TinySurveillance: An Extra Low-Power Event-Based Surveillance Method for UAVs

Uncrewed Aerial Vehicles (UAVs) have always been faced with power management challenges to extend their flight time. Managing power consumption becomes critical, especially in surveillance applications, where the longer flight time results in wider coverage and a cheaper solution. Most of the current studies show new methods for event detection without considering power consumption. This article presents an event-driven four-stage video surveillance pipeline with an efficient video transmission algorithm balancing power consumption and image quality. The surveillance starts automatically when the low-power AI-based onboard processor detects the desired event. When The edge node detects the defined event, a sample image is sent to the server for validation. After validation, a colored image accompanied by N grayscale images are sent to the server. The server colorizes the grayscale images using a convolutional neural network trained by the colored images. In this work, an application of wildfire detection and surveillance has been implemented to show the proof of concept of the TinySurveillance method. The results show that the power consumption of the onboard processing unit in detection mode reduces by at least 4 times; during the surveillance mode, the data transmission rate can be decreased by almost 66% while achieving a competent image quality PSNRAvg of 41.35 dB, PSNR of 30.94 dB, and output frame rate of 5.2.