树分类器在多界面粘接质量图像诊断中的应用

Application of tree classifier in the image diagnosis of the adhesive quality about the multi-layered structure

在对固体火箭发动机的钢、衬层、绝热层复合结构粘接质量进行超声图像诊断过程中 ,利用两检测回波之间的谐振关系确定作为特征值的延迟时间 ,并把特征值提取和多界面粘接状态识别归结为一个多模式识别问题 ,利用自行设计的二叉树分类器对其进行分类识别 。

When the adhesive quality in multi-layered structure (steel shell, under layer and insulation layer) which is used in the solid rocket motor is diagnosed, the resonance between the echoes is used to obtain the signal characteristics, i.e. delay time. Extracting the signal characteristics and recognizing the adhesive state are regarded as a pattern recognition process. The dichotomous branch tree classifier is designed to accomplish the pattern classification. The high quality diagnosis image is obtained.