Unmanned aerial vehicles (UAVs) equipped with cameras, capture images and videos from high altitude, usually for surveillance purpose. This paper investigates the power consumption of a drone processing unit under two different operational scenarios: (i) On-Board (local image processing using an onboard Raspberry Pi) and (ii) Off-Board (streaming video over Wi-Fi for remote processing). Therefore, we deployed the well-known YOLO (You Only Look Once) image detection algorithm on the Raspberry Pi and investigated the power consumption for the both scenarios. Contrary to the expectations, our findings prove that energy consumption is very similar in both scenarios. The results of this study are valuable for designers of energy-efficient UAV systems, particularly for applications where prolonged flight duration and real-time data processing are critical.