摘要
为提高遥感影像棉花识别的精度,提出一种基于反向传播(back propagation,BP)神经网络算法的棉花识别方法。利用单时相GF-1号和Ladsat8遥感数据,结合归一化植被指数(NDVI)、差值植被指数(DVI)、比值植被指数(RVI)、红波段亮度值(B3)和近红外波段亮度值(B4)等特征指数,依据野外GPS实测数据选择训练样本,通过不同的特征组合对BP神经网络进行训练。验证结果表明,该识别方法精度达到98.32%,较最大似然法和最小距离法分别提高8.27%和5.53%。实验结果表明,所提方法能够有效地提高棉花识别精度并简化识别过程。
To improve the accuracy of recognizing cotton using remote sensing data, a method was put forward for recognizing cotton based on BP (back propagation) neural network algorithm. Single-temporal GF-1 and Landsat8 remote sensing data wereused, and normalized difference vegetation index (NDVI) , difference vegetation index (DVI), ratio vegetation index (RVI) , red band brightness value and near-infrared band brightness value were combined, according to the field of GPS measurement data, the training sample was chosen, and the BP neural network was trained by different combination of features. It is verified that the recognition accuracy of cotton based on BP neural network reaches 98. 32%, which is 8. 27% and 5. 53% higher than that based on maximum likelihood and minimum distance respectively. Experimental results show that the proposed method can effec-tively improve the identification accuracy and simplify the identification process.
作者
范迎迎
钱育蓉
杨柳
黄震
FAN Ying-ying QIAN Yu-rong YANG Liu HUANG Zhen(School of Software,Xinjiang University,Urumqi 830008, Chin)
出处
《计算机工程与设计》
北大核心
2017年第5期1356-1360,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61562086
61462079
61363083
61262088)
关键词
遥感影像
棉花识别
BP神经网络
植被指数
GF-1
remote sensing image
cotton recognition
BP neural network
vegetation index
GF-1