A framework based on deep learning and mathematical morphology for cabin door detection in an automated aerobridge docking system

In this paper, a cabin door detection framework based on deep learning and mathematical morphology is proposed. It is applied to an automated docking system for airplane cabin door. This system needs to work under any weather condition like rain, shine, day and night. Limited by the number of videos, just a small dataset based on actual airport operation is established for aerobridge docking process. As the training dataset is small, the trained detector cannot identify all the cabin doors in this dataset. Some of the cabin doors, which are not detected, can be identified with the combination of deep learning and mathematical morphology. Experimental results show that the integration of deep learning and mathematical morphology performs better than the simple deep learning method.