Research On Multi-lane Detection Method Based On A Hybrid Model

Vehicle intelligent assisted driving technology is one of the key technologies to improve the existing traffic problems,and the environmental perception module is responsible for providing road information for the auxiliary driving system.Lane detection technology is the main means to perceive the road lane information,which is of great significance to the key technologies such as road identification,path planning and standardized driving.With the deepening and gradual product of automobile intelligent assisted driving technology,the combined application of this technology and other cutting-edge technology has become the hot research of the automobile industry.The requirements for lane detection have also changed,with higher requirements for the number,fit and robustness of degree of lane line detection.This paper focuses on multi-lane detection,considering the problem of insufficient robustness and low fit,image pretreatment,lane feature detection,lane fitting model and lane tracking,thus,a multi-lane detection method based on hybrid model is constructed.The main studies are as follows:1.The current research status of the key parts of the lane detection is summarized.The current lane detection algorithm is summarized from image pretreatment,lane target feature extraction and lane postprocessing technology,and analyzes the shortcomings,and puts forward the research objectives.2.This paper establishes a lane detection and tracking system based on a hybrid model.To meet the requirements of higher robustness in lane detection algorithm and better fitting results,First,To reduce the time-consuming factors and to filter out the interference factors,A dynamic area of interest based on historical frame detection and OTSU and contrast stretching are designed in the preprocessing phase;Secondly,the Combine Hove line transformation and sliding window feature screening to extract the lane feature pixels in the image;Further,Combining a linear line and a curve fitting model based on the Student-t noise hypothesis,Build hybrid models to depict lane lines,Reduce the noise impact,Improve the model fit;Finally,Lane line tracking based on Kalman filtering.3.This paper presents a side lane detection and determination method based on lane information and color constraints.First for the side lanes,In pretreatment stage,OTSU superposition gray scale diagram and bidirectional sobel operator were adopted to filter interference factors,Enhance the side lane edge features;Secondly,the During the feature extraction phase,Using Hove straight line transformation and screening combining the lane line detection results and the lane information constraint;Then,the Build the current lane color constraint based on the HSV image space transform,Reduce the impact of the guardrail and the direction lane on the side lane determination;Finally,Based on the interframe similarity of continuous detection results,the starting determination and error detection determination container and leak detection compensation methods are designed,Complete the side lane determination.4.This paper builds an on-board experimental platform.Simulation experiment and real scene test were conducted on computer terminal and on-board platform.In conclusion,the lane detection algorithm has good detection accuracy and robustness and real-time operation requirements to effectively meet the relevant function of vehicle lane.In conclusion,the detection method can detect multiple lanes,and has higher fit and robustness than the existing algorithms.The research work done in this paper provides the theoretical foundation and practical experience for the research of the lane detection technology in the environmental perception module and the practical engineering application.