基于图像分割的船舶涂装轮廓检测方法

Contour Detection Method for Ship Coatings Based on Image Segmentation

在船舶涂装轮廓检测中,检测到的轮廓点必须是连续的,轮廓的数量必须是既定的。目前常见的轮廓检测方法一般基于图像灰度的变化来判断图像中的点是否属于轮廓点,使用这些方法检测到的点可能不连续,包含大量非轮廓点,轮廓的数量也是不能设定的。针对这个问题,设计了一种基于图像分割的轮廓检测方法,首先对图像预处理,然后使用多阈值Otsu方法分割图像,从图像中提取各层涂装,生成对应的二值图像并进行形态学滤波,最后提取并筛选二值图像中的轮廓。实验证明,该方法不仅满足船舶涂装轮廓检测的要求,而且操作简单,计算效率高。

The process of contour detection for ship coatings should yield a specified number of contours and consecutive points in each contour. Common contour detection methods judge whether a image point belongs to a con- tour based on the change of image intensity, which may generate an undesired number of contours with inconsecutive or non-contour points in it. Focused on this issue, a contour detection method based on image segmentation was de- signed, during which the input image was firstly preprocessed, then segmented by multi-threshold Otsu method, each coating area was extracted to a binary image which was then filtered by a morphological filter, at last, contours from these binary images were extracted and selected. Experiments show that this method not only meets the requirement of ship coatings contour detection, but also is efficient and easy to operate.