Research On Classification Method Of Vehicle Driving Pavement Characteristics Based On Intelligent Perception

At present,the research direction of active suspension is mainly focused on the feedback of control algorithm to road information,and less on the active detection of road changes.Therefore,from a certain point of view,the lack of sufficient road surface information is the main factor restricting active suspension to further improve vehicle driving performance.This paper relies on the National Natural Science Foundation of China(Grant No 51605213)and Liaoning Province Science and Technology Department Joint Fund(Grant No 201602367),mainly aiming at the road ahead,using binocular vision system,through acquisition.The road two-dimensional image realizes the reconstruction of three-dimensional road surface information,and then classifies the characteristics of vehicle driving road surface,laying the foundation for the effective combination of vehicle active suspension and driving road surface.The specific research contents are as follows:Firstly,a vehicle-borne stereo camera test platform is built based on the binocular vision system,and the binocular vision system is calibrated online.A large number of road two-dimensional images are acquired through the real vehicle experiment as image samples for subsequent research.Secondly,the SGBM algorithm is used to obtain the road depth information.The method of restoring the three-dimensional road structure from the two-dimensional image and constructing the road vertical contour model is studied.The "clean" algorithm is used to effectively remove the wrong data points in the road vertical contour model.Thirdly,the characteristics of road gradient and bump are extracted,and the road feature data set is constructed based on these characteristics.The road feature data set is divided into four categories: uphill,downhill,straight road and continuous deceleration belt by using the Gauss mixture clustering method.Finally,multi-class support vector machines are trained with road feature data sets,and the training results are analyzed.The research results show that the two-dimensional road image acquired by binocular vision system can be used as the basis of intelligent perception of road surface characteristics.The classification model trained by road feature data set can also extract useful information from road image and classify it correctly.