【作者】 刘启海；

【导师】 林玉池；

【作者基本信息】 天津大学 ， 测试计量技术及仪器， 2011， 博士

【摘要】 可见光视觉测量技术是一种常用的非接触式尺寸测量方法,但是在高温、高速气流等恶劣环境下,可见光相机的成像效果会受到较大影响,并会产生光线偏折的问题。红外热成像技术以接收目标自身的热辐射并进行光电转换为基本工作原理,在解决可见光面临的上述挑战时具有较大的优势。以高温恶劣环境下的尺寸测量方法为主要研究目标,结合航天某院提出的某测量任务和要求,提出了高温构件三维尺寸测量的红外视觉方案,并围绕此方案的相关理论和关键技术展开研究,主要的研究内容如下:1.概括论述了高温恶劣环境下尺寸测量的需求和应用,总结分析了国内外高温构件尺寸测量方法的发展状况和红外热成像技术的应用研究进展,明确了研究方向和意义。2.从普朗克黑体辐射定律、可见光CCD的响应曲线等出发讨论了高温恶劣环境下影响视觉成像效果的因素。利用有限元方法分析了高温空气场内的温度分布,在此基础上,结合光的色散模型,选取0.38μm、0.65μm和14.0μm三个典型的波长,求解了空气场内各区域的折射率分布情况和光线在传输过程中产生的偏折程度。最后通过实验进一步验证了可见光视觉和红外视觉影像误差的大小。结果表明,构件处于约的高温状态时,在距离约1m处,可见光的影像误差约为远红外视觉的10倍。900?C3.从红外热像仪的理想透视模型出发,建立了高温构件三维尺寸红外视觉测量系统的数学模型。提出了两种比较容易实现的测量方案:简化的高温红外双目测量方案和基于单热像仪的测量方案。设计了红外视觉测量的总体方案,根据实际测量条件确定了方案中各硬件的型号参数。最后,搭建了高温环境和被测构件模拟系统、红外与可见光视觉成像对比实验系统、简化的高温红外视觉测量系统、基于单个热像仪的高温红外视觉测量系统以及高温红外视觉标定系统等实验平台,并进行了一些简单的验证实验。4.研究了高温构件红外图像非均匀性校正方法和红外图像增强方法。改进了传统BP神经网络应用于非均匀性校正时的一些不足,提高了校正的效果。此外,针对高温构件红外图像的纹理不足的特点,提出了基于方向波变换的红外图像增强方法,采用方向波变换对图像进行频域分解,建立多方向性的各向异性滤波器,提高了构件红外图像的增强效果。5.研究了高温红外视觉测量系统中红外热像仪内外参数的标定方法。结合红外热成像的原理,设计了一种专用的高温红外视觉测量系统的标定靶标。分析了可见光视觉测量中基于径向一致约束(Radial Alignment Constraint,RAC)的标定方法和张正友标定方法,并利用径向一致约束准则结合遗传优化算法对测量系统进行标定。结果表明,在双目热像仪前方距离约900mm处,传统标定方法的误差最大不超过0.85 mm,改进方法的误差最大不超过0.41mm,改进方法提高了系统标定的精度。6.研究了高温构件边缘提取、双目特征匹配和三维测量的方法。提出了一种基于脉冲耦合神经网络和空间灰度矩相结合的红外亚像素级边缘检测方法,有效地提高了检测的自适应性和精度。分析了基于极线约束关系的初始匹配方法,以及多种手段相结合的误匹配剔除方法。通过对三维离散坐标数据的处理,完成了高温构件主要三维尺寸的测量和形貌的重建。论文的主要创新工作包括:1.提出了基于红外热成像技术的高温构件三维尺寸测量方案,减小了在高温环境下光线偏折等原因对测量结果的影响。2.针对高温构件红外图像纹理不足的特点,提出了基于方向波变换的红外图像增强方法,获得了良好的效果。3.研究了一种基于径向一致约束准则与遗传优化算法相结合的红外双目视觉标定方法。获得了较高的标定精度。4.研究了一种脉冲耦合神经网络和空间矩相结合的高温构件亚像素级红外边缘检测方法,有效地提高了红外边缘检测的自适应性和精度。更多还原

