| According to statistic, since the1990s, because of predatory development, area of tropical forest is shrinking rapidly, its ecosystem is suffering unprecedented threat. It has become the major emitters that greenhouse gas emissions was caused by tropical deforestation and degradation in developing world. Human sustainable development put forward the urgent request to monitoring of tropical forest change.The basin of Mekong river is one of the important distribution of tropical forest, is also one of the hot area where the forest is dramatic changing. Remote sensing has been a major means of monitoring forest cover change. Landsat TM data has been widely used to forest cover change in regional level.In this paper, Cambodia was as the text zone, based on the feature change of re-mote sensing data was caused by forest change,such as spectral and texture, monitorin-g the forest cover change using2005/2010Landsat TM data, put forward the workfio wof monitoring quantificationally tropical forest change, on the basis of improving the m-ethod, improve the monitoring accuracy,the results will provide the basis to understa-nd forest cover change in Cambodia and be of guiding significance to monitoring tro-pical forest change in Mekong river basin or others, the main research contents and conclusions area as follows:(1) The methods of obtained samples of the classification basing on GlobCover2005and GlobCover2009data sets was put forwardIt is difficult to collect samples used in classification of land use and land cover in the large scale and a large amounts of samples are needed in supervised classification. In this paper, the samples was collected from consistency region of GlobCover2005and GlobCover2009data sets with a series of processing including removing the minimum spots and quality inspection, finally, satisfying the need of samples. (2) Finish the precision classification in Cambodia using a hybrid extraction approach though a hierarchical classification strategy (HCS)In tropical region, vegetation is very bloom and diverse, so,spectral characteristics is complex. It is difficult to classifying image directly. The hybrid method consisting of multi-level and many methods was used to classification,water wetland urban and built-up cropland was extracted and masked and forest, shrublandand grass was classified by supervised classification. Finally, obtain the map of land use and land cover in2005and2010year.In results, show that the accuracy of others is about80%in addition to the accuracy of the grass is70.37%in2005year. The accuracy in2010is better than them in2005.(3) Change detection method based on the fusion of different results was put forward and completed the maping of forest chenge in CambodiaAccording to comparison, adopting post-classification comparison and difference of the first three index of the Tasseled Cap transformation (TC), because post-classification can provide the type information and the direction of change and the prior index of TC is ability to detect the small spots of change. After post-processing, the map of forest cover change between2005and2010in Cambodia was obtained.(4) The methods of obtaining validating data was put forward and accuracy was validatedBase on GlobCover2005/2009which have beenpre-processed,Validation was collect by similar to methods of collecting training samples. Though validation, results show that the accuracy of the conversion of forest land to non-forest land and shrubland to non-forest land areO.7426and0.7533in kappa coefficient. It is high precision. The second is non-forest turn into forest land, the accuracy of other types are also perfect in addition to shrubland to forest land with the accuracy is0.4795in kappa coefficient.(5) Analysis of amounts and driving for forest cover change was completed at2005and2009years in CambodiaAs population is increasing and the development of economic in which wood processing is one of the main products of industry. In addition, Industrialization and urbanization are also the main cause. Forestland in Cambodia presents the downward trend.Instead of cropland and urban and build-up is increasing.forestland degradation speed is0.014%in every year. The speed is greater than the ones of addition of non-forest and shrub land with0.008%and0.003%.In a word, The entire workflow of monitoring forest cover change was established in this article. The results show the results of change detection can be satisfaction and ability to finish the task to monitor forest cover change through samples and validation data semi-automatic obtain. |