Abstract
One of the most noteworthy problems associated with conventional pattern recognition methods is that it is not easy to extract feature vectors from images which are translation, rotation, and scale change invariant in outdoor noisy environments. This research describes the development of an invariant traffic sign recognition system capable of tolerating the above variations. The signs are restricted to three types of warning signs and are all of red color. The procedure of the developed invariant pattern recognition system consists of three phases: (a) color image segmentation, (b) segmented image refinement, and (c) object recognition. In the color image segmentation phase, the proposed segmentation algorithm uses a partition and merge concept and the (u, v, h)-color coordinate system. A smoothing filter and morphological filters are used to refine the segmented binary image. The proposed smoothing filter is a nonlinear processing technique which is useful for filling little holes and eliminating small spots in an image. A unique combination of a centered polar-exponential grid, a Fourier transform, and a back-propagation network is used in the object recognition phase. The developed recognition system is insensitive to brightness changes as well as invariant to translation, rotation, scale change, and noise. The architecture of this system is based upon neural network supervised learning after geometrical transformations have been applied. The proposed system is tested on a large number of signs with different positions, rotations, scale changes, and backgrounds. The performance of this system is compared with other invariant recognition approaches in terms of the percentage of correct decisions in outdoor noisy environments.
Kang, Dae-Seong (1994). Invariant pattern recognition system based on sequential color processing and geometrical transformation. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -1551734.