Detection System of Smoking Behavior Based on Face Analysis

Since the implementation of Tobacco Hazards Prevention Act, the prohibition against smoking in public places tends to be stricter in Taiwan. Its emphasis level can be inferred from the frequent law amendments in recent years. In order to effectively improve the prevention effect and bring down the costs of human surveillance, this paper proposes a smoking behavior detection system based on the human face analysis, which can quickly and accurately detect whether the person in the image is smoking or not. First, the system makes use of the existing facial detection scheme to locate the mouth, and then conducts white balance processing on the detected face image so as to reduce the misjudgment caused by the light. It then implements color segmentation and noise elimination in HSV color system in order to determine the location of the cigarette. The range between the location and the mouth is used to determine whether the person is smoking or not. This study hopes to combine with the indoor monitoring system to automatically detect the illegal behavior and send out alarms to notify the responsible units for reference.