Research On Obstacle Reconition In Front Of Intelligent Vehicle

The key technologies of intelligent vehicle involve in the perception,decision-making and control techniques,in which the perception technique is a basis.The obstacle detection and recognition is an important part of the perception technology.To realize the intelligent driving and effectively detect the obstacles in front of the vehicle,the method of obstacle recognition based on monocular vision in view of the characteristics of obstacles on the road of intelligent vehicle is studied in this thesis.Firstly,the regions of interest for vehicles and pedestrians are identified.The images in front of the vehicle collected by the vehicle-mounted camera are pre-processed.By analyzing the characteristics of the images,the methods,such as the gray scale processing,median filtering and the gray scale equalization,are selected and applied to the collected images.For the preprocessed images,the interested regions of obstacles are determined by using the extracted rectangle feature of the shadow on the bottom of the vehicle and the depth-width ratio of the pedestrians.Secondly,the texture and HOG features of the target are extracted.In this thesis,by contrasting and analyzing the recognition results of vehicles and pedestrians,the texture and HOG features are finally selected for recognition of the obstacles,and then the two features of the target are extracted respectively in the regions of interest of the target.Thirdly,the Bayesian classifier is improved.Based on the texture and HOG features,the obstacles are first classified and identified by the traditional Bayesian classifier.It is found that the recognition results may be further improved.Thus,the traditional Bayesian classifier is modified by a weighted method with multi-feature fusion,thereby an improved version of Bayesian classifier.The obstacle recognition is realized based on the improved Bayesian classifier.Finally,the experimental verification of the proposed algorithm is carried out.The hardware system of obstacle recognition in front of the intelligent vehicle is constructed byconnecting the camera,vehicle power converter,computer and vehicle.An identification program is programmed by using the Visual studio 2010 software.The algorithm is tested and verified based on the system.The results show that,comparing with the traditional one,the improved Bayesian classifier has greatly improved in the recognition accuracy.Based on an improved Bayesian classifier,this thesis presents a recognition method for the obstacles in front of a vehicle,which might provide a reference for the research of decision-making and control technologies of intelligent vehicle.