A Danger Awareness System For Kids Based On Sensor Network And Computer Vision

Context Awareness as one of the key topics of Ubiquitous Computing, through the data processing on the data from ubiquitous network, detects the user's context, then provides appropriate services according to the detected context. When applied on detecting the danger of users, it is called danger aware.This paper discusses how to detect kids'danger in the indoor environment. Currently, most of the researches about the danger aware in the indoor environment used the sensor network to detect danger, and made the awareness more intelligent and accurate. Since there are two many dangerous object for kids, it brings great burden to developers/users to define a huge amount of contexts in advance. Furthermore, sensor network has the shortcomings for detecting the detail motions, so the results of the danger aware from the sensor network are not accurate enough.This research uses a ontology based reasoning mechanism proposed by Computer Network Lab of University of Aizu. This method can somehow reduce the number of predefined context, improve the accuracy of danger aware. To cover the shortage of detection of detail motions by the sensor network, the motion capture is combined with the sensor network in this research. A model for capturing the motion whether the kids are in the view of adult is proposed in this thesis. The method about motion capture of this thesis is developed based on the OpenCV(Computer Vision) package, using the Haar Face Tracking and NNC(Nearest Neighbor Classifier) -trees based Face Recognition to capture whether the kid is in the view of the adult.A danger awareness system for kids by combining sensor network and computer vision is developed in this research.Firstly, get the data of context from sensor network, and use the object ontology and danger feature ontology based reasoning mechanism to compute the danger degree of detected context.Then, according to the data of dangerous motion, adjust the danger degree, to make the result more accuracy.Finally, provide the warning service according to final danger degree.In the last part of thesis, there are two experiments for the system, the results of the experiments show that the results of the system about the danger degree of the detected context are close to the expected ones from the adults. The data from the results shows the accuracy rate and the error rate of the motion capture module.