【作者】 王晓宇；

【导师】 陈吉红；

【作者基本信息】 华中科技大学 ， 机械制造及自动化， 2008， 博士

【摘要】 随着第三代非致冷焦平面探测器的日益普及,红外热像仪作为一种非接触在线面扫描测温装置,已经广泛应用与电力、石油、建筑、军工等国民经济的重要行业。红外分析软件作为红外测温技术应用的核心,其涉及的红外图像处理技术近年来成为一热点研究方向。本文针对红外测温及分析应用的若干关键问题提出了一系列解决方案,对方案涉及的理论问题进行了系统研究与实践。针对早期慧眼红外热像仪在不同环境温度下存在测温整体偏移问题,本文提出了基于双向查表法的测温及环境温度补偿算法,该算法在原有热像仪测温档基础上增加了环境温度反查表,提高了热像仪环境温度适应性。虽然红外热像仪被宣传为一种预报性和预防性故障诊断设备,但本文指出了实践中存在的静态性、事后性和信息丢失性等问题。针对这些问题,本文提出了基于闭合等温线轮廓质心距离函数匹配的图像配准算法,算法充分利用了质心距离函数表达(CDF)的拓扑连续性、与轮廓形状的相关性、形状匹配时的鲁棒性和信息保留性等诸多优点,算法简单、高效,在此基础上提出了历史趋势分析方法,赋予了热像仪动态分析能力。针对低空间分辩率的红外图像不能满足建筑等应用中的大场景成像需求的问题,本文研究了基于对数极坐标映射及LMA迭代的红外图像拼接算法,算法利用了对数极坐标映射的非均匀空间采样策略和刚体不变性匹配得到变换初值,研究了以重叠区域的累加平方差为目标函数,采用Levenberg-Marquardt Algorithm(LMA)迭代算法的投影变换空间下高精度拼接,实验表明了该算法的精度、稳定性和适用性。针对红外热像仪在卫生防疫应用中存在的自动化水平低的问题,本文研究了支撑向量机(SVM)算法求解红外人脸检测问题,论证了SVM算法与红外目标识别问题的内在一致性,利用了红外人脸图像与光照无关性,从核函数的选择、学习机参数交叉检验以及步步为营的学习策略等几方面对SVM在红外人脸检测中的应用进行了深入探讨,获得了具有良好性能的SVM。作为本文的研究成果:“红外序列图像历史趋势分析方法”已被授予国家发明专利;“慧眼HY-2188G”型红外热像仪及“Imgsee”红外分析软件,已经成为华中数控重要产品,分别获得国家级新产品和软件著作权各一项,已取得良好的经济和社会效益。更多还原

【Abstract】 With the popularization of the third generation Uncooled Infrared Focal Plane Ar-ray(UFPA),Infrared camera, as a non-contact online thermal temperature measurementequipment, has been widly used in electrical power system, petrochemical, architecture,military industry and services. Infrared analysis software plays a kernel role in the appli-cation of Infrared Thermometry, and the relevant infrared image processing is becoming anew research hot issue. This paper proposes a series of solutions to some key problems inthe application, and in-depth study is exerted to solve the revelant theoretical problems.To improve the tolerance of IR camera to the variance of the environment temperature,a bidirectional table lookup algorithm is proposed. In addition to the temperature measuretable of the early HuiYan IR Camera, another table is introduced to measure the environmentvariance.Although infrared camera is propagandized as a predictive and preventive maintenancedevice, it is used in an ex-post way in fact. When an anomalous temperature pattern isbeing spotted, the equipment has been anomalous and the fault has occurred for some time.Without historical IR images being considered, this usage of infrared camera is static, ex-post and information discarding. This paper proposes a dynamic analysis method for IRimage series of an equipment taken at different times. Centroid Distance Function is used inimage registration for the first time. The spatial distortions will be removed by an automaticregistration algorithm. Every image will share the same spatial structure as a referenceimage. The temperature variation is brought into prominence, history and tendency analysiscan be carried out, with surface scanning ability remained, and the hand-work-load is greatlyreduced.To fulfil the big IR scene analysis need, this paper study the Log-Polar mapping andLevenberg-Marquardt Algorithm for IR image mosaicing. The spatial variant samplingstructure and rigid variance properties are fully utilized. Experiment result shows this algo-rithm is precise, stable and wide applicable.The SARS epidemic resulted in the introduction of IR camera for non-voluntary screen-ing on human face for fever symptoms, but they are not working in an automatic way. Thispaper present an human face auto detection method based on SVM. The inherent consistencyof SVM with the problem is discussed. A smart biometrics system that automatically de-tects human face in infrared video and performs temperature measurement is implemented.The potential for illumination invariant face recognition using thermal IR imagery is fully utilized.As results of this paper,”HY-2188G”IR thermal camera and”Imgsee”IR analysissoftware have become important products of HuaZhong Numeric Conrol System Company.更多还原