都市道路圖自動輸出入之程序設計－應用微電腦影像處理技術

Abstract

由於影像處理的技術能使道路圖形的處理邁入自動化的領域，因此本論文嘗試以此項技術中多項方法的組合，設計一個適用於簡易都市道路圖－街廊和道路中心線網路圖的自動輸出入程序，並在NEC-98 XA PC上設計應用程式。經測試結果，除特徵點辨識外，其餘方法顯示結果尚可。 本論文的重點。在於將簡易都市道路圖的掃描資料經過一連串的程序之後，轉化成向量式資料加以儲存，並作為日後處理較複雜的道路圖形以及與道路管理資料庫結合的基礎；其中的各項主要方法簡述如下： (一)細線化：以直覺的簡易方法，把輸入之人工手繪粗線圖，利用水平與垂直方向的細線化，抽取其骨幹而留下一點寬度的細線。 (二)設定圖形單元：利用(一)之結果，以“＋”字形狀的角落符號之版比圖找到角落點，利用角落點算出轉換矩陣，把圖形轉正，再把角落點範圍外的各點去除。 (三)尋找特徵點：在(二)取得的範圍內，先以5×5的臨近範圍補點，再以3×3的臨近範圍找到各組封閉線段和各組線的端點和叉點，最後以5×5所建立的臨近斜率指標抽出折點，同時以包含叉點的組別為網路線。 (四)儲存圖形資料檔：把(三)之結果儲存在都市道路圖形資料的結構中，作為道路的描述與查詢之用。

For the reason of entering the automatic processing field of road maps by the use of image processing techniques, this thesis adopts combination of many methods included in these techniques to design a satisfactory procedure of automatic input/output for simple city road maps--street blocks & networks of road center lines & program on NEC-98XA personal compuer. Through the tests on computer we have fare results besides feature points detecting. The part underlined of this thesis is to transform raster datas of simple road maps into vector datas of those by a serious of steps & to save that vector datas; furthermore, these steps can be used as a beginning to develop more powerfully to hand over complicated road maps & have a good Link with road management data base. The essential method is described briefly below: (1)Thinning: by a intuitive simple method, to extract the skeleton from man-made thick lines input horizontally & perpendicularly, leaving thin lines of one pixel. (2)Setting map units: when (1) is complete, proceedig to extract cornor points by cornor sign---"+" template matching, then, using coordinates of cornor points as parameters to calculate the transformation matrix & turn maps right & clip the maps into map units. (3)Searching feature points: in the areas of map units; first, to compensate points on 5x5 neighborhood if there is no points besides current point on 3x3 neighborhood; secondary, to make points clustered as closed lines & search extreme points & crossing or branch points by 3x3 neighborhood; Last, to found turning points by slope index on 5x5 neighborhood & classify closed lined including crossing or branch points as network lines. (4)Generating graphic data files: to save the result of (3) into data structure of city map for the use of road map description & retrieval.