Comparative Image Processing and Object Detection in Aerial Search and Rescue

Aerial search and response plays an important role in finding and rescuing persons in need. Unmanned Aerial Vehicle (UAV) -acquired aerial images provide an intensive profile search area and facilitate identification of potential targets. To improve the effectiveness of aircraft Search and Rescue missions, this study compares many object detection algorithms and image processing methods. The investigated object detection methods are Mask R-CNN, SSD, and YOLOv8, and image processing techniques are contrast improvement and image brightness control. The effectiveness of these strategies has been evaluated by using real aerial images that include different targets such as humans, vehicles, and fire. The results show how well the object detection algorithms work to accurately distinguish targets, and how an attractive image processing approach improves the image performance of the best object detection model, as well as how the performance of the best object detection model with image processing can improve the performance of model. This research shows that combining image processing (contrast and brightness) with an optimal object detection algorithm (Mask R-CNN) significantly improves target detection in aerial search and rescue. While only Mask R-CNN performs best, adding image processing increases its accuracy to 0.92, mAP to 0.89, and makes it more reliable in finding people in need. This shows the potential of this joint approach to save people in air rescue missions.