航空发动机孔探损伤识别方法

针对目前航空发动机孔探检测不能对损伤类型自动识别现状，将支持向量机与孔探检测技术相结合，提出基于支持向量机（SVM）的损伤图像识别方法。该方法将损伤图像进行二值化分割，利用链码跟踪及灰度共生矩阵分别提取损伤区域的形状特征和纹理特征，组成多维特征向量，输入支持向量机进行分类识别。分类器设计阶段，组建性能优越的二叉树支持向量机以减少训练样本，提高分类效率。CFM56发动机实验结果表明：该方法的识别性能明显优于传统SVM多分类器和BP神经网络方法。

Aero-engine interior damage recognition based on support vector machine

Targeted at the failure of most of the current aero-engine borescopic inspection system to identify the kind of interior damages automatically, this paper introduces a new damage recognition method which combines support vector machine （SVM） with borescopic inspection technology. The method consists of converting the damage image to a binary image, extracting five shape features and four texture features from the chain-code and gray-level co-occurrence matrix of the image respectively and putting these features into SVM to carry out automatic classification of damages. The design of the classifier involves the development of a high-performance binary tree-SVM which decreases the number of training sample and improves the efficiency of SVM. CFM56 aero-engine shows a higher recognition accuracy than traditional SVM multi-class method and BP neural network method.