东北大学学报(自然科学版) ›› 2013, Vol. 34 ›› Issue (12): 1792-1795.DOI: 10.12068/j.issn.1005-3026.2013.12.028

• 资源与土木工程 • 上一篇    下一篇

利用BP神经网络预测大伙房水库叶绿素a质量浓度

王琦1,孟伟2,马云峰2,胡筱敏1   

  1. (1东北大学资源与土木工程学院,辽宁沈阳110819;2沈阳航空航天大学能源与环境学院,辽宁沈阳110136)
  • 发布日期:2013-07-09
  • 通讯作者: 王琦
  • 作者简介:王琦(1965-),男,吉林梅河口人,东北大学博士研究生;胡筱敏(1958-),男,江西婺源人,东北大学教授,博士生导师.
  • 基金资助:
    水体污染控制与治理科技重大专项(2012ZX07202-004-02).

Prediction of Chlorophylla Concentration in Dahuofang Reservoir Based on BP Neural Network 〓

WANG Qi1, MENG Wei2, MA Yunfeng2, HU Xiaomin1   

  1. 1. School of Resources & Civil Engineering, Northeastern University, Shenyang 110189, China; 2. School of Energy and Environment, Shenyang Aerospace University, Shenyang 110136, China.
  • Published:2013-07-09
  • Contact: MENG Wei
  • About author:-
  • Supported by:
    -

摘要: 利用HJ-1卫星CCD数据,通过MATLAB软件计算,分析了多光谱CCD数据的4个波段(近红外波段、红色波段、绿色波段、蓝色波段)参与的65个波段组合与叶绿素a质量浓度之间的关系.研究发现波段组合T1=B4/B3与叶绿素a质量浓度相关性最高,并以此为自变量建立一维线性模型.利用BP神经网络进行模型建立与预测,对比两者的拟合度R2和均方根误差RMSE以及验证点的相对误差.结果表明,利用BP神经网络预测的大伙房水库叶绿素a质量浓度与实测值较为接近,并且效果优于线性模型.

关键词: HJ-1, 大伙房水库, 叶绿素a, 线性模型, BP神经网络

Abstract: With the CCD data of HJ1 satellite and the calculation of MATLAB, this paper analyzes the relationship between chlorophylla concentration and 65 band combinations which consist of four bands(nearinfrared band, red band, green band and blue band) of multispectral CCD data. The study finds that band combination of T1=B4/B3 has the highest correlation with the chlorophylla concentration. Then onedimensional linear model was established by taking T1 as the argument. The degree of fitting R2, the root mean square error and the relative error of the verification points were compared based on the BP neural network. It shows that it is feasible to predict the chlorophylla concentration of Dahuofang reservoir with BP neural network, and the method is better than linear model.

Key words: HJ1, Dahuofang reservoir, chlorophylla, linear model, BP neural network

中图分类号: