Research On Dynamic Detection Method Of Highway Smog Visibility Level

In recent years,the proportion of the traffic accident rate in the haze weather in the total traffic accident rate has increased year by year.Due to the reduced visibility,people’s travel and activities cannot go smoothly.In particular,low visibility makes it even more difficult for autonomous vehicles to drive on highways.Therefore,this paper has done research on the visibility detection method of haze in front of autonomous vehicles on highways.The main research contents are as follows:(1)Aiming at the problem that the existing haze visibility detection equipment or method is difficult to realize real-time dynamic detection,we have studied a dynamic detection method of highway haze visibility based on threshold band segmentation.Firstly,through image preprocessing,the maximum connected area is obtained to determine the position of the "threshold band",and the gradient value is calculated through the threshold band;secondly,the threshold value is obtained by analyzing the calculated gradient value matrix and the relationship between it and the visibility value of the haze is established.Determine whether there is haze in front of the self-driving car and determine the visibility level of the haze.The simulation results show that the detection accuracy of this method for road visibility levels in a non-interference environment reaches 91.88%,and the average detection time for a single image is 0.332 s.(2)Aiming at the problem of low accuracy of the traditional methods for the dynamic detection of highway haze visibility levels in a non-interference environment,We have studied an improved AlexNet network and applied it to the visibility level detection of haze.First,the image is flipped along the vertical axis and moved by 30 pixel units in the horizontal and vertical directions for data enhancement.Second,the number of neurons in the classification output layer in the AlexNet network is adjusted to 5 according to the haze visibility level,and the last layer is full The learning rate of the weights and bias in the connection layer is adjusted to 20,and finally the initial learning rate and other parameters are set.The simulation results show that the improved AlexNet network has an accuracy rate of 98.5% for the visibility detection of highway smog and haze in a non-interference environment.(3)Aiming at the problem of low accuracy of dynamic detection of highway haze visibility levels in interference environments,we have studied a method for detecting haze visibility levels based on vehicle information.Firstly,detect the area of the interfering vehicle,calculate the distance between the vehicles,then use the WCK-Alexnet network to classify the clarity level,and finally establish the mathematical relationship between the distance,the clarity level and the haze visibility level,and use it to judge the level.The simulation results show that the accuracy of this method for the detection of highway haze visibility levels in interference environments is improved by more than 6.25%.