【作者】 田裕鹏；

【导师】 于盛林；

【作者基本信息】 南京航空航天大学 ， 测试计量技术及仪器， 2009， 博士

【摘要】 红外辐射成像检测由于其非接触、快速、一次检测面积大等特点,成为引人注目的无损检测方法。本文对红外辐射成像检测中的关键技术及其定量化问题进行研究,所涉及的主要关键技术有:红外辐射成像检测实验及序列热图的预处理技术;红外辐射成像检测的热传导模型及有限元分析计算;脉冲相位辐射成像检测技术;红外辐射成像检测中的图像融合技术。针对红外辐射成像检测定量化水平不高的问题,分别在时间域、频率域中对缺陷检测的定量化问题开展研究。全文的主要内容如下:构建了红外辐射成像检测的实验系统,针对热激励源控制、检测参数设置、检测条件及检测结果等,开展红外辐射成像的实验研究。主动式红外辐射检测面临加热不均、检测对象表面发射率、表面状况对检测的影响等问题,为此根据红外图像检测的特点和需要,提出并实现了图像k近邻高斯平滑滤波、缺陷中心位置检测、旋转跟踪最大梯度缺陷边缘检测及改进的判断分析缺陷分割等预处理功能,为进一步的缺陷定量化分析打下基础;首次采用本征模分解这一具有局部自适应分析能力的信号分析方法,对红外辐射检测的热图像进行处理,指出热图像噪声和加热不均的本征模式特征,通过一次本征模分解实现了热图像噪声抑制和加热不均的修正,改善了图像对比度。在红外辐射热传导分析的基础上,对缺陷深度和面积大小的红外定量检测问题进行实验和有限元模拟计算研究。采用对时间序列图像求强度平均的方法,由信号本身确定检测中激励开始和结束时间点,为分析缺陷与时间信息的关系建立时间基准。分析了缺陷深度与序列图像对比度时间变化的关系,给出了起始分离时间点的概念,提出并实现了根据温度信号随时间的变化确定上述时间点的方法,对缺陷大小、激励强度、激励时间对缺陷深度分析的影响进行研究,验证了根据上述时间点确定缺陷深度的可靠性,并与对比度最大值时间点等进行对比。在缺陷面积大小的定量分析中,根据时间序列图像中缺陷面积大小检测变化的规律,确定缺陷面积大小检测的最佳时间段,根据对实际缺陷温度异常分布及热传导的分析,对半最大对比度宽度方法进行了改进,提高了检测精度。脉冲相位热成像检测技术是近年来出现的新型红外无损检测技术,由于其抗干扰能力、探测深度缺陷的能力强,引起人们广泛关注。本文对脉冲相位热成像检测技术中频谱分析的有关问题进行研究,指出合理选择采集频率、采集窗口大小要考虑的影响因素;对通过拟合扩展和插值,相应改变采样时间和采样频率,实际上提高了傅立叶变换的窗口长度和频谱范围进行探讨,一定程度上改善了缺陷检测效果;并分别采用组合平滑滤波、分段拟合的方法改善了频谱变换的实际效果。对脉冲相位辐射检测的定量化问题进行研究,采用有限元方法对缺陷深度检测的盲频率理论进行了计算分析;对缺陷面积大小检测中热扩散对相位、振幅、对比度图像的影响及其造成的误差进行分析,总结其规律,指导缺陷的定量化分析工作。任何一种无损检测技术都有其局限性,探索多种检测方法的融合处理是无损检测的发展方向。针对红外辐射成像检测易受干扰、检测结果解释困难等问题,开展了红外辐射成像检测中可见光与红外图像的像素级图像融合及其应用的研究。首先根据像素级图像融合的要求,进行基于特征点的多模式图像配准研究,采用CCS方法提取角点作为特征点,设计了一种新的形状描述子完成特征点的匹配。在像素级图像融合方法方面,根据红外辐射成像检测的目的和要求,实现了对比度调制融合、金字塔变换法、小波变换法和EMD分解的多分辨图像融合、透明度图像融合等多种图像融合算法。实验和现场检测结果都证明图像融合有助于红外无损检测中的缺陷解释和定位,明晰了检测结果。更多还原

