Design And Development Of Water And Snow Pavement Identification System Based On Caffe

With the economic development and the cross-integration of various road trunk lines,more and more people choose to travel by car,accompanied by this,the frequency of traffic accidents has increased year by year,according to statistics,on average,there is a traffic accident every three minutes in China,and the state of the road in bad weather conditions is an important inducement,the vehicle is more prone to accidents under this condition;In addition,with the development of intelligent driving technology,it is becoming more and more important to obtain real-time water and snow pavement status in intelligent driving,guide the normal driving of the car by obtaining pavement status.Therefore,the detection and identification of the real-time condition of pavement plays a great role in providing realtime warning information for travel vehicles,ensuring road safety operation and personal safety.In recent years,the rapid development of deep learning and big data has provided a new method for the realization of pavement condition detection.Using digital image processing and deep learning technology,this paper puts forward a method for detecting the state of water and snow pavement based on the framework of deep learning.First,rewrite the content of the recognition module to suit the detection requirements and compile it into a dynamic-link library file for detection system calls;secondly,various pavement state data sets are collated and produced,the sample pictures are classified and labeled,and largescale data training is carried out after conversion of image data format to generate pavement state recognition model files;then the system initializes the detection network and model file,combined with the obtained camera real-time stream data,detects and identifies the pavement state of the frame screen,and output pavement recognition results,if there is a trigger warning section,you can obtain its alert information,save the screenshot and video file at that time,and record the recognition results in the system log;finally display the identified frame screen in the interface.The system shows good performance in real-time and uninterrupted identification of water accumulation and snow pavement,and provides pavement information.,and can provide theoretical and technical support for pavement management monitoring and intelligent driving.