复杂背景下目标树叶的分割与补缺

Segmentation and replenishment of the target leaves in complex background

为了进一步提高复杂背景下目标树叶分割算法的运行效率和准确性,在超绿算法的基础上结合模糊C聚类算法提高分割的运行效率,利用凸包填充法提高分割的准确性。首先利用超绿算法去除复杂背景中的非绿色部分,然后利用模糊C聚类算法去除与目标树叶颜色差异较大的绿色背景,而对于颜色差异较小的绿色背景,则先利用底帽变换得到目标树叶的边缘信息,再通过腐蚀和比较连通区域的大小进行去除。采用凸包填充法对目标树叶的边缘缺口进行补缺可以提高分割的准确性,从而降低错分率。超绿算法结合模糊C聚类算法可提高分割的运行效率。研究结果表明,与原有算法对比,错分率平均降低了1.05%,分割效率平均提高了13.82%。

In order to furtherly improve the operation efficiency and accuracy of the segmentation algorithm of target leaves in complex background,fuzzy C-cluster algorithm was adopted to improve the operation efficiency of the segmentation on the basis of the extra-green algorithm,and the convex hull filling method was used to improve the accuracy of the segmentation. First of all,the non-green part of the complex background was removed by using the extra-green algorithm. The green background which was largely different from the color of the target leaves was removed by using fuzzy C-cluster algorithm. While for the green background with smaller color difference,the edge information of the target leaves was obtained by using bottom-hat transformation,and then the area was removed by corrosion and the comparison of the connected regions. The edge gap of the target leaves was replenished to improve the segmentation accuracy by convex hull filling,reducing the rate of error rates. In addition,the operation efficiency of segmentation could be increased by extra-green algorithm combined with fuzzy C-cluster algorithm. The experimental results showed that the rate of error rate was by an average of 1.05%,and compared with the original algorithm,the segmentation efficiency was increased by 13.82% averagely.