【Abstract】 Visible vision technology is a non-contacting method widely used in geometry size measuring, however, the visible imaging result will be greatly affected in harsh environments such as high temperature, high-speed airflow, etc. Infrared thermal imager receives the radiation of target and converted it to electrical signals through photoelectric conversion. It has many advantages in dealing with the challenge in some harsh environments compared with visible camera.Centering on three-dimensional size measuring methods in harsh environments of high temperature, and in combination with the task and demand from a space research institute, a novel method based on infrared vision is proposed to measure 3-D size of high-temperature components, and the related theory and key techniques in this method are researched. The main research contents are as follows:1. The thesis discusses the demand and application of size measuring at high temperature, and it also summarizes the comtemporary state-of-art size measuring methods of high-temperature components and the development of infrared thermal imaging application technology. The research direction and meaning are then determined.2. Based on the Planck blackbody radiation law and CCD’s response curve,?factors that affect the vision imaging result in high temperature harsh environment are detailed discussed. The field distribution of high-temperature air is analyzed by means of finite element method.Then the refractive index distribution and the bending extent of light at the selected wavelengthes of 0.38μm, 0.65μm and 14.0μm are both calculated.The bending extent of visible light and infrared light are further verified through some imaging experiments, and results illustrate that the bending extent of visible light is about 10 times compared with far-infrared light at ? and a distance of 1.0 m. 900?C3. A mathematical model of 3D infrared vision measurement systems used for high-temperature component size measurements is established based on the ideal perspective model of infrared thermal imager. Two measurement molds are proposed, which are the simplified high-temperature infrared binocular measurement mold and the measurement mold based on single thermal imager. Both of them are easy to implement. An overall scheme for the infrared vision measurement is presented, and the hardware parameters in the scheme are provides. Some experiment platforms are set up, including the imitation of the actual measured workpiece and the actual high-temperature environment, the comparative experiment system of infrared and visible vision imaging, the simplified high-temperature infrared measurement system, the high-temperature infrared vision measurement system based on single thermal imager and the high-temperature infrared?binocular vision calibration system.4. Non-uniformity correction methods and image enhancement methods of infrared image from high-temperature components are studied. By improving the traditional non-uniformity correction method based on BP neural network, the non-uniformity of infrared image is reduced. In addition, to overcome the lack of sharpness for infrared image from high-temperature component, several measures are developed to heighten the enhancement effect of infrared image, including a novel enhancement algorithm named directionlet transform. In this method, infrared image is decomposed in frequency domain by means of directionlets, which uses some anisotropic and multi-directional filters.5. Calibration methods of internal and external parameters of infrared thermal imagers in high-temperature infrared vision measurement system are researched. A calibration target is designed for infrared high-temperature vision measurement according to the characteristics of infrared thermal imaging. The binocular calibration method based on radial alignment constraint and Zhang Zhengyou algorithm are mainly analyzed. On this basis, the infrared binocular measurement system is calibrated with a method which combines radial alignment constraint and genetic optimization algorithm. The results illustrated that the calibration error of the novel method is less than 0.41mm, and that of traditional methods is less than 0.85mm when the measured workpiece is about 900mm away.6. The methods for edge extraction, binocular feature matching and 3D reconstruction are studied. A sub-pixels method by combining pulse coupled neural network(PCNN) and spatial gray moment is applied to effectively improve the adaptivity and accuracy of edge detection. The approaches to binocular feature initial matching are figured out by using epipolar geometric constraint, and system works efficiently with some other methods which can reduce the possibility of mismatches. Finally, three-dimensional critical sizes measurement and shape reconstruction of high-temperature components are accomplished via data processing of 3D discrete coordinates. The main innovations of this study include:1. For high temperature environment, a solution for 3D size measurement of high-temperature components based on infrared thermal imaging technology was proposed, which can decline the influence of light bending on vision measurement.2. An infrared image enhancement algorithm based on directionlet transform is developed to enhance the feture such as edge and texture of high-temperature component. The results showed that it gets a better enhancement effect compared with some common methods.3. A novel infrared binocular calibration method combining radial alignment constraint with genetic optimization algorithm is researched. Experiments illustrated that it can effectively improve the accuracy of infrared binocular vision system, and thus is more suitable for infrared binocular calibration.4. A sub-pixel level edge detection method based on PCNN (Pulse Coupling Neural Network) and spatial gray moment is proposed, with which the adaptivity and accuracy of infrared edge detection were both improved.更多还原