Abstract
The current chapter presents a color texture-based method for object detection in images. A support vector machine (SVM) is used to classify each pixel in the image into object of interest and background based on localized color texture patterns. The main problem in this approach is high run-time complexity of SVMs. To alleviate this problem, two methods are proposed. Firstly, an artificial neural network (ANN) is adopted to make the problem linearly separable. Training an ANN on a given problem to achieve low training error and taking up to the last hidden layer replaces the kernel map in nonlinear SVMs, which is a major computational burden in SVMs. As such, the resulting color texture analyzer is embedded in the continuously adaptive mean shift algorithm (CAMShift), which then automatically identifies regions of interest in a coarse-to-fine manner. Consequently, the combination of CAMShift and SVMs produces robust and efficient object detection, as time-consuming color texture analyses of less relevant pixels are restricted, leaving only a small part of the input image to be analyzed. To demonstrate the validity of the proposed technique, a vehicle license plate (LP) localization system is developed and experiments conducted with a variety of images.
Preview
Unable to display preview. Download preview PDF.
Author information
Authors and Affiliations
Editor information
Rights and permissions
About this chapter
Cite this chapter
Kim, K., Jung, K., Kim, H. Fast Color Texture-Based Object Detection in Images: Application to License Plate Localization. In: Wang, L. (eds) Support Vector Machines: Theory and Applications. Studies in Fuzziness and Soft Computing, vol 177. Springer, Berlin, Heidelberg. https://doi.org/10.1007/10984697_14
Download citation
DOI: https://doi.org/10.1007/10984697_14
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-24388-5
Online ISBN: 978-3-540-32384-6
eBook Packages: EngineeringEngineering (R0)