Research On Vision-based Road Detection Technology In Complex Environments

Road detection technology is one of the basic key technologies for autonomous driving environment perception and an important reference for decision-making and planning of autonomous vehicles.Compared with developed countries,road conditions of our country have the characteristics of diverse road environment configurations and a high degree of mixing of people and vehicles,which brings great challenges to the development of road detection technology for complex conditions.The main research work and conclusions of this paper are as follows:(1)Aiming at the problems of low efficiency and poor robustness of structured road lane line detection in complex environments,a complex environment structured road detection method based on multi-constraint PPHT(Progressive Probabilistic Hough Transform)algorithm was proposed.The contrast-limited adaptive histogram equalization was used to enhance the road image,the progressive probabilistic Hough transform algorithm was improved,and the improved PPHT algorithm based on multiple constraints was proposed to extract the left and right lane lines.(2)Aiming at the problem of low accuracy in structured road images such as occlusion,defacement,shadows,etc.,the multi-constrained PPHT algorithm was improved,the lane line tracking algorithm based on unscented Kalman filter was designed,and the lane line observation model was established in real time.By observing the lane line model parameters in the driving state of the vehicle,an effective estimate in the driving state of the vehicle was obtained;By comparing and analyzing with the existing classic algorithm of structured road lane line detection,the improved algorithm was verified to have higher accuracy,real-time performance and resistance Intrusive.(3)Aiming at the problem of complex environment variables and many interference factors in unstructured road images,an unstructured road boundary detection method based on region growth and image block was designed.Unstructured road preprocessing methods were researched,HSV space conversion and median filter was used to enhance the image;Otsu threshold segmentation algorithm was used to binarize the image;Morphological methods were used to remove discrete noise points and smooth edges;The growth algorithm based on the region of the maximum area was used to extract the road area,and the image block algorithm was combined to extract the road edge points and fit the road boundary.Compared with the existing classic algorithm of unstructured road boundary detection,unstructured road detection method proposed in this paper was verified to be feasible and advanced.