【Abstract】 Infrared Thermography has been found to be a very valuable nondestructive evaluation (NDE) due to its non-contactness, rapidity, capability of imaging large area. Research on the key technologies and quantitative infrared thermography is carried through in this dissertation. The key technologies include: infrared thermography experiment and pre-processing technology of sequence images, heat conduct model and finite element analysis of infrared thermography, pulsed phase thermography (PPT), and pixel-level image fusion for infrared thermography. On the quantitative infrared thermography, the research is carried through in time and frequency domain respectively. The main contents of the dissertation are as follows:The experiment system of infrared thermography is set up and experiment studies are carried through on heating stimulator, experiment parameter, experiment condition and result. To remove the influence of uneven heating, emissivity and status of the sample surface, according to the characteristics and needs of infrared thermography, k-neighbor gauss image averaging, defect center location, rotary tracking edge detecting, and improved estimation analysis image segmentation are achieved. This lay a foundation of quantitative infrared thermography. For the first time, an empirical mode decomposition (EMD) method, which is an adaptive decomposition method, is applied to image processing of infrared thermography. Through analysis of mode function of image EMD, the characterizations of noise and uneven heating are given. Using EMD method, the noise and uneven heating are reduced at the same time and the contrast of thermal image is improved. Good result shows this method is simple and effective to overcome noise and uneven heating in infrared thermography.Experiment and finite element analysis of quantitative evaluation of defect depth and size are carried through based on the heat conduction. Firstly, the time points of beginning and end of stimulation are found out by averaging the intensity of sequence images and the base origination time point is set up for evaluation of defect depth. The relation between defect depth and sequence images contrast changing with time is analyzed. The concepts of initialization separation time point is given. The method of achieving the time points according to temperature changing with time is realized and comparison was made with maximum contrast time point. The influence of defect size, external heat stimulation intensity and duration on acquiring defect depth was analyzed. The reliability of acquiring the defect depth according to the time points above is proved. The changing rule of defect size was found and best measurement time region was given. Based on the abnormity and heat conduction of defect, full width half maximum method is improved and the measurement precision is advanced.Pulsed Phase Thermography (PPT) is a new infrared nondestructive evaluation method appeared recently. It brings abroad attention due to its advantage of strongly ability of detecting deeper defects and anti-disturbing. In this paper, The problems of FFT(fast Fourier Transform) of PPT were studied. The key factors for choosing the sampling frequency and truncation were given. This paper also studies how to change the value of sampling frequency and window width through fitting extend and interpolation methods. And, to a certain extent, the result of measurement is improved. The result of fourier transform was improved practically by combining low-pass filters and subsection curve fitting. The theory of phase blind frequency with defect depth of the quantitative pulsed phase thermography is studied through finite element analysis. Finally, the influence and error analysis of thermal diffusion on phase, amplitude and contrast image are studied .No NDT is perfect and it is the developing orientation of NDT to combine different NDTs. In this paper, for reducing noise and exactly interpretation of thermography, the pixel-level image fusion of infrared light and visible light image for infrared thermography and its application are studied. Firstly in the pixel-level image fusion, different mode image registration has been done based on feature corners. The corners and curves relative to each corner are extracted using the CSS (Curvature Scale Space) method. A new shape context descriptor for each corner is given as the criterion to match the corners. According to the purpose and need of infrared thermography, fusion methods of contrast modulation fusion , multi-scale fusion methods of Laplacian pyramid, wavelet transform and EMD, transparent fusion are studied. The experimental and application results indicate image fusion for infrared thermography can facilitate defect location and interpretation of inspection results.更多还原