Airport Ground Object Localization Based on Monocular Camera

For the foreign objects on airport pavements, this study examines and compares the accuracy of ground object localization using geometric mapping estimation method and one multi-layer perceptron approach based on one monocular camera. It quantitatively evaluates the impact of observed height, pitch angle, field of view, and coordinates of objects on localization. The findings indicate that localization error positively correlates with distances of objects, while image distortion may exacerbate errors in certain scenarios. Overall, even geometric mapping estimation can be easily and quickly deployed, the multi-layer perceptron approach performs better adaptability with sufficient training dataset.