Runways Detection from Sentinel Images using Enhanced Yolov8 Model

Automation is what future is looking for. Aircrafts and autonomous vision-based vehicles are facing challenging consequences during take-off and landing due to the inefficient detection of runways. The ability to accurately detect airport- runways is critical for enhancing the safety and efficiency of airport-operations. Traditional methods for runway-detection are based on edge-detection or thresholding methods that segment the image by converting pixel intensity thresholds to binary values. Such methods often struggle during undesirable weather conditions and obstructions. The integration of YOLOv8, has shown significant promise in overcoming these challenges. This work enlightens the usage of enhanced YOLOv8 model with better preprocessing techniques for accurate detection of airport- runways. This automated way of detecting the runways will be beneficial during natural calamities, when manual verification is impossible. The proposed model is tested on 880 images and gave the accuracy of 93.2 and false positive rate of 6.32 ensuring the reliability of the model.