CBERS-02B与资源三号多光谱影像的光谱可分性实验对比

Comparison of Spectrum Separability between CBERS-02B and ZY-3 Multi-spectral Images

以江苏省徐州市郊区为实验区,采用Jeffries-matusita距离方法分析了CBERS-02B与资源三号卫星(ZY-3)多光谱影像的光谱可分性。采用最小距离法、最大似然法和支持向量机3种分类算法进行影像分类,以用户精度、制图精度、总分类精度和Kappa系数作为分类精度的评价指标,探讨了CBERS-02B与ZY-3多光谱影像的光谱可分性。实验表明,CBERS-02B和ZY-3多光谱影像均有利于区分农田和水体类,而对建筑物、裸地及道路的光谱可分性较弱;ZY-3影像对建筑物类的光谱可分性优于CBERS-02B影像,但对裸地及道路类的光谱可分性不及CBERS-02B影像。研究结果对于卫星传感器设计和遥感数据应用具有一定的参考价值。

In this paper,the spectrum separability has been analyzed in details with experiments on both CBERS-02B and ZY-3 multi-spectral images covering Xuzhou,Jiangsu Province.Jeffries-Matusita distance is first employed as separability index to analyze the spectrum properties,quantitatively.Then,multi-classifiers including Minimum Distance,Maximum Likelihood and Support Vector Machines are employed for further visualizing the spectrum separability of two datasets,qualitatively.Moreover,user＇s accuracy,produce＇s accuracy,overall accuracy and Kappa coefficients are selected as indicators for crossing assessment of the spectrum separability.Experimental results demonstrate that both CBERS-02B and ZY-3 datasets have good performance in separating plant and water from other articles,while have little ability in separating buildings,roads and bare land from each other.Also,the spectrum sensitivity of ZY-3 multi-spectral image is superior to that CBERS-02B image in separating buildings from other articles whereas it is inferior to that in bare land and roads distinguish.It has certain reference value in satellite sensor designing and remote sensing image application.