即時影像辨識應用於智慧型機器人自走車之導航系統

Real-time Imaging Recognition used in the Intelligent Mobile Robot Navigation System

陳坤雄(Kuen-Shiung Chen)

指導教授 ： 林春宏

國立臺中科技大學/資訊與流通學院/資訊工程系碩士班/碩士(2012年)

https://doi.org/10.6826/NUTC.2012.00113

摘要

本論文主要目的在應用即時影像處理技術以及智慧型自走車本身感測器，以實現自走車之即時導航的功能。在本研究中所使用的智慧型自走車具備影像傳輸的CCD攝影機，可作即時影像信號處理、顏色辨識、移動偵測及遠端環境影像監測、智慧型感測器應用及自主性可程式化開發，讓使用者快速導入希望完成的應用實驗工具與開發工程。
本論文透過電腦視覺與影像處理的技術，規畫設計了有關自走車應用於「物件追蹤」、「方向標誌辨識」與「路邊停車」的三個重要導航系統功能。以期達成讓自走車去模擬現實生活中汽車導航的環境，其中包括自動辨識方向標誌、依照方向標誌指示轉向行進、搜尋停車格及自動路邊停車等之自動導航功能。此外，為了讓使用者能透過簡易的操作介面控制自走車之導航功能，我們特別設計了PC介面與WEB遠端操控介面，以提供使用者能自行選擇操作各項功能，以體現自走車導航系統之功能多元性與操作便利性。
在「物件追蹤」的過程中，自走車上的無線影像傳輸攝影機能將即時攝影得的影像資訊，經由無線傳輸回伺服機端做進一步的影像辨識處理，再經過影像處理與相關演算法計算之後，以無線傳輸機制回傳演算結果資料，便能得知追蹤物體與自走車之間的相對關係，進而達成追蹤目標物體之導航功能。
在「方向標誌辨識」偵測部分，是依據自走車利用顏色偵測技術去搜尋並偵測到需要辨識的圖樣區塊，並且把圖樣從原本複雜的背景影像中抽離出來，並同時濾除掉在背景抽離時所殘留的雜訊。接著在圖樣擷取時，則是利用三角形區塊面積比例判定方法，來判別可以用來判斷特徵的方向標誌符號及其指向的方向性。在取得可以用來判斷特徵的方向標誌圖樣之後，我們即能讓自走車執行方向標誌之判定與方向導航之功能。
在進行「路邊停車」的實作時，我們是依據自走車上的無線影像傳輸CCD攝影機所拍攝得到的影像進行分析，接著依照本論文所設計的演算法步驟，以執行路邊停車之控制流程，依序進行以下之導航功能判別：停車格之搜尋、判斷停車格之外圍停車線、確定停車格內無障礙物、進行路邊停車。
最後，我們透過系統分析的方法，規劃設計了PC介面與WEB遠端操控介面，並經由實際操作方式，以驗證介面的可行性與操作便利性，以達成自走車導航系統操作之功能多元性與操作便利性。
經由所完成之自走車導航系統、辨識與功能應用的各項實驗結果展示，證明我們所提出的自走車導航系統，有助於推廣至真實之停車自動導航上的應用。

關鍵字

影像處理 ； 圖型辨識 ； 視覺感測器

並列摘要

The purpose of this thesis in the application of real-time image processing technology and intelligent sensors to achieve real-time navigation capabilities of intelligent mobile robot. With wireless image transfer CCD camera used in this study, an intelligent mobile robot, can be used for real-time video signal processing, color recognition, motion detection and remote environment image monitoring, smart sensor applications and autonomous programmable development, allowing users to quickly import the hope that the completion of the application of experimental tools development engineering.
In this thesis, through computer vision and image processing technology, to planning design on the vehicle use the object tracking, direction sign recognition and on-street parking, three important navigation system functions as listed above. In order to reach a self-propelled vehicle car navigation environment to simulate real life, including automatic identification of direction signs, turned to the road, follow the direction signs to indicate automatic navigation function to search for parking space and automatic on-street parking. In addition, in order to let the user through a simple interface to control self-propelled car navigation features, specially designed PC interface and WEB remote control interface to provide users to choose the operating various functions to reflect the intelligent mobile robot navigation system versatility and operating convenience.
In"Object tracking" process, self-propelled image information of the wireless video transmission on the car camera can of instant photography, via wireless transmission back to the server for further image recognition processing, and then after the image processing algorithms to calculate theconvenience can be learned to track objects with intelligent mobile robots between the relative and thus reached the navigation features to track the target object.
In the part of detection direction of Mark Recognition is the basis for the use of intelligent mobile robot color detection technology to search for and detect the need to identify the pattern blocks, and pulled out of the drawings from the original background image, and at the same time filterget rid of the residual noise in the background is detached. Then, in the capture of the drawings, it is a triangular block area ratio to determine the method to distinguish the direction signs can be used to determine the characteristics of symbols and their point of directional Obtained can be used to determine the characteristics of directional signage, that is, allow intelligent mobile robot is performing the direction signs to determine the function of navigation and orientation.
During the implementation of the on-street parking, based on images taken since the car-go wireless image transfer CCD camera, then the algorithm steps in accordance with the paper, to perform the on-street parkingcontrol flow, sequentially following navigation feature discrimination: parking grid search to determine the perimeter parking lines of parking space, sure no obstructions within the parking space, on-street parking.
Finally, through the methods of systems analysis, planning and design of the PC interface and WEB remote control interface and through the practical way to verify the feasibility of the interface and operating convenience has reached a intelligent mobile robot navigation system operated by the versatility and convenience of operations.
Through the completion of car navigation systems, Identification and functional applications through the completion of the experimental results show, to prove that our proposed intelligent mobile robot navigation system, helps to automatically navigate to the real parking on the application.

並列關鍵字

Image processing ； Pattern recognition ； vision sensors

參考文獻

[26] 陳冠宇、張永鵬、傅培耕，“具物體追蹤功能之立體視覺導引自走車”，中原大學機械工程學系，2004.

[1] Leese, J. A., Novak, C. S., and Taylor, V. R. “The Determination of Cloud Motion Patterns from Geosynchronous Satellite Image Data,” Pattern Recognition, Clustering, Statistics, Grammars,Learning, Vol.2,pp. 272-292,1970.

[2] Horn, K. P. and B. G. Schunck,“Determining optical flow,” Artificial Intelligence, Vol.17, pp. 185-203, 1981.

[4] Tan, Y. P., Kulkarni S. R., and Ramadge P. J., “A New Method for Camera Motion Paramater Estimation,”IEEE.International Conference onImageProcessing, vol. 1, pp. 406-409, 1995.

[5] 吳良杰，「視覺感測器自動追蹤地面移動目標」，碩士論文，元智大學機械學系，1996.