A noise-tolerant traffic sign recognition method based on color images

Due to the character of the original source materials and the nature of batch digitization, quality control issues may be present in this document. Please report any quality issues you encounter to digital@library.tamu.edu, referencing the URI of the item.Includes bibliographical references.Issued also on microfiche from Lange Micrographics.Devising a pattern recognition methodology for traffic sign images captured in noisy outdoor environments poses a challenging problem. Such a method has to cope with occlusion noise as when part of the region of interest (in this case, the traffic sign) is blocked by an interfering object, with centroid noise as when an interfering object is present in the vicinity of the region of interest, etc. In this research, a noise-tolerant method is designed such that the recognition rate will not be adversely affected by noisy conditions that can be expected in realistic environments. The performance of various color coordinate systems (RGB, ISH, XYZ, YIQ, uvY) is evaluated via several separability measures in order to obtain the color coordinate system giving the best separation between traffic signs and other objects in the scene. Color segmentation is performed by using a self-organizing neural network, which generates a segmented binary image of the traffic sign. A signature of the traffic sign is extracted that is invariant to changes in scale, location, and orientation based on the log polar exponential grid and Fourier transformations. Finally, a backpropagation neural network trained by using signatures of various traffic sign images is used to classify the traffic signs