基于数学形态学的相接触草莓果实的分割方法及比较研究  

Comparison of two algorithms based on mathematical morphology for segmentation of touching strawberry fruits

作　　者：周天娟[1] 张铁中[1] 杨丽[1] 赵金英 

机构地区：[1]中国农业大学工学院,北京100083 [2]现代农装科技股份有限公司,北京100083

出　　处：《农业工程学报》2007年第9期164-168,共5页Transactions of the Chinese Society of Agricultural Engineering

摘　　要：多个草莓在田间经常相互接触。为了便于机器人的自动化采摘,需将相接触的多个成熟草莓果实分开,并给出各个草莓的重心坐标。基于数学形态学的方法,研究了两种针对这种较复杂情况的成熟草莓果实分割的方法,即聚类快速分割法和分水岭区域分割法。首先对成熟草莓果实和背景使用BP神经网络方法进行分割,然后进行灰度化、二值化、孔洞填充等初步处理,最后分别利用聚类快速分割法和分水岭区域分割法分割相接触成熟草莓果实图像。两种分割方法可分别得到两个果实区域,通过对这两个区域计算重心即可为机器人采摘提供重心数据。结果表明,两种分割方法都能将相接触区域分开,各有优缺点和适用性。More than two strawberry fruits always contact with each other in the natural condition. The touching mature strawberry fruits have to be separated and their centers of gravity have to be found in order to be picked automatically by robot. Based on the mathematical morphological algorithm, two methods to solve this complexity were proposed, namely, Clustering Fast Segmentation and Watershed Region Segmentation. First, BP neural network algorithm was used to separate the strawberry fruits from the background. Second, several initial processes were applied to prepare for the following operation, including graying, threshold and holes filling. In the end, the close fruits were separated by these two methods, and then the gravity centers of the fruits were calculated. According to the comparison and analysis of the results of the two methods, it is proved that both methods can successfully separate the touching fruits and both of them have their own advantages and disadvantages, and they can be applied in different areas.

关 键 词：草莓 图像分割 数学形态学 聚类快速分割法 分水岭区域分割法 

分 类 号：S233.4[农业科学—农业机械化工程] TP391.4[农业科学—农业工程]