图像

The research work in this dissertation aim to apply the advanced technique in practice, that ramifies three parts: at first, the basic study of the computer vision, especially the stereo vision, include their technical mechanisms, methods and algorithms resolves the same technical problems of endoscopy and ferrograhpy based on stereo vision.

本文的研究工作以技术应用为目标，分为三大部分：首先是对计算机视觉，特别是立体视觉技术机理、方法和算法的基本研究，主要解决了立体视觉在内窥探伤与铁谱检测中应用的技术共同点；其次是对基于立体视觉的发动机内窥探伤的技术研究和系统开发，解决了立体视觉用于内窥图像这一宏观图像的技术和算法实现问题，其中包括对内窥场景的视觉重建和对发动机内部典型损伤的分析；最后对基于立体视觉的磨粒分析技术的系统构架、实施方法和相关算法进行了研究，主要解决了立体视觉用于磨粒显微图像的技术问题，并对融合三维描述参数的磨粒形态综合识别方法进行了初步研究与探讨。

The research work in this dissertation aim to apply the advanced technique in practice, that ramifies three parts: at first, the basic study of the computer vision, especially the stereo vision, include their technical mechanisms, methods and algorithms resolves the same technical problems of endoscopy and ferrograhpy based on stereo vision. Secondly, the endoscopy of engine fault inspection based on stereovision is researched and the system faced to this application is developed, thus the problems of stereovision applied to macroscopical images such as endoscopic images are resolved in method and computation. In addition, inside scenes are reconstructed and typical damages of engine inside are analyzed. Thirdly, the investigations of system formation, methods applied and algorithms study of debris analysis based on stereovision is carried out. Thus the problems of stereovision applied in debris analysis are resolved mainly. Furthermore, the integrated method combine with 3D symbolic parameters is studied and discussed for the recognition of debris configuration in pilot.

本文的研究工作以技术应用为目标，分为三大部分：首先是对计算机视觉，特别是立体视觉技术机理、方法和算法的基本研究，主要解决了立体视觉在内窥探伤与铁谱检测中应用的技术共同点；其次是对基于立体视觉的发动机内窥探伤的技术研究和系统开发，解决了立体视觉用于内窥图像这一宏观图像的技术和算法实现问题，其中包括对内窥场景的视觉重建和对发动机内部典型损伤的分析；最后对基于立体视觉的磨粒分析技术的系统构架、实施方法和相关算法进行了研究，主要解决了立体视觉用于磨粒显微图像的技术问题，并对融合三维描述参数的磨粒形态综合识别方法进行了初步研究与探讨。

The research work in this dissertation aim to apply the advanced technique in practice, that ramifies three parts: at first, the basic study of the computer vision, especially the stereo vision, include their technical mechanisms, methods and algorithms resolves the same technical problems of endoscopy and ferrograhpy based on stereo vision. Secondly, the endoscopy of engine fault inspection based on stereovision is researched and the system faced to this application is developed, thus the problems of stereovision applied to macroscopical images such as endoscopic images are resolved in method and computation. In addition, inside scenes are reconstructed and typical damages of engine inside are analyzed. Thirdly, the investigations of system formation, methods applied and algorithms study of debris analysis based on stereovision is carried out. Thus the problems of stereovision applied in debris analysis are resolved mainly. Furthermore, the integrated method combine with 3D symbolic parameters is studied an d discussed for the recognition of debris configuration in pilot.

本文的研究工作以技术应用为目标，分为三大部分：首先是对计算机视觉，特别是立体视觉技术机理、方法和算法的基本研究，主要解决了立体视觉在内窥探伤与铁谱检测中应用的技术共同点；其次是对基于立体视觉的发动机内窥探伤的技术研究和系统开发，解决了立体视觉用于内窥图像这一宏观图像的技术和算法实现问题，其中包括对内窥场景的视觉重建和对发动机内部典型损伤的分析；最后对基于立体视觉的磨粒分析技术的系统构架、实施方法和相关算法进行了研究，主要解决了立体视觉用于磨粒显微图像的技术问题，并对融合三维描述参数的磨粒形态综合识别方法进行了初步研究与探讨。

This method reshapes the SAR images matrix into vectors, and analyzes the matrix which is composed of the two vectors with PCA process, getting a conclusion that the minor component image is the change portion.

该算法将SAR图像转化为列向量，对两个图像向量组成的矩阵进行主分量分解，其图像的变化部分则表征为矩阵的次分量成分。

With the help of localizer, the magnified size can be identified and the image can be restituted correctly with no photogrammetric parameters.

面部定位器还可以确定定位摄影图像的放大倍率，准确进行图像复原，无须记录摄影参数，其图像复原的变异系数CV＜0.03。

It's application supplies a new means for image process.Key words: fractal theory;medical image;edge sharping and detecting

结果表明，经处理过的图像与原图像相比较，不同结构的边界有了明显的增强，特别是对图像中的肿瘤结构能更好的识别出来，该方法的应用为医学图象处理和分析领域提供了新的手段和方法。

Keywords: Virtual reality;Panoramic image;Edge sharping;Image matching

基于图像的绘制技术是虚拟现实中建立虚拟环境的重要方法，而全景图像是基于图像绘制的关键。

We use sharping processing, binary processing, smoothing processing to separate the face from the background, and then we use Integration projection to realizes the precise location of the face.

在人脸的自动识别算法中，本系统对定位后的人脸图像做了几何标准化和灰度标准化操作，采用KL变换进行降维处理，提取出人脸图像矢量，与样本库中的特征样本进行匹配，选取最为匹配的人脸图像作为输出。

This model can measure statistical differences between original image and spliced image.

使用图像质量评价量和统计特征量来建立模型，以得到原始图像和拼接图像之间的统计差异。

In order to have the thermal image sight with the same functions of low level sight sight as the tank does,a realtime image processing system that superposes sight graduation image and thermal image, based on DSP high speed image processing function is offered to realize the goal.

为了实现热成像瞄准镜同样具有坦克微光瞄准镜所具有的功能，本文利用数字信号处理器DSP高速图像处理功能实时地将热成像图像和瞄准镜分划图像进行叠加处理，从而解决了上述问题。

But we don't care about Battlegrounds.

但我们并不在乎沙场中的显露。

Ah! don't mention it, the butcher's shop is a horror.

啊！不用提了。提到肉，真是糟透了。