《无损探伤检测中的二维直方图分割算法研究》》-毕业论文.doc

PAGE 精品 摘 要 本文主要是针对在无损检测中得到的荧光磁粉裂纹图像进行处理，通过MATLAB软件来实现图像处理的各种算法，首先将荧光磁粉裂纹真彩图像转化为灰度图像，然后对灰度图像分别采用了基于一维灰度直方图的最大类间方差法，基于二维灰度直方图的最大类间方差法和改进的基于二维灰度直方图的最大类间方差法，求取三组最佳阈值，利用得到的阈值分别对荧光磁粉裂纹图片进行分割，然后根据它们的结果来比较这三种方法的优劣。这些算法主要是基于一维直方图原理和二维直方图原理并结合最大类间方差法（Otsu算法）实现的。二维直方图可以反映出像素与邻域的空间相关信息，大大提高了分割的准确性和抗噪能力。结果表明基于二维灰度直方图的分割算法明显优于基于一维灰度直方图的分割算法，可以更好的把裂纹从工件背景中分割出来。 关键词：荧光磁粉探伤，Otsu法，一维灰度直方图，二维灰度直方图，阈值分割 Abstract This paper aims at the procession of a damaged image in non-destructive testing and fluorescent magnetic particle inspection,and realizes algorithms by means of MATLAB.First,we need convert true colour images into gray images,then we can use three different kinds of methods to process the picture we got,the two of which are respectively based on one-dimensional histogram and two-dimensional histogram.The.last.one.is.an.advanced.algorithm.which.is.based.on.two-dimensional histogram also.Three thresholds are gained by which we can respectively partition the picture we got .Finally these partitioned pictures are compared to decide which method is the best.These algorithms are based on the principle of one-dimensional histogram and the principle of two-dimensional histogram combined with otsu method.Two-dimensional histogram can show the relation of a.pixel.with.pixel.nearby,which.can.materially.increases.the.acuracy.of.partitioning a image and the ability of anti-noise.The result of a experiment show that an algorithm based on one-dimensional histogram is superior to an algorithm based on two-dimensional histogram,the former can partition an image better. Key word: fluorescent magnetic particle inspection,Otsu, One-dimensional histogram, two-dimensional histogram, thresholding 目 录 TOC \o 1-3 \h \z \u HYPERLINK \l _Toc231897350 第一章 绪论 PAGEREF _Toc231897350 \h 1 HYPERLINK \l _Toc231897351 1.1 引言 PAGEREF _Toc231897351 \h 1 HYPERLINK \l _Toc231897352 1.2荧光磁粉探伤 PAGEREF _Toc231897352 \h 3 HYPERLINK \l _Toc231897353 1.3论文的主要研究内容 PAGEREF _Toc231897353 \h 5 HYPERL