Abstract: In order to achieve parking position detection that plays a great important role in the automatic parking system and its accuracy directly affects the accuracy of path planning, this paper has put forward a method that can fast and accurate detect the parking position. Firstly, laser rangefinder was used for gathering environmental information around the vehicle and then median filtering and least square method were used for original laser data filtering, which can reduce noise interference. Finally, grid-based clustering method and K-means clustering method were used for detecting the feasible parking spaces. This paper simulated the two methods on the VC++ platform and made a comparative analysis of them. Experimental results show that this method has high robustness and accuracy.