【作者】 曾俊；

【导师】 李德华；

【作者基本信息】 华中科技大学 ， 控制科学与工程， 2011， 博士

【摘要】 数字图像处理技术的应用范围越来越广,渗透到社会的各个领域。图像的边缘特征是图像的重要特征。图像边缘检测技术是数字图像处理、计算机视觉、模式识别的基础。数字图像边缘检测技术广泛应用于图像分割、运动检测、目标跟踪、人脸识别等领域。因此图像边缘检测技术是图像处理技术的研究热点之一,提高边缘检测精度和探索边缘检测技术在实际工程中的应用是边缘检测技术的重要研究内容。本论文的研究工作是结合项目“多谱图像配准技术研究”、“货车故障动态图像检测系统中故障自动检测”和“基于双目立体视觉的动作捕捉系统”,对数字图像边缘检测技术的基础理论展开了研究,并将数字图像边缘检测技术应用于上述项目中。本论文的主要研究内容主要包括以下几个方面：首先,本论文对国内外关于数字图像边缘检测技术的研究成果和现状进行了系统总结,阐述了经典的灰度图像边缘检测方法和彩色图像边缘检测方法。介绍了图像边缘检测技术在图像处理、模式识别、计算机视觉中的应用,展望了数字图像边缘检测技术的发展趋势。其次,详细介绍了经典的灰度图像Canny边缘检测方法,分析了此方法存在的缺陷,提出了一种基于直方图凹度分析的灰度Canny边缘检测方法。改进的Canny边缘检测方法在高斯滤波之前用开关型中值滤波器滤除脉冲噪声,利用直方图凹度分析来自动选取双阈值。改进的方法能有效地滤除图像中的脉冲噪声,自动地选取双阈值。为了将灰度图像SUSAN边缘检测算子扩展到了彩色图像。我们提出了用于彩色图像边缘检测的SUSAN边缘检测方法。首先将待检测的彩色图像从RGB空间转换到CIELAB空间,然后利用利用色差来计算核值相似区域的大小,最后通过阈值化来提取彩色图像的边缘。与传统的边缘检测方法相比,改进的新方法能有效地检测出彩色图像的边缘。检测出的彩色图像边缘比较符合人眼视觉特性。此外,本论文回顾了偏微分理论及其在彩色图像滤波中的应用,提出了一种基于快速向量全变分和Sobel色差算子的Canny彩色图像边缘检测方法。改进的彩色图像边缘检测方法利用快速向量全变分方法滤除噪声,同时能保留图像边缘细节信息,然后利用色差Sobel算子计算彩色图像的色差和方向,用于非极大值抑制,最后通过双阈值的方法来提取彩色图像的边缘,这种方法能有效地检测出图像的边缘。针对彩色图像中的脉冲噪声,设计了一种基于CIELAB空间抗脉冲噪声的彩色Canny边缘检测算子。最后,介绍了货运列车故障动态图像检测系统的组成及工作原理,并对货运列车故障动态图像检测系统中的常见故障,给出了具体的分类和分析。本文将图像边缘检测技术用于列车故障图像检测系统,提出了一种基于边缘检测定位和匹配的心盘螺栓丢失故障自动检测方法,该算法通过边缘检测提取图像中心盘螺栓所在的感兴趣区域,然后通过特征点检测来定位故障可能出现的位置,最后通过边缘匹配来实现候选故障验证。实验结果表明本方法能快速地检测出心盘螺栓丢失位置。更多还原

【Abstract】 Digital image processing is increasingly being used for a wide range of applications and infiltrating to all areas of society. The edge feature of digital image is an important feature of the image. Digital image edge detection technology is the basis of image processing, computer vision, pattern recognition, which is widely used in image segmentation, motion detection, object tracking, face recognition and other fields. Thus, image edge detection is one of the hot research areas in image processing. Improving the accuracy of edge detection and exploring its practical applications are the important research contents of edge detection. The study is combined with three projects, "Research on Multi-spectral Image Registration"、"Automatic Fault Detection in Trouble of Moving Freight Car Detection System" and "Human Motion Capture System Based on Stereo Vision". The basic theory of edge detection method is researched in this paper. Edge detection technologies are applied in above three projects. The content studied of this thesis mainly includes the following several respects:Firstly, we systematically summarize the research achievements and the curruent situation of research related to image edge detection technology at home and abroad, and categorizes the edge detection technology and describes the classic gray image and color image edge detection methods. This paper introduces the applications of edge detection technology in the field of image processing, pattern recognition and computer vision, and prospects the development trend of image edge detection technology.Next, the traditional Canny edge detector is introduced in detail, the drawbacks on Canny detector are analysised in this thesis. An adaptive Canny edge detector based on histogram concavity analysis is proposed, it uses switching median filter to remove the impulsive noise before Gaussian filtering, and selects the optimal dual-threshold through histogram concavity analysis. The improved detector can efficiently remove the impulsive noise and automatically select the optimal dual-threshold.Then, in order to this extend the gray SUSAN edge detection method to the color image. A method of SUSAN edge detection method is presented to detect the edges of color image. This new method first convert the the RGB image into CIELab color space, and then use the color difference to calculate the size of Univalue Segment Assimilating Nucleus (USAN) area. The final edges are extracted by thresholding. This new method can effectively detect edges of color images compared with traditional methods, the detected edges more consistent with the human visual characteristics.Moreover, the theory of partial differential equation and its application in image processing are reviewed first, an improved Canny color edge detection method based on fast vectorial total variation (VTV) minimization denoising model and color difference Sobel operator is proposed, it uses fast vectorial total variation (VTV) minimization model to remove noise while preserving the image edges. Then calculates the color difference and direction in CIELAB color space, which is used for non-maximal suppression. Finally, the improved method extracts the edges by the double-threshold method. This detector can efficently detects the edges of color image. As to impulsive noise in the color image, we designed an improved Canny edge detector against impulsive noise based on CIELAB space.Finally, this thesis describes the basic principle and composition of the trouble of moving freight car detection system. According to several common faults of the TFDS system, it offers the classifying, analyzing in detail. Image edge detection technology is applied successively to the trouble of moving freight car detection system. An automatic detection method for loss of freight car center plate bolts based on edge location and match is presented.The algorithm firstly uses sobel edge detection to extract the ROI (region of interest) of center plate bolts.Then locates all possible faults in the ROI using feature point detection and verify the candidate fault by edge matching.The experimental results show that:The proposed algorithm can rapidly detect the trouble of freight car center plate bolts.更多还原