关键词: 图像分割, 人工蜂群, 多阈值, 叶片, 边缘检测

Abstract: Concerning the problem that the segmentation of plant leaf images in natural scenes is not effective enough, it is difficult to extract the complete leaf region from the images containing many leaves. Aiming at this problem, a fast multi-threshold extraction method of leaf region is proposed. Firstly, the artificial bee colony algorithm is used to optimize the Otsu multi-threshold selection process. The optimal multi-threshold is obtained by using the between-class variance as the fitness function. The suitable number of segmentation thresholds for the leaf image is determined adaptively in a iterative manner in the process of obtaining the optimal multi-threshold. Then the edge detection, logical operation and morphological methods are used to remove the remaining background elements and extract the leaf region from the results of the multi-threshold segmentation. The experimental results show that the method can obtain a more complete and accurate leaf region when the plant leaf image with one or more leaves are operated.

Key words: image segmentation, artificial bee colony, multi-threshold, leaves, edge detection