Research Of Environment Perception Technology For Automatic Parking Based On Laser Scanner

This paper is committed to research the parking space detection of the automatic parking based on the experimental platform of intelligent vehicle research and the requirement of modern automobile technology. Laser scanner could perceives barriers such as pedestrian, wall and so on so as to analyze the available parking spaces. The main content is as follows:(1) Due to the measurement of laser scanners may contains noise affecting test results, median filtering and least square method is applied to process the laser data for filtering noise in this paper. Median filter could filter isolated noise greatly. Therefore, median filter is good for filtering white noise and outlier data in the laser data; Due to the obstacles of detection has a commendable linear characteristic, so this paper applies linear least square method to smooth the laser data after median filtering process. The laser data after processing will be more conducive to the follow-up work.(2) This paper applies two methods including the way based on grid partition and the way based on clustering division to classify the laser data. The method based on grid partition is relatively simple for processing data so that the speed for processing data is fast and at the same time the method could be greatly good at data classification. The paper applies k-means clustering algorithm that is the most basic clustering method. As k-means clustering applied to divide two-dimensional laser data is worse than dividing one-dimensional and three-dimensional data. The two-dimensional laser data is not ideal after partition. So this paper applies k-means clustering and threshold segmentation to divided laser data.(3) Endpoint and endpoint coordinates of each category could be analysed based on the laser data partition so that the available parking spaces could be found. Finally, the method for detecting parking has the very good feasibility through the experiment analysis.