Abstract—Recently, drones are used in all fields. The video captured by this drone is sent to the terminal for analysis. In terms of speed, performance, and latency, it would be an advantage if the analysis of the image or video is done onboard, the drone, and the result is sent to terminal, this is called onboard processing. For faster recognition speed and higher frame rate, YOLOv5 is used for image detection along with EfficientNet-b0 for classification and de-blurring with DeblurGan v2. A custom dataset of 6999 military vehicle images is created and annotated. This model is loaded in Raspberrypi4 as it is used as a platform to implement real-time image processing applications since their framework can leverage spatial and temporal parallelism. Integrate the Raspberry Pi board into the drone. The classified images are received in a telegram at the terminal. The accuracy of the model is 88%.
 
Keywords—object detection, military vehicle classification, YOLOv5, drone image, Raspberry Pi 4


Copyright © 2023 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.