This paper introduces a method for robust back ground subtraction suitable for continuous 24-hour video surveil lance in traffic environments. The technique aims to function in real-time, withstand varying weather conditions, and maintain the detection of foreground objects for extended periods. The primary application is a monitoring system for highway safety, capable of detecting events like wrong-way driving, accidents, or pedestrians on the road. The method is based on the Pixel-Based Adaptive Segmenter (PBAS), which models the background using a history of recent pixel values and dynamically adjusts decision thresholds and learning parameters. The algorithm is fine-tuned to ensure robust performance under diverse conditions