【作者】 马艳；

【导师】 张治辉；

【作者基本信息】 浙江工业大学 ， 控制理论与控制工程， 2004， 硕士

【摘要】 计算机视觉检测理论的发展，给自动测量带来新的解决方法。边缘检测技术是计算机自动检测技术的一个重要方面。图象边缘包含了图象的绝大部分有用信息，所以边缘检测在模式识别、计算机视觉、图象分析和图象编码等领域都有重要的研究价值。本文对国内外边缘检测方法的现状和发展进行了综述。 边缘检测有很多方法，根据课题实际背景，选择了几种比较经典的边缘检测方法，即一阶算子(Roberts算子、Sobel算子、Prewitt算子)、二阶算子(Laplacian算子和Kirsch算子)、LOG算子和Canny算子，对他们进行理论和实际应用研究。 近几年来迅速发展起来的小波分析，由于其具有时频局部变换的特性及多尺度性，能敏感地反映突变信号及良好的去噪效果，因此能准确、有效地确定图象边缘位置，因此是一种理想的边缘检测方法。本文用B—样条双正交小波构造滤波器组，探索基于多尺度分析的小波变换边缘检测方法的应用。 本文设计了在Visual C++6.0平台下的图象边缘检测软件。对于输入系统的图象首先进行预处理：灰度转换和滤波去噪；并用几种经典的边缘检测方法对PVC型材产品图象进行边缘检测。用多尺度小波变换的方法在Matlab中对PVC型材产品进行边缘检测，并把检测结果与前面几种经典的边缘检测方法进行比较，认为多尺度小波变换边缘检测是一种效果较好的检测方法。最后对检测的结果进行定性分析。更多还原

【Abstract】 Computer visual inspection is rising and advancing. It brings a new way to auto-measure. And edge detection is one of great important parts in computer auto visual inspection. Image edge contains much useful information, and it is very important to pattern recognition, computer vision, image analysis and image code. This thesis summarized the current state and development of edge detection in china and abroad.There are many methods of edge detection. According to practical task, we choose some classical methods of edge detection including first differential(Roberts operator, Sobel operator and Prewitt operator), second differential(Laplacian operator and Kirsch operator), LOG filter and Canny operator. They are studied in theory and practice applying.Wavelet transform is a kind of time-frequency analysis method in signal processing and is becoming more and more popular during these years. Its multi-scale property, localization both in time and frequency domain, sensitivity to the abrupt change of signals, and good to denoising make it an ideal way for edge detection. We use B-spline biorthogonal wavelet to form filter groups, and explore application of edge detection based on multi-scale wavelet transform.A analytical and processing software is designed under the environment of Visual C++6.0. Firstly, color intensity is transformed to gray intensity and image is filtered out some noise. Secondly, we use some above classical methods of edge detection. And we use multi-scale wavelet transform to measure the edge of PVC product in Matlab. After comparing these methods, we draw a conclusion: multi-scale wavelet transform is a better method in detecting edge of PVC product. Finally, we make the qualitative analysis of edge detection.更多还原