基于边缘流的煤电厂炉膛火焰图像分割方法

Image Segmentation for Furnace Flame Based on Edge Flow

针对炉膛火焰灰度变化不明显采用传统的分割算法无法达到满意分割效果的问题,采用了基于边缘流的分割方法对电厂炉膛火焰图像进行分割;首先选用高精度的灰度边缘流对图像进行分割,得到边缘流的矢量方向和边缘能量,然后再对图像闭合处理得到初始分割图像;最后,再利用颜色相似度对相邻区域进行处理,得到最终分割图像,该方法不仅避免了单纯边缘流算法提取边界不连续无法检出较弱边界的问题,而且也可以避免过分割问题;通过实验表明,基于边缘流的方法是对炉膛火焰图像进行分割的有效方法。

In view of the problem of over segmentation to the underground objects of the traditional edge flow algorithm,the paper proposes a segmentation method of combining edge flow and fuzzy clustering based on genetic algorithm.Firstly,the energy and direction of edge flow is acquired by using the edge flow algorithm on the initial underground image,Then the initial segmentation image is got though closed treatment.Finally,the final segmentation image is obtained by the color similarity of adjacent areas.Experiments show that the method has good segmentation effect in the underground objects image segmentation.