BP神经网络的煤矿塌陷区水深反演
Water depth inversion in coal mine subsidence area based on BP neural network
付俊 1毕京锐 1付翔2
作者信息
- 1. 安徽省核工业勘查技术总院,安徽 芜湖 241000
- 2. 安徽理工大学 空间信息与测绘工程学院,安徽 淮南 232001
- 折叠
摘要
为探测煤矿塌陷区的水深,利用Sentinel-2 遥感影像数据和无人船实测的水深数据,分析蓝波段、绿波段和红波段与水深实测值的相关性较高,确定其为反演因子,通过提取水深点的辐射亮度值并作归一化处理,建立单波段和多波段对数比值模型,以及BP神经网络模型反演矿区塌陷水域的水深.结果表明,BP神经网络模型的决定系数为 0.895,均方根误差为 0.428,平均相对误差绝对值为8.12%,预测结果最好,精度最高,优于传统的线性模型.
Abstract
This paper aims to detect the water depth in coal mine subsidence area.The study in-volves using Sentinel-2 remote sensing image data and the water depth data measured by unmanned ship to analyze the higher correlation between the blue wave band,green band,red band and the measured water depth value respectively,which is determined as the inversion factor;extracting the radiation brightness value of the water depth point and performing the normalization processing;establishing the single band and multi-band logarithmic ratio model and the BP neural network model,which is used to retrieve the water depth of mine subsidence.The results show that the BP neural network model with R2 by 0.895,ERMSR by 0.428,EMAPE by 8.12%is the best prediction result and the highest precision,and better than the traditional linear model.
关键词
塌陷水域/Sentinel-2/水深反演/BP神经网络Key words
collapsed waters/Sentinel-2/water depth inversion/BP neural network引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据