Smart entrance monitoring system using clothing color analysis in multi-camera environment

This paper introduces an automatic method for detecting intruders based on their clothing features extracted from CCTV camera images. The proposed method is used as an auxiliary entrance authentication for a smart entrance monitoring system in a multi-camera environment. Whenever an individual is authenticated at the entrance of a building, typically using an entrance card or finger print, their clothing feature is extracted and added to a database. When an arbitrary individual is captured by different cameras installed inside the building, their clothing feature vector is automatically extracted and compared with the reference feature vectors in the database. If a matching clothing feature vector is not found in the database, the individual is identified as an intruder. The color features of the clothing areas from different camera images are also corrected according to the illumination conditions. Experiments demonstrate the success rate for detecting intruders and the effectiveness of the color correction.