Drowsiness Detection System For The Sentry Based On The Surveillance Video

The sentry is an important foundation for the CAPF(Chinese People’s ArmedPolice Force) to complete the duty work satisfactorily. If the sentry on duty nap not onlyaffects the image of the sentry but also may lead to a serious accident. Therefore，it is ameaningful thing to detect the drowsiness of the sentry on duty. By employing themotion detection and the detection of the head’s state，with the warning mechanism，thisthesis uses the technology of machine vision to analyze the surveillance videos of thesentry for judging whether the sentry is drowsy. The main works of research are asfollows:1. A scheme is proposed about the drowsiness detection system for the sentrybased on video surveillance in this thesis. With Considered the actual conditions， thehead of state is used as the feature of drowsiness detection，and we gives the descriptionof the system model and the introduction of the function of each module.2. A motion detection method is introduced in this thesis. The method bases on theentropy about the projection of the difference image, and overcomes the affect of thebad weather and the motion of the small target，it is more suitable for the applicationenvironment.3. A segmentation method is presented in this thesis. The method which bases onhuman physiological characteristics is mainly used for image segmentation of thesurveillance videos about the static standing sentry，it have a simple principle and a fastoperation.4. This thesis focused on the target recognition algorithm based on kernelregression， and used this algorithm to solve the problem of identification of the state ofhead. During the study，we realized the algorithm and analysed the advantages of thealgorithm and the applicability of the recognition of the head’s state by experiment.5. With based the platform of DirectShow，the thesis completed the capture ofvideo images， decoding， analysis， display and timely alarm.