Yonghwa Kim, KyoungYeon Kim, and Yoo-Sung Kim
Loitering, multiple cameras, trajectory summarizing features, data mining
In this paper, a loitering detection model for multiple camera environments in which multiple cameras are of location-synchronized is proposed. This model transforms the location of a human in each two-dimensional video into the integrated three-dimensional (3D) coordinates of the multiple camera environments and extracts the trajectory summarizing features of the person within the integrated 3D coordinates. Then, using 3 machine learning algorithms with the trajectory summarizing features as the training data, loitering event detection model which can distinguish loitering event from normal walking in multiple camera environments is created. From the results came out from the experiments with the self-taken test video data, we confirmed that the proposed detection model is able to detect loitering events accurately that were not recognizable with a single camera.
Important Links:
Go Back
