A technique that improves precision in classification results using information extracted from video features is introduced. It combines fuzzy rule-based classification with the detection of changing regions in outdoor scenes. The approach is invariant to extreme illumination changes.
