SEARCH
MENU

History of Robotics Research and Development of Japan1976SensingResearch on Desk Scene Analysis

Yoshiaki ShiraiElectrotechnical Laboratory (ETL)
A method was developed which recogninizes various objects in a gray image of a complex scene. The method was an extention of the heteroarchical method[1] for block world recognition. The system extracts obvious edges, obtains sequences of the edges by edge tracking, approximates each edge sequence with lines and ellipses, and then tries to recognize objects. Next, using the previous recognition result, the system extracts less obvious features to recognizes other objects. By repeating this process, objects are recognized more accurately and efficiently(FIg.1 shows an example). This scheme was widely accepted for recognition of complex scenes[2,3,4] which were previously regarded as difficult .
図1(a)机上シーン
Fig. 1 (a) Desk scene
図1(b)最初に得られたエッジ列
Fig. 1 (b) Initially obtained edge sequences
図1(c)直線と楕円で近似
Fig. 1 (c) Approximation by lines and ellipses
図1(d)認識結果(0:スタンド,1:本立,2:電話機)
Fig. 1 (d) Recognition result (0: lamp, 1: book stand, 2: telephone)
図1(e)4回の繰り返しで得られたエッジ列
Fig. 1 (e) Edge sequences obtained after 4 repetitions
図1(f)認識結果(カップ,パイプ,鉛筆類などが追加)
Fig. 1 (f) Recognition result (cup, pipe, pencil, etc. are added)

Movies


Correspondence papers


Yoshiaki Shirai:A Method for Recognition of Complex Objects from Light Intensity Data

Journal of the IPS,Vol. 17, No. 7, pp. 611-617, 1976 (in Japanese).

Related papers


[1]Y. Shirai: A Context Sensitive Line Finder for Recognition of Polyhedra, Artificial Intelligence, Vol.4, No.2, pp.95-119 1973

[2]Y. Shirai: A Step toward Context Sensitive Recognition of Irregular Objects, Computer Graphics and Image Processing, Vol.2, No.3/4, pp.298-307 1973

[3]Y. Shirai: Edge Finding, Segmentation of Edges and Recognition of Complex Objects, 4th Int. Joint Conf. on Artificial Intelligence, pp.674-681 1975

[4]Y. Shirai: Recognition of Real-World Objects Using Edge Cue, Academic Press, pp.353-362 1978

Related Article

記事はありません。