The Comparison Of Two Ways That Extracting Subcompartment Information Of Returning Farmland To Forest Based On SPOT5Remote Sensing Images

As one of the key projects of the forestry ecological construction,converting cropland to forest project has provide an important ecological security for the national ecological security construction and the economic social development.Using different classification techniques to extact the subcompartment dynamic information of converting cropland to forest project,and to explore more appropriate extraction methods of subcompartment information and precision testing methods is a effective way of mastering the detaile of the project progress and implementation effect, can provide scientific and technological support for the ecological engineering decision making and effective management, and offers a technology reference for the forest resources survey technology development.This study choose Liujiang town of HongYa county in sichuan province as the test area,based on two issue SPOT5images data, by overlay analysis, using the comparison patterns after classification to extract the subcompartment of returning farmland to forest which turned out form farmland in2004year to forest in2008year.This test mainly selected two classification methods—supervised classification and object-oriented classification, compared and evaluated them form three aspects:the classification process, classification effects and the extraction accuracy of subcompartment of converting cropland to forest. The results showed that:(1) Supervised classification made the ground boundary was not integrity and existed " salt and pepper noise ", the SPOT5image of2004year and2008year were86.69%and92.16%based on supervised classification; object-oriented classification is better able to retain the integrity of object, and the boundary is clear, the SPOT5image of2004year and2008year were95.71%and96.15%based on object-oriented classification,the classification accuracy is better than that of supervised classification, especially visible on farmland, water, and building land.(2)By overlay analysis, based on the classification thematic map of different periods to extract subcompartment change information of converting cropland to forest is feasible, the extracting accuracy of change information depend on the overall classification accuracy.(3) On the whole, the quantity precision of subcompartment、the accuracy of subcompartment area and the woodland types、location accuracy by object-oriented classification was higher than that by supervised classification, which could explain that object-oriented classification was better than supervised classification in extracting subcompartment information of converting cropland to forest.