A vision-based lighting conditions classification method for video surveillance

To enhance the robustness of the vehicle detection system, an effective algorithm to identify the lighting conditions (daylight, night, lowlight (dawn, dusk)) based on histogram analysis is presented in this paper. The algorithm consists of two procedures: extracting and updating background image, and generating a lighting conditions classifier based on background image analysis. The algorithm is based on naive Bayes algorithm, and it is classified by the Bhattacharyya distance of the histogram of each color component(R-component, G-component, and B-component) in RGB space. The experiment has shown that the lighting conditions classification method proposed improves the performance in lighting classification.