红外技术, 2017, 39 (11): 1018, 网络出版: 2017-11-27  

基于主成分分析的热图像序列盲源分离

Blind Source Separation of Thermal Image Sequences Using Principal Component Analysis
作者单位
装甲兵工程学院机械工程系,北京 100072
摘要
在涡流脉冲热像技术中,高频涡流瞬时加热被测物体时,不同区域的热响应会发生混叠现象,这势必影响缺陷区域热响应信号的判别。本文以红外图像序列为观测信号,建立热响应信号的混叠模型;其次,利用不同区域的热响应彼此独立的特点,开展了基于主成分分析的盲源分离数据处理方法研究;最后,建立仿真模型研究了不同区域的热响应形态,采用了基于混叠向量和峰度系数定量分析主成分强化的区域。实验结果表明该方法能够实现不同生热区域的盲源分离,为缺陷的特征提取和识别提供了理论支撑。
Abstract
: In eddy current pulse thermography, test specimens are heated instantaneously by a high-frequency eddy current, and the thermal response of different region aliasing occurs, which inevitably affects the identification of the thermal response of a defect area. In this paper, a blind source separation model is developed that takes the thermal response as an observation,. Since the responses of different areas are independent, principal component analysis is employed to separate data. Then, a simulation model is established to study the thermal responses of different regions. Based on this model, we propose a method of identifying enhancement regions using principal components based on an aliasing vector and kurtosis coefficient. Experimental results show that the method can separate the principal components that describe different heating areas. This result provides theoretical support for the feature extraction and automatic identification of defects.

徐超, 冯辅周, 闵庆旭, 孙吉伟, 朱俊臻. 基于主成分分析的热图像序列盲源分离[J]. 红外技术, 2017, 39(11): 1018. XU Chao, FENG Fuzhou, MIN Qingxu, SUN Jiwei, ZHU Junzhen. Blind Source Separation of Thermal Image Sequences Using Principal Component Analysis[J]. Infrared Technology, 2017, 39(11): 1018.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!