采摘点的识别与定位是智能采摘的关键技术,也是实现高效、适时、无损采摘的重要保证。针对复杂背景下番茄串采摘点识别定位问题,提出基于RGB-D信息融合和目标检测的番茄串采摘点识别定位方法。通过YOLOv4目标检测算法和番茄串与对应果...采摘点的识别与定位是智能采摘的关键技术,也是实现高效、适时、无损采摘的重要保证。针对复杂背景下番茄串采摘点识别定位问题,提出基于RGB-D信息融合和目标检测的番茄串采摘点识别定位方法。通过YOLOv4目标检测算法和番茄串与对应果梗的连通关系,快速识别番茄串和可采摘果梗的感兴趣区域(Region of Interest,ROI);融合RGB-D图像中的深度信息和颜色特征识别采摘点,通过深度分割算法、形态学操作、K-means聚类算法和细化算法提取果梗图像,得到采摘点的图像坐标;匹配果梗深度图和彩色图信息,得到采摘点在相机坐标系下的精确坐标;引导机器人完成采摘任务。研究和大量现场试验结果表明,该方法可在复杂近色背景下,实现番茄串采摘点识别定位,单帧图像平均识别时间为54 ms,采摘点识别成功率为93.83%,采摘点深度误差±3 mm,满足自动采摘实时性要求。展开更多
Taihu Lake region is one of the most industrialized areas in China, and the surface water is progressively susceptible to anthropogenic pollution. The physicochemical parameters of surface water quality were determine...Taihu Lake region is one of the most industrialized areas in China, and the surface water is progressively susceptible to anthropogenic pollution. The physicochemical parameters of surface water quality were determined at 20 sampling sites in Taihu Lake region, China in spring, summer, autumn, and winter seasons of 2005-2006 to assess the effect of human activities on the surface water quality. Principal component analysis (PCA) and cluster analysis (CA) were used to identify characteristics of the water quality in the studied water bodies. PCA extracted the first three principal components (PCs), explaining 80.84% of the total variance of the raw data. Especially, PC1 (38.91%) was associated with NH 4 -N, total N, soluble reactive phosphorus, and total P. PC2 (22.70%) was characterized by NO 3 -N and temperature. PC3 (19.23%) was mainly associated with pH and dissolved organic carbon. CA showed that streams were influenced by urban residential subsistence and livestock farming contributed significantly to PC1 throughout the year. The streams influenced by farmland runoff contributed most to PC2 in spring and winter compared with other streams. PC3 was affected mainly by aquiculture in spring, rural residential subsistence in summer, and livestock farming in fall and winter seasons. Further analyses showed that farmlands contributed significantly to nitrogen pollution of Taihu Lake, while urban residential subsistence and livestock farming also polluted water quality of Taihu Lake in rainy season. The results would be helpful for the authorities to take sound actions for an effective management of water quality in Taihu Lake region.展开更多
密度峰值聚类(density peaks clustering,DPC)是一种基于密度的聚类算法,该算法可以直观地确定类簇数量,识别任意形状的类簇,并且自动检测、排除异常点.然而,DPC仍存在些许不足:一方面,DPC算法仅考虑全局分布,在类簇密度差距较大的数据...密度峰值聚类(density peaks clustering,DPC)是一种基于密度的聚类算法,该算法可以直观地确定类簇数量,识别任意形状的类簇,并且自动检测、排除异常点.然而,DPC仍存在些许不足:一方面,DPC算法仅考虑全局分布,在类簇密度差距较大的数据集聚类效果较差;另一方面,DPC中点的分配策略容易导致“多米诺效应”.为此,基于代表点(representative points)与K近邻(K-nearest neighbors,KNN)提出了RKNN-DPC算法.首先,构造了K近邻密度,再引入代表点刻画样本的全局分布,提出了新的局部密度;然后,利用样本的K近邻信息,提出一种加权的K近邻分配策略以缓解“多米诺效应”;最后,在人工数据集和真实数据集上与5种聚类算法进行了对比实验,实验结果表明,所提出的RKNN-DPC可以更准确地识别类簇中心并且获得更好的聚类结果.展开更多
文摘采摘点的识别与定位是智能采摘的关键技术,也是实现高效、适时、无损采摘的重要保证。针对复杂背景下番茄串采摘点识别定位问题,提出基于RGB-D信息融合和目标检测的番茄串采摘点识别定位方法。通过YOLOv4目标检测算法和番茄串与对应果梗的连通关系,快速识别番茄串和可采摘果梗的感兴趣区域(Region of Interest,ROI);融合RGB-D图像中的深度信息和颜色特征识别采摘点,通过深度分割算法、形态学操作、K-means聚类算法和细化算法提取果梗图像,得到采摘点的图像坐标;匹配果梗深度图和彩色图信息,得到采摘点在相机坐标系下的精确坐标;引导机器人完成采摘任务。研究和大量现场试验结果表明,该方法可在复杂近色背景下,实现番茄串采摘点识别定位,单帧图像平均识别时间为54 ms,采摘点识别成功率为93.83%,采摘点深度误差±3 mm,满足自动采摘实时性要求。
基金Project supported by the Knowledge Innovation Key Project of the Chinese Academy of Sciences (No. KZCX1-YW-14-5)the National Natural Science Foundation of China (No. 30600086)
文摘Taihu Lake region is one of the most industrialized areas in China, and the surface water is progressively susceptible to anthropogenic pollution. The physicochemical parameters of surface water quality were determined at 20 sampling sites in Taihu Lake region, China in spring, summer, autumn, and winter seasons of 2005-2006 to assess the effect of human activities on the surface water quality. Principal component analysis (PCA) and cluster analysis (CA) were used to identify characteristics of the water quality in the studied water bodies. PCA extracted the first three principal components (PCs), explaining 80.84% of the total variance of the raw data. Especially, PC1 (38.91%) was associated with NH 4 -N, total N, soluble reactive phosphorus, and total P. PC2 (22.70%) was characterized by NO 3 -N and temperature. PC3 (19.23%) was mainly associated with pH and dissolved organic carbon. CA showed that streams were influenced by urban residential subsistence and livestock farming contributed significantly to PC1 throughout the year. The streams influenced by farmland runoff contributed most to PC2 in spring and winter compared with other streams. PC3 was affected mainly by aquiculture in spring, rural residential subsistence in summer, and livestock farming in fall and winter seasons. Further analyses showed that farmlands contributed significantly to nitrogen pollution of Taihu Lake, while urban residential subsistence and livestock farming also polluted water quality of Taihu Lake in rainy season. The results would be helpful for the authorities to take sound actions for an effective management of water quality in Taihu Lake region.
文摘密度峰值聚类(density peaks clustering,DPC)是一种基于密度的聚类算法,该算法可以直观地确定类簇数量,识别任意形状的类簇,并且自动检测、排除异常点.然而,DPC仍存在些许不足:一方面,DPC算法仅考虑全局分布,在类簇密度差距较大的数据集聚类效果较差;另一方面,DPC中点的分配策略容易导致“多米诺效应”.为此,基于代表点(representative points)与K近邻(K-nearest neighbors,KNN)提出了RKNN-DPC算法.首先,构造了K近邻密度,再引入代表点刻画样本的全局分布,提出了新的局部密度;然后,利用样本的K近邻信息,提出一种加权的K近邻分配策略以缓解“多米诺效应”;最后,在人工数据集和真实数据集上与5种聚类算法进行了对比实验,实验结果表明,所提出的RKNN-DPC可以更准确地识别类簇中心并且获得更好的聚类结果.
