Title:	MULTI-VIEW OBJECT DETECTION USING REGION-BASED RANDOM FOREST CLASSIFIERS
Author(s):	Ji-Hun Jung, Byoung-Chul Ko, Jae-Yeal Nam, Young-Do Joo
ISBN:	978-972-8939-89-2
Editors:	Yingcai Xiao
Year:	2013
Edition:	Single
Keywords:	Object detection, region-based classifier, random forest, PASCAL, multi-view
Type:	Short Paper
First Page:	129
Last Page:	132
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	This paper presents a robust method for detection and classification of objects that is invariant to intrinsic and extrinsic variations in a complex background. To improve the object detection performance, we first select semantic regions and train their classifiers by considering the specific characteristics of regions. Each region-based classifier is boosted using a random forest classifier that is an ensemble of decision trees. An integration of random forest classifiers of single views is used to detect the most likely object position in the image. Then, because each view overlaps with neighbouring views, a weighted sum of the probabilities of three neighbouring views is used as the final score to determine the location and viewing angle of the object. The proposed algorithm is successfully applied to various PASCAL images, and its detection performance is better than the performances of other methods.
	Go Back