以對應點為基礎應用立體視覺之自動停車系統

Development of an Automatic Parking System Using Stereo Vision Based on Corresponding Points

廖育聖(Yu-Sheng Liao)

指導教授 ： 蕭耀榮

國立臺北科技大學/機電學院/車輛工程系所/碩士(2008年)

https://doi.org/10.6841/NTUT.2008.00375

摘要

本論文中，我們發展出一套可裝置在自動導航車上的立體視覺系統，藉此系統收集停車環境中的資訊，將車輛自動停置於停車格之中，目的在於減低駕駛者對於停置車輛時的負擔，並減少停車所需花費的時間。
本研究分為三個主要部分，首先是影像辨識方面，使用雙CCD(Charge Coupled Device，電荷耦合元件)相機擷取含有停車格之左右成對影像，經由中值濾波與sobel運算子邊緣化後，利用霍夫轉換最後切割出含停車格線的影像。第二部份是立體視覺感測系統，以Zhang的相機校正方法求出左右相機的內外部參數，接著利用共軛幾何關係推導出基礎矩陣，依等極線的特性將原本特徵對應點的二維搜尋化簡為一維，再利用視差關係來量測出自動導航車與停車格之間的距離(深度)，進而計算出停車格的方位與大小。第三部份則是停車控制系統，藉由視覺系統與影像處理後的停車格幾何關係，判別停車模式為路邊停車或倒車入庫模式，再以車輛與停車格的相對位置，角度以及停車模式為參考依據，計算停車路徑，最後以模糊邏輯控制器控制車輛依循目標路徑將車輛正確的停置於停車格之中。

關鍵字

自動停車 ； 立體視覺 ； 相機校正 ； 基礎矩陣

並列摘要

In this thesis, a set of stereo vision system was developed for Autonomous Mobile Robot(AMR). This system was used to collect environmental information about the parking space, and then parked the AMR into the desired parking space automatically. This designed system would decrease the burden for vehicle drivers, and reduced the process time in parking.
This research was divided into three main parts. The first part is image recognition. A set of double CCD(Charge Coupled Device) camer was used to catch the images of parking space extract. The images were processed through medium filter, sobel edge diction, and Hough transform to the lines of parking space.
The second part was stereo vision sensing system. This system found the camera’s intrinsic and extrinsic parameters, using Zhang’s method. Epipolar geometry relation was used to reason fundamental matrix. Then the 2D images could be transferred into 1D images according to epipolarline’s characteristic of line. Finally the depths between AMR and parking space as well as dimensions of parking space could then be calculated.
The third part is parking control system. The information of the geometry of parking space was used to decide the parking mode: parallel parking or garage parking . A parking path was then calculated according to the relative positions between AMR and parking space. Finally, a fuzzy logic controller was used to guide the AMR into the destination parking space following the designed parking path.

並列關鍵字

automatic parking ； stereo vision ； camera calibration ； fundamental matrix

參考文獻

[4] O. Faugeras and Q. T. Luong, The Geometry of Multiple Images, MIT Press, Cambridge, MA, 2001.

[6] R. C. Gonzalez, R. E. Woods, Digital Image Processing, 2nd ed., Prentice-Hall, New Jersey, 2002.

[7] C. G. Choi, D. S. Kim, H. G. Jung, and P. J. Yoon, “Stereo Vision Based Parking Assist System,” SAE technical paper series, 2006-01-0571, 2006.

[8] M. Usami, K. Ohue, H. Ikai, M. Ohta, and T. Tamaoki, “Stereo Vision System for Advanced Vehicle Safety System,” SAE technical paper series, 2007-01-0405, 2007.

[9] S. Hiramatsu, A. Hibi, Y. Tanake, T. Kakinami, Y. Iwata, and M. Nakamura, “Rearview Camera Based Parking Assist System with Voice Guidance,” SAE technical paper series, 2002-01-0759, 2002.