期刊文献+
共找到2篇文章
< 1 >
每页显示 20 50 100
Drivers,Trends,and Patterns of Changing Vegetation-greenness in Nansha Islands,China from 2016 to 2022
1
作者 tang jiasheng FU Dongjie +2 位作者 SU Fenzhen YU Hao WANG Xinhui 《Chinese Geographical Science》 SCIE CSCD 2024年第4期662-673,共12页
Changes in vegetation status generally also represents changes in the ecological health of islands and reefs(IRs).However,studies are limited of drivers and trends of vegetation change of Nansha Islands,China and how ... Changes in vegetation status generally also represents changes in the ecological health of islands and reefs(IRs).However,studies are limited of drivers and trends of vegetation change of Nansha Islands,China and how they relate to climate change and human activities.To resolve this limitation,we studied changes to the Normalized Difference Vegetation Index(NDVI)vegetation-greenness index for 22 IRs of Nansha Islands during normal and extreme conditions.Trends of vegetation greenness were analyzed using Sen's slope and Mann-Kendall test at two spatial scales(pixel and island),and driving factor analyses were performed by time-lagged partial correlation analyses.These were related to impacts from human activities and climatic factors under normal(temperature,precipitation,radiation,and Normalized Difference Built-up Index(NDBI))and extreme conditions(wind speed and latitude of IRs)from 2016 to 2022.Results showed:1)among the 22 IRs,NDVI increased/decreased significantly in 15/4 IRs,respectively.Huayang Reef had the highest NDVI change-rate(0.48%/mon),and Zhongye Island had the lowest(–0.29%/mon).Local spatial patterns were in one of two forms:dotted-form,and degradation in banded-form.2)Under normal conditions,human activities(characterized by NDBI)had higher impacts on vegetation-greenness than other factors.3)Under extreme conditions,wind speed(R^(2)=0.2337,P<0.05)and latitude(R^(2)=0.2769,P<0.05)provided limited explanation for changes from typhoon events.Our results provide scientific support for the sustainable development of Nansha Islands and the United Nations‘Ocean Decade’initiative. 展开更多
关键词 island and reefs(IRs) Normalized Difference Vegetation Index(NDVI) vegetation-greenness change-rate Sen's slope Nansha Islands China
下载PDF
基于改进卷积神经网络的在体青皮核桃检测方法 被引量:12
2
作者 樊湘鹏 许燕 +3 位作者 周建平 刘新德 汤嘉盛 魏禹同 《农业机械学报》 EI CAS CSCD 北大核心 2021年第9期149-155,114,共8页
采摘机器人对核桃采摘时,需准确检测到在体核桃目标。为实现自然环境下青皮核桃的精准识别,研究了基于改进卷积神经网络的青皮核桃检测方法。以预训练的VGG16网络结构作为模型的特征提取器,在Faster R-CNN的卷积层加入批归一化处理、利... 采摘机器人对核桃采摘时,需准确检测到在体核桃目标。为实现自然环境下青皮核桃的精准识别,研究了基于改进卷积神经网络的青皮核桃检测方法。以预训练的VGG16网络结构作为模型的特征提取器,在Faster R-CNN的卷积层加入批归一化处理、利用双线性插值法改进RPN结构和构建混合损失函数等方式改进模型的适应性,分别采用SGD和Adam优化算法训练模型,并与未改进的Faster R-CNN对比。以精度、召回率和F1值作为模型的准确性指标,单幅图像平均检测时间作为速度性能评价指标。结果表明,利用Adam优化器训练得到的模型更稳定,精度高达97.71%,召回率为94.58%,F1值为96.12%,单幅图像检测耗时为0.227 s。与未改进的Faster R-CNN模型相比,精度提高了5.04个百分点,召回率提高了4.65个百分点,F1值提升了4.84个百分点,单幅图像检测耗时降低了0.148 s。在园林环境下,所提方法的成功率可达91.25%,并且能保持一定的实时性。该方法在核桃识别检测中能够保持较高的精度、较快的速度和较强的鲁棒性,能够为机器人快速长时间在复杂环境下识别并采摘核桃提供技术支撑。 展开更多
关键词 青皮核桃 采摘机器人 目标检测 卷积神经网络 改进Faster R-CNN
下载PDF
上一页 1 下一页 到第
使用帮助 返回顶部