中文摘要：

为解决荔枝收获机器人对采摘目标的识别与定位的关键问题,荔枝串、荔枝果与结果母枝各部位的识别成为采摘机器人视觉系统的研究重点.首先,对采集荔枝图像的感兴趣区域进行荔枝串、荔枝果与结果母枝的图像分类并给出了荔枝各部位分类识别的图像分割策略;然后,采用探索性实验分析的方法获取了识别各类光照条件与各类颜色的荔枝串的固定阈值为0.502 0,0.530 0,0.510 8及0.533 2、0.501 0、0.520 8,并利用所获取固定阈值进行荔枝串的识别;最后,以不同光照条件的荔枝图像90幅进行荔枝各部位分类识别的实验验证,荔枝串与荔枝果的平均识别率为91.67％,结果母枝的平均识别率为86.67％.实验结果表明：基于探索性分析方法识别荔枝各部位图像分割策略与识别各类荔枝串的固定阈值的设定是有效、可行的.

英文摘要：

To solve the problems on recognition and positioning of picking object for litchi harvesting-robot, recognizing litchi cluster, litchi fruits and their main fruit bearing branch become key points so that the research focuses on vision system of litchi picking robots. Firstly, image sorts on litchi cluster, litchi fruits and their main fruit bearing branch in the interesting region were given, and segmentation strategy based on exploratory analysis for recognizing all parts of litchi image was then proposed. Secondly, fixed thresholds for segmenting all sorts of litchi clusters by method of exploratory analysis were obtained, they are 0. 5020, 0. 5300,0. 5108 to segment litchi clusters-influencing illumination with highlight, normal light and backlighting, respectively, and 0. 5332,0. 5010,0. 5208 for litchi clusters with main fruit bearing branch in partial red, partial green and partial brown. By these thresholds, recognition on litchi cluster was carried out successfully. Finally, taking 90 differently illuminated litchi images as test object, this robot effectively recognized ripe litchi clusters with recognition ratio 91.67%. After that, litchi fruits and the main fruit bearing branch were successfully extracted from the recognized litchi cluster by segmenting the recognized images in fuzzy c-means clustering （with respective recognition ratio 91.67% and 86.67% ）. The result of the testing experiment also attested that the given strategy for recognizing each part of litchi and the obtained fixed threshold for recognizing all kinds of litchi clusters based on exploratory were feasible and effective.