VisionLab

Signal Processing and Anomaly Detection for Multi-Sensor Systems

This project investigates the use of multiple sensors in Anomaly Detection. Smartphones are of particular interest as they provide network capabilities along with the processing power. Having network of sensors is beneficial in several ways, such as, computation and communication offloading for saving energy, sharing large databases, pre-empting of targets or danger, target Re-identification etc. Experiments show that offloading may help save energy and time. Below is the output showing time and energy required to send image as opposed to running vision algorithms locally. The figure shows runtime and energy consumption of popular Histogram of Gradients (HOG) and Viola Jones Face Detection on an Android mobile phone (Sony Z1 compact). Currently we are working towards target re-identification across cameras in multiple smartphones.