δ-ATP: Towards Fast and Reliable Aerial Traffic Patrol

In this work, we present our efforts to develop δ-ATP, a working system for fast and reliable traffic patrol using drones in the sky. Compared to the state-of-art approaches, δ-ATP improves accuracy and responsiveness in detecting speeding incidents by focusing only on essential "delta" (δ) information - regions of interest (say, roads) and key frames needed for speed estimation. We have implemented and evaluated δ-ATP in real-world experiments, successfully detecting all the speeding vehicles and outperforming all other approaches with a higher accuracy for real-time traffic enforcement. The datasets collected in our field test are available at [7].