Research On Lane Line Recognition And Path Planning Of Intelligent Vehicles

In the context of the integration of traditional machinery and intelligent information technology in Industry 4.0,the automotive industry has ushered in rapid development and new changes.Due to the increase in car ownership causing traffic pressure and driving difficulty,intelligent driving has become a key development direction of automotive technology.With the continuous improvement of intelligence and driving safety requirements,accurately perceiving the external environment and planning paths reasonably and efficiently is one of the hotspots in intelligent vehicle research.Scholars at home and abroad have made many achievements in lane detection and path planning technology,but there are still many problems.During the driving process of intelligent vehicles,they will face various unexpected situations,requiring the vehicle to have the planning ability in known maps and the ability to collect lane and obstacle information in unknown environments using sensors such as cameras to complete real-time online planning.The research work conducted in this article on intelligent vehicle lane recognition and path planning is as follows:(1)Research on lane recognition and detection.Preprocess the vehicle driving environment image captured by the camera,such as image denoising,image binarization,image filtering,image transformation,etc.In the grayed image,the region of interest containing the complete road information is determined,and the aerial view is obtained through reverse perspective transformation to improve the running speed of subsequent image processing.Two image processing methods,edge segmentation and threshold segmentation,are proposed for structured and unstructured roads respectively.Then,the feature points of lane lines are extracted by sliding window method,and clear and complete lane lines are extracted by Hough line detection and Bézier curve fitting.(2)Plan safe and feasible paths in a global static environment.Based on the principle of the Fast Random Tree Algorithm(RRT *),an improved RRT * algorithm is proposed.By introducing a bidirectional random tree search strategy,the convergence time of the algorithm is shortened and the computational efficiency is improved.Adopting redundant point clipping and B-spline curve fitting methods for the preliminarily planned path,the algorithm generates a relatively smooth path.Carry out simulation verification of narrow channels and multiple obstacles in the MATLAB software environment.(3)In a dynamic environment with unknown information,a heuristic based rolling window method is proposed,which divides the one-time global planning into local planning with sub objective points established within multiple windows to improve information prediction ability.Integrating rolling window method,artificial potential field method,and improved RRT * algorithm,guiding RRT * algorithm nodes to sample towards the target point direction,while the repulsive force field of obstacles keeps vehicles away from obstacles,improving obstacle avoidance ability.Finally,simulation was conducted in a two-dimensional environment to verify the effectiveness of the fusion algorithm based on rolling optimization.(4)Build an intelligent vehicle experimental platform based on the ROS system.Establish a WORLD road environment model in the system and use SLAM technology to create maps.Through the ROS system experiment,analyze and compare the advantages and disadvantages of the improved algorithm in global planning and local planning.