve traffic sign detection accuracy,reduces the false detection rate and leak detection rate,but also not restrict the efficiency of the intelligent transportation system and has high real-time due to the complexity is relatively low.