Multispectral imaging: Monitoring vulnerable people

摘要： This paper describes the development of a new healthcare monitoring system for vulnerable people which uses a visible image sensor and passive infrared sensors, in an unconventional manner, to monitor daily living activities. It based on a novel method, using chromatic methodology, to process spatially and temporally the incoming multispectral data from the visible and infrared parts of the spectrum, to overcome the impact of noisy environments, illumination changes and a dynamic background. An efficient chromatic descriptor is suggested to improve activity recognition of vulnerable people. The new monitoring system is robust to distortions associated with healthcare systems and its descriptor has an improved quality of description. System performance was evaluated using a series of experimental data, the results showing the efficacy of using both spatial and temporal domains of multispectral data to deal with events that disturb monitoring systems. The chromatic descriptor achieved a better performance in comparison to traditional methods when describing daily living activities.