|
Title:
|
AN IMPROVED RED COLOR SEGMENTATION ALGORITHM AS PART OF AN AUTOMATIC SPEED LIMIT SIGN DETECTION |
|
Author(s):
|
Hanene Rouabeh, Chokri Abdelmoula, Mohamed Masmoudi |
|
ISBN:
|
978-989-8533-56-2 |
|
Editors:
|
Hans Weghorn |
|
Year:
|
2016 |
|
Edition:
|
Single |
|
Keywords:
|
Image processing; Color segmentation; Color spaces; Thresholds |
|
Type:
|
Full Paper |
|
First Page:
|
131 |
|
Last Page:
|
138 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Color segmentation is one of the most important tasks in the image processing field. It is used for many vision based applications. In the presented work, a red color segmentation approach was presented in purpose to be used as a mandatory task of a complete vision based intelligent speed limit sign recognition system. The main challenging problem in realizing such system, resides in the achievement of a compromise between real-time processing and accuracy constraints. For this purpose, a simple and accurate segmentation technique was proposed. It presents a modified and improved thresholding based method that uses the Red, Green and Blue color space components. This technique is divided into three steps; the first one is classification which is done based on the Hue, Saturation and Lightness color space components thresholding. The aim is to classify images according to different illumination conditions. The second one consists in color adjustment and correction. And the third one consists in thresholding. This approach was introduced to solve problems caused firstly by the use of same thresholding values for different images, and secondly by the sensibility of colors to lightening conditions variation. The comparison with existent methods has shown robustness and accuracy mainly for images of different day times and for faded and blurred signs. In addition to that low computation time is achieved. |
|
|
|
|
|
|
