Vehicle Detection in Different Environments

In this paper, we presented a vehicle detection method based on RGB color space components analysis. The proposed approach is mainly focused on designing the system which is applicable in the case of different weather conditions (rainy, snowy, misty etc), different times during the day and night (daylight, night, noon, afternoon), heavy traffics, the existence of the shadows and different road conditions. Most of the vehicle detection methods utilized background model generation. Since even slight changing in the brightness could decrease the detection quality, in these kinds of methods the background image needs to continuously be updated. In this paper, we presented the method in which the vehicle detection process is performed without any need to generate and update the background model. In the presented approach, we utilized the histogram normalization in order to alleviate the problems caused by brightness change in the case of different weather conditions. We also extracted moving objects using optical flow. Finally, we utilized the HOG descriptor and SVM classifier in order to detect vehicle objects. The performance of the proposed method is tested using VDTD dataset and the results illustrate that the proposed method provides acceptable results specially in heavy traffics and different weather conditions.