基于多源遥感数据的遥感影像生态地块划分方法  

Method for remote sensing image ecological block division based on multi⁃source remote sensing data

作　　者：李双营[1] LI Shuangying(Qinghai Minzu University,Xining 810007,China)

机构地区：[1]青海民族大学,青海西宁810007

出　　处：《现代电子技术》2025年第5期142-146,共5页Modern Electronics Technique

基　　金：青海省科技计划项目(2022-QY-222)。

摘　　要：为资源合理利用、生态保护与修复提供科学依据,文中提出基于多源遥感数据的遥感影像生态地块划分方法,实现了高精度生态地块划分。采用高频调制融合法逐像素融合处理采集的生态环境多源遥感影像;构建新的卷积神经网络(CNN),以融合后的高光谱影像为输入,通过在CNN中引入分组卷积和残差学习,实现输入高光谱影像多尺度特征提取,经过全连接层和softmax层的处理后,输出生态地块划分结果,并在softmax层中引入多分类Focal loss损失函数,解决生态地块划分结果产生的类别不平衡问题,提升生态地块划分精度。实验证明,该方法能够准确划分生态地块,划分精度平均值达到95.38%。融合后的多源遥感影像光谱扭曲度数值均低于20,可以确保融合影像在光谱信息上的高保真度,提高生态地块划分的准确性。A method for remote sensing image ecological block division based on multi-source remote sensing data is proposed to achieve accurate ecological block division,and provide scientific basis for rational resource utilization,ecological protection and restoration.The high-frequency modulation fusion method is adopted to fuse pixel by pixel and collect multi-source remote sensing images of ecological environment.A new convolutional neural network(CNN)is constructed.The fused hyperspectral images are taken as the input.Multi-scale feature extraction of input hyperspectral images is achieved by introducing group convolution and residual learning into the CNN.The results of ecological block division are output after the processing of fully connected(FC)layers and softmax layers.And multi-class Focal loss function is introduced into the softmax layer,so as to eliminate class imbalance caused by the results of ecological block division and improve the division accuracy.Experimental results have shown that the proposed method can divide ecological block accurately,with an average accuracy of 95.38%.The spectral distortion values of the fused multi-source remote sensing images are all below 20,which can ensure the high fidelity of the fused images in spectral information and improve the accuracy of ecological block division.

关 键 词：多源遥感 遥感影像 生态地块 划分方法 高通滤波融合 高光谱影像 融合影像 特征提取 

分 类 号：TN919-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]