Evolutionary segmentation of road traffic scenes

Segmenting a car region is an essential stage in the automatic car identification. It is difficult to segment the car region correctly, because road traffic scenes are usually degraded and processing the images is computationally intensive. In this paper, we propose a method of extracting a car region based on color image processing. To segment the color image, we use a distributed genetic algorithm and a Hue-Saturation-Intensity (HSI) color space as a measure of distance. The method offers robustness in dealing with deformation of road scenes and inherent parallelism to improve processing time. A test with road scenes shows an extraction rate of 92.5%. This result suggests that the proposed method works well with real-world situations, and is pertinent to be put into practical use.