نوع مقاله : مقاله پژوهشی
نویسنده
گروه مهندسی بیوسیستم، دانشکده علوم کشاورزی، دانشگاه گیلان، رشت، ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسنده [English]
Background and Objective: The drastic changes of land use in northern provinces of Iran have been one of main issues of the policy-making during the last two decades, and so it is always important to find an appropriate methods. The present research aims to evaluate the performance and compare three plant spectral indices for land use/cover classification in Lahijan and improve the obtained results.
Materials and Methods: OLI images of Landsat 8 were used to produce maps of NDVI, LAI and EVI. The classification was done using decision tree and thresholding according to six land uses/covers of the region, and land use/cover maps were developed. To improve the classification performance, a new decision tree was proposed by combining NDVI and DEM.
Results: The validation results using the confusion matrix showed that the performance of NDVI with overall accuracy of 87.7% and kappa coefficient of 0.83 was more than the others. The overall accuracy of LAI and EVI was 71.4 and 71.6%, respectively. The kappa coefficients were 0.828, 0.609 and 0.614 respectively. The proposed method increased the overall accuracy by 4.94% and reached 92.60%. It also increased the percentage of true classification in the tea class from 50.75 to 86.87% and in the forest from 95.08 to 98.18%.
Conclusions: In areas with dense vegetation, it is possible to improve the results of land use/cover classification by combining NDVI and DEM in a decision tree. The results can be used for quick mapping and detecting of land use/cover of the region.
کلیدواژهها [English]
)