Research On Road Parameter Recognition Algorithm Based On Visual Inertial Navigation Information Fusion

The safety performance of vehicles has always been the focus of people’s attention.The effective identification of road surface parameters is an important aspect of the safe and stable running of vehicles,among which the identification of road surface type and road adhesion coefficient is the main content of the research of road surface parameters.The analysis of the current situation of road condition detection at home and abroad shows that the accuracy of traditional road type detection is low,so the realization of accurate and efficient driving of intelligent vehicles has become one of the hot issues in this field.Nowadays,visual inertial navigation information fusion technology is widely used in vehicles.How to achieve accurate recognition of road adhesion coefficient is the focus of this thesis.This thesis mainly carries out research from three aspects:Firstly,aiming at the low detection accuracy of different pavement types in traditional methods,the algorithm model of YOLOv5 is used to effectively identify the problem.In this thesis,image sensors were used to collect different types of pavement information,and YOLOv5 algorithm was used to detect five types of pavement,namely asphalt pavement,concrete pavement,soil pavement,stone pavement and snow and ice pavement.In this process,Labelimg was used to mark the image features separately for 20,000 pictures collected first,so as to eliminate the influence of interference information on classification and recognition tasks.Then,the training and evaluation process of YOLOv5 pavement type detection was realized by python programming.Experimental results show that the algorithm can detect the pavement type accurately and effectively.Secondly,the pavement parameters based on vehicle dynamics are studied,and the extended Kalman filter algorithm is used.By constructing the seven-degree-of-freedom dynamic model and Dugoff tire model,the force analysis of moving vehicles was carried out,and the real-time recognition of road adhesion coefficient was realized.The algorithm was simulated in Car Sim and Matlab/Simulink software,and the experimental data were verified by angular step condition and serpentine condition.The experimental results show that the extended Kalman filter can be used to accurately identify the road adhesion coefficient,and the relative error of the recognition results can be controlled within 5%,which can deal with the situation of road mutation,and verify the feasibility and effectiveness of the model.Finally,in order to further improve the accuracy of the road adhesion coefficient recognition algorithm,this thesis proposes a road parameter recognition algorithm based on visual inertial navigation information fusion.The pavement image recognition results were incorporated into the dynamic recognition algorithm as a limiting condition,and the range values of common pavement types were established by using the comparison table of pavement categories and range values of adhesion coefficient at different running speeds,so as to improve the accuracy of recognition of pavement adhesion coefficient.By comparing the reference value table with the state data of the vehicle in operation,the experiment verifies that the fusion algorithm can effectively improve the accuracy of the identification value of the adhesion coefficient and optimize the convergence speed of the algorithm.