活立木图像分割算法研究

RESEARCH ON THE METHOD OF SEGMENTATION ON COLOR IMAGE OF STANDING TREES

为了更好地避免图像过分割的缺陷,提高机器识别的准确度,提出一种基于分水岭算法的活立木图像分割算法。通过分析分水岭算法在实际分割中存在缺陷的原因,使用前处理与后处理相结合的方法克服过分割的缺陷。使用形态学重建技术进行图像前处理;使用分水岭算法对图像进行分割,并且利用标记对象法对图像进行标记;利用目标对象的色彩信息进行区域合并以及去除多余水坝,最终得到分割图像。针对10幅活立木图像,通过与人工分割的方法进行比较,所提出的基于分水岭算法的活立木图像分割算法能够有效地提取活立木的目标区域,分割的正确率达到88.42%。

In order to avoid the defects of image segmentation and improve the accuracy of machine identification, a new algorithm of segmentation image based on watershed algorithm is proposed. By analyzing the causes of the defect in the actual segmentation, the method of combining pre-treatment and post-processing was used to overcome the shortcomings of over-division. Firstly, the image pre-processing was carried out by using the morphological reconstruction technique. Then the watershed algorithm was used to segment the image, and the image was marked by the marker object method. Finally, the color information of the target object was used to merge the area and removed the redundant dam, and finally the segmented image was obtained. According to the method of artificial segmentation, the proposed algorithm based on watershed algorithm can effectively extracted the target area of living wood, and the correct rate of segmentation was 88.42%.