两种热带森林覆盖变化检测方法比较研究

[目的]以热带森林复杂区域为对象,对两种缨帽变换“衍生数据”检测热带森林变化的方法的优劣进行比较.[方法]对两期数据进行缨帽变换,同时结合黑暗对象掩膜与局部直方图阈值提取等方法,获取亮度、绿度、湿度指数组合(MKT)和干扰指数(DI).采用MKT差值(MKT-D)、干扰指数差值(MDI-D)进行变化信息识别,然后,根据植被覆盖与亮度、绿度、湿度之间的变化关系,通过决策树分类提取变化信息,最后,对不同检测结果进行验证与比较分析.[结果]结果显示两种方法都能检测出森林内部的细微变化,但对小图斑变化,MKT-D检测优势明显,且MKT-D的kappa系数为0.763 0,MDI-D的kappa系数为0.655 9,两者相比,MKT-D方法优于MDI-D方法.[结论]MKT能够增强短波红外与近红外波段对森林变化信息的敏感性,有效地消除噪声等非目标信号,突出目标信号,此外,MKT-D为RGB彩色图像,更利于变化信息的目标提取与解译.

A Comparison of Two Methods for Detecting Tropical Forest Change Cover

Objective In this paper, two methods from two "variant data" to detect tropical forest change were compared. Method After Tasseled Cap transformation, the combine of Brightness index, Greenness index, Wetness index (MKT) and Disturbance index (DI) were obtained by masking the dark object and extracting local histogram threshold. Change information was discovered by the differences of MKT and DI, According to the connection between vegetation cover and Brightness index, Greenness index, Wetness index, the change information was extracted by decision tree classification, and the results were assessed and compared. Result The results showed that they were all able to detect the subtle change of tropical forest, but, MKT-D take an obvious advantage to detect the small change spots, and the total kappa coefficient from MKT-D (0.763 0) was higher than that from MDI-D (0.655 9). For tropical forest change detection in this research, the result from MKT-D was better than that from MDI-D. Conclusion MKT could enhance the effect of near-infrared and short wave infrared band to forest change information. Moreover, MKT-D is accessible to extraction and interpretation of target change information.