Autonomous Parking System Research And Design Based On Bird’s Eye View

As an effective and quick traffic tools, automation take greatlyconvenient to people for their daily life and work. However, it is also bringseriously trouble for our city. With increasing traffic problems such astraffic congestion and parking problems happens in anytime. As aneffective solution for traffic problems, drive assistance technology istaking more and more attention by many countries at moment to solve theparking problems for safety and efficiency. Panoramic and autonomousparking system which based on vision method is becoming an importantbranch area in Intelligent Transportation System. This paper mainly studiesthe low cost of panoramic parking assistant system and automatic parkingsystem. This paper makes the following researches:A parking assistance system (PAS) has been proposed in the paper.According to the requirements of parking task, PAS is divided intopanoramic parking assistant module and automatic parking module. Firstly,in the panoramic parking assistant module, an important problem is how toeliminate blindness and visual dead angle during parking environments. So,a low cost parking assistance system with the bird’s eye view which wasmosaicked based on four image of cameras amounted around vehicle wasdesigned. What’s more, in order to solve the problems of image distortionand four images mosacing, the derivation of camera model is introduceddetailly. Then a convenient calibration method based in grid pattern is usedto join the vehicle and four cameras coordinate together. Lastly, withinverse perspective transformation, an image mosaicing method based on the feature points was proposed. Experiment shows that the proposedmethod was effectively and reliability.Secondly, in the autonomous parking module, the core problem ishow to detect and recognize parking space more accuracy and reliability.Therefore, the paper proposed a new method based on the bright featuresin Radon space to recognize parking space. Then in order to get the freeparking space for autonomous parking, the method based on interestwindow is proposed to extraction free parking space. The experimentshows that the methods have good accuracy and high robustness in parkingspace detection with the influence of noise, clutter and variation inillumination and other cars.Finally, to solve problems of path planning and path tracking controlduring parking process, the paper analyze vehicle’s kinematic model at thelow speed in parking conditions. Then, considering the original positionbetween the vehicle and detected free parking space, the improvedmethods based on double arcs for vertical and parallel parking pathplanning were proposed. Simulation and experiments show that theplanned path is easy and safe. Combing with the PID control method andposition estimation method, the path tracking is effectively for autoparking.