摘要
针对目前飞机蒙皮无损检测,目视检测存在人为因素影响严重,其他检测手段检测成本高,检测过程复杂。采用基于飞机蒙皮图像灰度共生矩阵的检测方法,通过机器视觉系统,采集飞机蒙皮图像,建立了正常、裂缝、腐蚀和撞击四种飞机蒙皮图像库;根据飞机蒙皮不同损伤表面纹理不同,提取图像的灰度共生矩阵,提出了基于样本紧密度模糊支持向量机的蒙皮图像损伤识别方法。该方法在欧氏空间考虑损伤样本类别的距离度量特性,通过定义模糊连接度,提高了样本的区分度。实验表明,采用样本紧密度模糊支持向量机识别率要优于支持向量机以及普通模糊支持向量机,蒙皮损伤识别率可以达到93.333%。
In the non destructive testing field of aircraft ,visual inspection is affected seriously by human factors, other detection methods cost high, the detection processes are complicated, or only certain materials could be detected. In view of the current detection and the urgent need of the aircraft skin, a recognition method based on Gray Level Co-ccurrence Matrix （GLCM） is presented. The images （Normal, Crack, Corrosion and Impact）of the aircraft skin were collected by machine vision systems, then the image database of aircraft skin was established, ac- cording to different textures with different damages, the gray-level co-occurrence matrixes of the images were extrac- ted, and the Fuzzy Support Vector Machine （FSVM）based on sample affinity method has been used to classify the damages, which considers the distance measuring characteristics between sample classes in Euclidean space. The samples＇ separating capacity was improved by defining the fuzzy connectivity. The experiment results show that the presented method is superior to the SVM and FSVM at a 93. 333% recognition rate.
出处
《科学技术与工程》
北大核心
2013年第10期2901-2905,2910,共6页
Science Technology and Engineering
关键词
飞机蒙皮
损伤识别
灰度共生矩阵
模糊支持向量机
样本紧密度
aircraft skinmachine sample affinitydamage identificationgray level co-ccurrence matrix fuzzy support vector
作者简介
王昊（1987-），男，河北沧州人，硕士研究生。研究方向：模式识别与智能控制。E-mail：fighterwans01@163．com。
