基于计算机视觉的客机舱门识别与定位技术研究

Research on Recognition and Positioning Technology of Passenger Aircraft Door Based on Computer Vision

摘要:

登机桥是机场将航站楼与飞机连接的活动通道，登机桥与客机舱门对接系统的智能化变得尤为重要。对于基于计算机视觉的客机舱门识别与定位系统，其关键组成部分是目标检测系统。传统的目标检测算法通过提取传统手工特征进行学习，不能达到鲁棒性好、速度快、准确性高的检测要求。基于迁移学习在深度学习上的应用，利用SSD（Single Shot Multibox Detector）算法，以轻量化的MobileNet作为特征提取网络，实现了鲁棒性好、准确度高的目标检测模型，完成对客机舱门的识别与定位，对不同样式的舱门、部分遮挡、背景变化、光照变化、运动模糊具有鲁棒性，能准确完成识别功能，完成对舱门在当前视觉图像中的相对位置的解算。

Abstract:

The boarding bridge is the active passageway connecting the terminal with the aircraft at the airport. The intelligentization of the docking system between the boarding bridge and the passenger cabin door becomes particularly important. For the passenger cabin door recognition and positioning system based on computer vision, the key component is the target detection system. The traditional target detection algorithm learns by extracting traditional manual features, which cannot meet the detection requirements of good robustness, fast speed and high accuracy. Based on the application of transfer learning in deep learning, using the SSD (Single Shot Multibox Detector) algorithm, using lightweight MobileNet as a feature extraction network, the target detection model with good robustness and high accuracy is achieved, and the passenger cabin door The identification and positioning of the model are robust to different styles of doors, partial occlusion, background changes, lighting changes, and motion blur, and can accurately complete the recognition function and solve the relative position of the doors in the current visual image.