Identification Of Common Domestic Traffic Warning Signs Based On Fully Convolution Network

There are four types of traffic signs.The first type is usually a circle or rectangle with a white pattern on a blue background,which is often referred to as a traffic sign.It is used to indicate that the vehicle can go straight ahead,detour,or turn left and right,etc.;The second type is an equilateral triangle with mostly yellow and black edges,which is called a traffic warning sign.For example,it warns that there may be falling rocks in front of a moving vehicle,and a sharp turn and a slow moving ahead.Circular,called a prohibition sign,such as reminding drivers to prohibit traffic,parking,etc.;the fourth type is a guide sign that is generally rectangular on a white background with a blue background,such as indicating the junction and distance between cities,towns,and villages The distance of the target location,etc.,indicates the direction for distant drivers.In summary,of the four types of traffic signs,traffic warning signs are particularly important in terms of vehicle driving safety.They are mainly used to remind drivers that road conditions ahead are not conducive to driving and road sections may be dangerous.Drive carefully.Before the deep learning method was widely used,most of the research on traffic warning signs in China was based on the edge information of triangle warning signs to detect and identify,for example,to locate the warning signs by finding their centroids and vertex positions,and then The well-located signs are identified,but the real-time performance of this method is poor.In recent years,people have begun to use neural networks to detect and recognize traffic warning signs.Although it takes less time,it still fails to meet the basic real-time requirements,and the accuracy is relatively low.In order to reduce the incidence of traffic accidents caused by ignoring traffic warning signs,the main data source of this article is the common traffic warning signs in China under natural backgrounds,and they are detected and identified.I hope that the images can be quickly detected while ensuring the accuracy The target is detected and identified.We hope that the detection and recognition method of traffic warning signs and its model structure can be applied to the auxiliary traffic driving system,and the detection results of traffic warning signs can be notified to the driver as soon as possible under the condition of ensuring accuracy,so as to assist the driver better Understand the road conditions ahead to further ensure driving safety.After understanding the theory and structure of neurons and neural networks,this paper chooses the Full Convolutional Network to locate and detect common traffic warning signs in China.Using common traffic warning signs in China under natural circumstances as a data set The model is trained.First,in order to enhance the generalization ability of the model,the image was enhanced with data.Second,the convolution layer and the residual connection layer were alternately connected to extract the color and shape features of the image data.Finally,the feature pyramid network structure Perform cross-scale predictions to predict object positioncoordinates and class probabilities on different scales.The experimental results show that the method uses the deep learning capability of the full convolutional neural network to achieve fast and accurate recognition of traffic warning signs,and can basically meet the real-time requirements for detection and recognition of common traffic warning signs.