Deep learning-based Vision System for Human Detection in Drone Videos

Human detection from a drone-based videos has many potential applications such as searching for missing persons, surveillance of illegal immigrants, and monitoring of critical infrastructure. However, it is very challenging computer vision problem to be tackled. The difficulties are related to many aspects including the variations in camera view, the changes in illuminations and weather conditions as well as the variations in the surrounding objects. Recently, deep learning-based vision systems have been proven a great success in many object detection problems. Therefore, this work aims to develop deep learning-based vision system which applied for the problem of human detection from videos captured by a drone-based camera. Particularly, the presented system comprises a detection approach which consists of Faster R-CNN deep learning model to detect the human inside the captured drone-based images. To assess the performances of the proposed vision model, various videos were recorded using drone at different places, from various views and various weather conditions. The outcomes show the effectiveness of the proposed system for human detection in drone-based videos.