Title:	HOUGH TRANSFORMED CONTOUR LINE SEGMENTS FOR OBJECT DETECTION
Author(s):	Weize Zhang, Jean-Marc Vézien, Ruofeng Tong, Patrick Bourdot, Jinxiang Dong
ISBN:	978-972-8939-22-9
Editors:	Yingcai Xiao, Tomaz Amon and Piet Kommers
Year:	2010
Edition:	Single
Keywords:	Hough transform, implicit shape model, object centroid, distance measurement, shape matching.
Type:	Full Paper
First Page:	235
Last Page:	242
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	This paper presents an Implicit-Shape-Model-like framework for object detection, however with only shape cues. A new shape representation called Hough Transformed Line Segment (HTLS) is proposed, which encodes the geometric relation between the line segment and the object centroid implicitly, while handles rotation, translation and scaling easily. In the HTLS space, a simple weighted distance measurement is introduced to calculate the distance between two groups of the same size of connected HTLSes. The weights are adjusted to make the distance distinctive while tolerant to partial matching and other noises introduced during the approximation of the contour by line segments. With a simple shape matching strategy in this framework, the experimental results on two categories, motorbikes (man-made object) and cows (articulated animal), show that shape cues have a great potential for object detection task.
	Go Back