Abstract: Forestland is the carrier and an important component of forest resources. How to accurately and quickly obtain information of forestland change is of important significance to monitoring and management of forest resources. In this study, we improved the training sample selection method in the decision tree classification to enhance the accuracy of forestland classification and change detection. The reliability of the improved method was tested with Yichun as a case. The results showed that the improved decision tree classification method outperformed traditional method with the accuracy of change detection being improved by 4.01%. Furthermore, the method improved the detection accuracy in regions with shadow, mist and arable land. Therefore, our method is more reliable for the detection of forestland changes.

Key words: quality., LED, seedling quality, eggplant, light quality, yield