Cone Transform for Size-Tolerant Object Recognition
Chia Shyan Lee, J. Lim, Seongyoun Woo · 2018
In this paper, we propose a cone transform method, which is robust against size variations in object recognition. A main problem in computer vision is that objects may show large size variations. This makes it difficult to develop robust object recognition algorithms since most recognition algorithms assume certain target size ranges. The proposed method represents an input image by two angles. The first angle is measured from a center point counter-clockwise and the other angle is measured from the radius. The proposed method shows better tolerance against size variations. Experiment results show that the proposed method provided better classification performance when objects were reduced in size.