论文部分内容阅读
针对铁路货车铸件DR(Digital Radiography)图像数据量大、背景亮度不均匀以及对DR图像中铸件薄壁区域的缺陷提取效果不理想和基于C-V模型的分割方法存在计算量大的问题,研究基于DR图像多重快速分割的铸件缺陷提取方法。先利用OTSU双阈值分割算法对铸件DR图像进行分割,然后应用区域生长技术得到铸件薄壁区域的图像;对该图像进行剥皮和归一化线性灰度拉伸处理后,再次利用OTSU双阈值分割算法和区域生长技术对图像进行分割,从中提取出含有和可能含有铸件缺陷的局部图像;对可能含有铸件缺陷的亮背景区域图像用C-V模型再次进行分割,最终提取出所有铸件缺陷的图像。
In order to solve the problem of large amount of image data, uneven background brightness and defects in the thin wall of castings in DR images, and the large amount of calculation based on the CV model, Image Fast and Multiple Segmentation Method for Casting Defect Extraction. Firstly, the OTSU dual threshold segmentation algorithm is used to segment the DR images of castings, and then the image of the thin-walled region of the castings is obtained by applying the region growing technique. After the image is peeled off and normalized to the linear grayscale stretching, OTSU dual threshold segmentation Algorithm and region growing technology to segment the image and extract the partial images which contain and may contain the defects of the castings. The images of the bright background regions which may contain the casting defects are re-divided by the CV model and finally the images of all the casting defects are extracted.