| 摘要: |
| 【目的】研究枝条遮挡情况下单个苹果目标的识别，为复杂生长环境下苹果目标的准确识别与定位及实现自动采摘提供支持。【方法】针对果实目标受枝条遮挡影响被分割成几个部分，从而严重影响果实目标准确识别的现状，以枝条遮挡下的苹果目标为研究对象，选用基于Lab颜色空间的K-means聚类算法对苹果目标进行分割，再通过数学形态学方法对目标苹果轮廓进行提取，然后根据最小外接矩形法去除目标苹果的伪轮廓，最后利用轮廓的曲率特征对目标苹果进行重建，并对分割与重建结果进行了方法验证。【结果】利用基于Lab颜色空间下的K-means聚类算法和最小外接矩形法可有效提取出苹果目标的真实轮廓，能够与苹果目标边缘线达到高度重合，同时可获得较准确的重建结果。对10幅枝条遮挡果实目标的识别、定位与重建的验证结果表明，该方法对目标苹果进行分割与重建的平均分割误差为13.83%，平均重叠系数为88.08%，假阳性率和假阴性率分别为1.22%和11.92%，目标苹果重建准确率均在84.00%以上，平均重建时间为24.40 s。【结论】应用本研究中的方法可对枝条遮挡下的苹果目标进行准确识别、定位与重建，有效缩短重建时间。 |
| 关键词: 苹果 自动采摘 枝条遮挡 识别与定位 K-means聚类算法 轮廓曲率 |
| DOI： |
| 分类号: |
| 基金项目:国家高技术研究发展计划(863计划)项目(2013AA10230402)；陕西省农业科技创新与攻关项目(2016NY-157)；中央高校基本科研业务费项目(2452016077) |
|
| Recognition and reconstruction of single apple occluded by branches |
|
SUN Sashuang,WU Qian,TAN Jianchang,et al
|
| Abstract: |
| 【Objective】The identification of a single apple target occluded by branches was studied to provide support for the identification,localization and automatic picking of apples in complex growing environments.【Method】The fruit targets occluded by branches are often divided into several parts,which affects the accurate identification of fruit targets seriously.In this study,the K-means clustering algorithm based on Lab color space was used to segment apple targets,then the contours were extracted by mathematical morphology method.After the false contours of target apple were removed according to the minimum external rectangle method,the targets were reconstructed by the curvature features of the contours,and the results were verified.【Result】The above method could effectively extract the real contours of apple targets and achieve high coincidence with the edge lines of targets based on K-means clustering algorithm in Lab color space and minimum external rectangle method.More accurate reconstruction results were also obtained.A total of 10 experimental target fruit images occluded by branches were tested for identifying,locating and reconstructing.The average division error was 13.83%,the average overlapping coefficient was 88.08%,and the false positive and false negative rates were 1.22% and 11.92%,respectively.The reconstruction accuracy rate of target apples was above 84.00%,and the average construction time was 24.40 s.【Conclusion】The methods developed in this study could be used for accurate identification,location and reconstruction of apples occluded by branches,which could shorten the time of reconstruction effectively. |
| Key words: apple automatic picking branch occlusion identification and location K-means clustering algorithm contour curvature |
