Recognition Method Of Road Condition For Active Safety System

At present,the active safety system has large limitations.Developers usually calibrate the system with good road conditions,which is difficult to adapt to different working conditions.The road surface condition has a great influence on the vehicle’s handling stability and braking performance,and is closely related to driving safety.The road surface recognition system is a subsystem located in the active safety system.Accurately identifying the road surface can effectively expand the working range of the active safety system.Therefore,it is of great significance to explore pavement recognition methods.In this paper,the reason-based image processing is used to identify the road surface and the effect-based vehicle dynamics response is used to estimate the road surface adhesion coefficient.The independent operation and verification of the two methods are completed.The results are fused,and the fusion results have the advantages of both methods.Regarding road surface recognition based on image processing,the recognition process in this paper is: after inputting the original image,the first convolutional neural network performs semantic segmentation on the image and extracts the road surface area in the image,and sends the extracted image with only the road surface area to The second convolutional neural network classifies the pavement,maps the pavement type to the adhesion coefficient by looking up the table and combines the vehicle speed,and outputs the corresponding adhesion coefficient value and other reminder signals.The article first established the data set required by the image recognition method,compared,modified and trained different neural network models for the tasks of road classification and semantic segmentation,and explored the training methods and modifications of the convolutional neural network to improve accuracy according to the tasks of this article The strategy explores the factors that affect the accuracy of the recognition system.Finally,write code to realize that the system can output the corresponding attachment coefficient and other reminder information for the system input video,direct connection camera or input picture.In terms of estimating the adhesion coefficient based on the vehicle dynamic response,this paper estimates the road surface adhesion coefficient based on the recursive least square method of the vehicle longitudinal dynamic response,and realizes the recognition of the road surface adhesion coefficient based on the vehicle dynamic response through thejoint simulation of Carsim and Matlab / Simulink.Since the vehicle involves acceleration or deceleration during driving,the vehicle speed,wheel speed,utilization coefficient,slip rate,longitudinal force,and normal force can be obtained according to the dynamic model and the driving balance equation,in the linear range of small slip rate The slope of the μ-s curve can represent the adhesion coefficient;recursive least squares method is used to estimate the adhesion coefficient in the saturation range of large slip rate.Because different recognition methods have different characteristics,the article finally puts forward a strategy that combines the results of vehicle dynamic response method and image recognition method.The image recognition method has predictability of road surface condition and dynamic response method under partial steady-state conditions Combining the advantages of more accurate recognition results,the fusion strategy is verified using a common pavement mutation condition,and it is concluded that the fusion strategy can identify the pavement more accurately and have the ability to predict the pavement mutation condition.