部分重疊工件的辨識

Recognition of Partially Occluded Parts with Hough-like Clustering

龐君豫(Chun-Yu Pang)

指導教授 ： 蔡篤銘

元智大學/工程學院/工業工程與管理學系/碩士(1992年)

摘要

無資料

關鍵字

機器視覺 ； 形狀描述 ； 形狀比對 ； 霍氏轉換 ； 頂角偵測 ； 關聯圖

並列摘要

h本研究乃是利用機器視覺技術來辨識部分重疊的工件和其位置，並且不
需要依賴工件的特徵點來做辨識。在本研究中是以平面工件為辨識對象，
而對於工件的外觀形狀只考慮直線和圓弧，因為許多工業界中的工件其厚
度較薄且其外觀形狀是由直線和圓弧所組成，所以仍然能有廣泛的應用。
本研究主要分為兩大部分─形狀描述和形狀比對。在形狀描述部分是先對
工件邊緣的形狀以直線和圓弧描述，在此利用兩種方法：一為霍氏轉換
法(Hough transformation)；另一為頂角(vertex)偵測法，來做直線和圓
弧的偵測，此兩種方法在計算時間及準確性方面各有不同，然後再以描述
工件形狀的直線和圓弧所構成的關聯圖(relational graph)來進行比對，
在本研究中的比對方法，稱之為類霍氏群集法(Hough-like clustering)
，乃利用霍氏轉換法中累積方格的觀念，來偵測標準物和待測物之間的旋
轉角度和位移量，再經過假設檢定的檢查，以增加其辨識的可靠度。所以
本研究方法並不受重疊部分是否為工件之重要特徵之限制，並且如果工件
形狀有部分描述的不完全或不準確仍然能辨識出來，因此亦可增加其使用
範圍。本研究之結果能應用於工件在輸送帶上自由流動所造成的部分重疊
情況，以方便機器人對工件做加工處理等。如此便能增加工廠自動化的彈
性，而且因為不需要以人工的方法將工件安排在拖板或固定的位置，也不
需要依賴特定的物料供應系統，所以將可節省人力和週邊設備的投資。
The problem of recognizing partially occluded parts is of
considerable interest and challenge in the filed of industrial
automation. While it is possible to employ carriers or pallets
to separate or prearrange the parts for easy recognition, a
vision system which can recognize the parts even though they
may be partially occluded and in random position is much
flexible. An occluded object can be due to the occurrence of
touching and overlapping objects,or the occurrence of defects
and burrs of objects. In this research, we consider the problem
of identifying and locating partially occluded industrial parts
lying on a flat surface; and the contours of the two-
dimensional parts parts can be represented by linear and
circular segments. Most of man-made objects have such
geometrical features. This research consists of two phases :
shape represention and shape matching. Two methods that
identify the line and circular primitives of objects are
developed in this research. One is based on the Hough transform
that incorporate the information of the gradient directions of
the boundary points. The other is to find the vertices of the
object contours and then use the least-squares method to
identify a linear or circular segment between every two
adjacent vertices. In matching phase, the structure of the
shape primitives is represented by an attributed relational
graph where a node represents a line or circular segment, and
an arc between any two nodes represents the the geometrical
relation of two primitives. A matched pair of arcs between the
model-object graph and scene-object graph has the same
geometrical relation and node properties. A Hough-like
clustering procedure is developed that accumulates the number
of arcs which give the same rotation angle and displacement for
every matched pair of arcs. The developed procedure is
insensitive to noise and does not rely on the presence of
special features of an object such as interior holes or shape
corners.

並列關鍵字

Machine vision ； Shape representation ； Shape matching ； Hough