【作者】 赵静；

【导师】 高庆吉； 牛国臣；

【作者基本信息】 中国民航大学 ， 控制科学与工程， 2017， 硕士

【摘要】 传统的飞机油箱检查方式是人工检查方式,工作人员往往需要钻入环境恶劣的油箱内部作业,危险性高且效率低下。为实现飞机油箱内部结构状态的视觉检测,课题对基于图像的飞机油箱内窥形貌建模方法及基于特征分析的缺陷检测方法进行研究,对提高飞机油箱检查效率,保证工作人员人身安全具有重要意义。第一,研究了飞机油箱内窥形貌全景图建模和缺陷检测过程中的图像获取问题。根据飞机油箱内部的结构特点及飞机油箱检修的需求,设计了飞机油箱内窥形貌观测系统,来完成图像采集;针对飞机油箱内窥形貌观测系统存在几何变形的问题,进行标定及畸变校正;研究了基于图像的飞机油箱内窥形貌全景图模型的三种投影方法。第二,设计了飞机油箱内窥形貌全景图建模过程中的图像拼接方案。拼接过程首先要对飞机油箱内窥图像进行特征点检测、初匹配和筛选;针对飞机油箱内窥图像的配准过程实质上属于不同平面内的场景匹配问题,研究了一种基于结构相似度的飞机油箱内窥图像配准算法;并研究了图像拼接效果的评价指标。第三,设计了基于特征分析的飞机油箱内窥图像缺陷判别方法。利用基于缺陷特征向量的判别方法判断飞机油箱图像中有无缺陷,而后根据复合特征分析方法找出缺陷分类依据并判断缺陷类别,最后采用基于运动恢复结构的单目定位算法对飞机油箱内部目标区域进行定位,为工作人员维修油箱提供参考。最后,对飞机油箱内窥形貌观测系统软、硬件平台进行了简要说明,并分别对本文所研究方法进行实验分析,验证了飞机油箱内窥形貌建模方法的有效性与缺陷识别方法的正确性。更多还原

【Abstract】 The traditional manual aircraft tank inspection requires the maintenance crew to enter the tank to perform the inspection and maintenance tasks,which have the problems of high risk and low efficiency.In order to realize the visual inspection of the aircraft fuel tank,the image-based endoscopic modeling method was adopted,and the defect detection method based on the feature analysis was put forward.The research is of great significance to improve the aircraft fuel tank inspection efficiency and ensure the safety of the maintenance crew.Firstly,the image acquisition approach for panorama modeling and defect detection was proposed.According to the internal structural characteristics and the inspection requirements of the aircraft fuel tank,the observation system of the aircraft tank endoscopic appearance was established.The geometric deformation of the observation system was calibrated and corrected.In addition,three kinds of projection methods of the image-based aircraft tank endoscopic appearance panorama were analyzed.Secondly,the image stitching scheme in the process of endoscopic panorama modeling of aircraft fuel tank was presented.Detecting,the initial matching and the screening of the feature points extracted from the aircraft fuel tank endoscopic image were carried out.In fact,the registration process of aircraft fuel tank endoscopic image is essentially a scene matching problem in different planes.So the aircraft tank endoscopic image registration algorithm based on structural similarity was proposed to solve this problem.And the evaluation indexes of image mosaic effect were presented.Thirdly,the method of defect discrimination based on feature analysis was proposed.The defect identification method based on feature vector was used to determine whether there are defects in the image.Furthermore,the composite feature analysis was adopted as the defect classification to judge the defect category.The monocular localization algorithm based on structure from motion was presented to locate target area of the aircraft tank and provide reference for the maintenance crew.Finally,the prototype was established to verify the methods proposed in this paper.The experiments showed that the validity of the modeling method and the correctness of the defect discrimination method.更多还原