Event-driven dynamic platform selection for power-aware real-time anomaly detection in video

In surveillance and scene awareness applications using power-constrained or battery-powered equipment, performance characteristics of processing hardware must be considered. We describe a novel framework for moving processing platform selection from a single design-time choice to a continuous run-time one, greatly increasing flexibility and responsiveness. Using Histogram of Oriented Gradients (HOG) object detectors and Mixture of Gaussians (MoG) motion detectors running on 3 platforms (FPGA, GPU, CPU), we characterise processing time, power consumption and accuracy of each task. Using a dynamic anomaly measure based on contextual object behaviour, we reallocate these tasks between processors to provide faster, more accurate detections when an increased anomaly level is seen, and reduced power consumption in routine or static scenes. We compare power- and speed-optimised processing arrangements with automatic event-driven platform selection, showing the power and accuracy tradeoffs between each. Real-time performance is evaluated on a parked vehicle detection scenario using the i-LIDS dataset. Automatic selection is 10% more accurate than power-optimised selection, at the cost of 12W higher average power consumption in a desktop system.