Data-Driven Logical Topology Inference for Managing Safety and Re-Identification of Patients Through Multi-Cameras IoT

Abstract

As Internet of Things (IoT) develops, IoT technologies are starting to integrate intelligent cameras for managing safety within mental health hospital wards and relevant spaces, seeking out specified individuals from these surveillance videos filmed by the various cameras. Because monitoring is one of the important application of IoT based on distributed video cameras. In order to fine-grained re-identification of patients and their activities against the very low resolution, occlusions and pose, viewpoint and illumination changes, we propose a novel data-driven model to infer multi-cameras logical topology and re-identify patients captured by different cameras. In our model, we employ a Time-Delayed Mutual Information (TDMI) model in order to address multi-cameras logical topology inference. Additionally, we use a well-trained Deep Convolutional Neural Network (DCNN) to extract characteristics. Moreover, we employ a name-ability model to discover deep attributes and a classifier based on a structural output of attributes is designed to tackle the re-identification of patients, especially who possess psychiatric behaviour. In order to improve the present model's performance, we resort to the parallelized implementations. Experimental results show that our model possesses the best performance as compared to state-of-the-art model,especially, when the semantic restrictions are imposed onto the production of patients' specific attributes with structural output. Further, the deep learning model is used to produce characteristics when there is no supervision on the learning model of attributes.

Keywords

• Canonical correlation analysis
• time delayed mutual information (TDMI)
• deep convolutional neural network (DCNN)
• multi-camera topology inference