摘要
快鸟影像是目前世界上空间分辨率达到亚米级的商用卫星之一,辐射亮度级为2048级、11 bit,将其转换为256级、8 bit,然后在ERDAS软件中进行多光谱、全色分辨率的融合,输出能够被MATLAB软件接受的JPG图形格式文件;基于快鸟影像有4个波段,彩用最佳指数(OIF)进行最佳组合波段的选择,经计算4、3、2组合波段的OIF值最大为76.53684,作为下一步自组织分类的图像;自组织神经网络分类在MATLAB中完成,选择函数newsom,待地物类别数为5类,网络训练最大次数为100次,对分类结果计算出了总体分类精度为83.21%,Kappa系数为78.66%.
Quick Bird is the world′s spatial resolution commercial satellite yami one level,gray-scale for 2048,11bit,will be 11bit to 8bit,then ERDAS software in multi-spectral,panchromatic resolution integration,the output can be accepted by MATLAB software JPG graphics format;quick bird are based on the four-band,color with OIF to choose the best combination of bands,calculated by the combination of 4,3,2-band maximum of the OIF value 76.53684,as self-organizing mapping next image classification;self-organizing mapping completed by the middle of classification in MATLAB,select the function newsom,when the number of features for the 5 categories,the largest number of network training,100,calculated on the classification results of the overall classification accuracy for 83.21%,Kappa coefficient of 78.66%.
出处
《湖北民族学院学报（自然科学版）》
CAS
2010年第2期233-235,F0003,共4页
Journal of Hubei Minzu University(Natural Science Edition)
基金
湖北省自然科学基金项目(2008CDB049)
湖北民族学院博士基金共同资助项目
关键词
快鸟影像
自组织神经网络
分类
Quick Bird
self-organizing mapping
classification
