1. INTRODUCTION
As one of the most vital components in terrestrial ecosystem, forest resources are not only the crucial material foundation for the country, but also the indispensable renewable resources for the establishment of ecological environment and economic construction [1]. They have many ecological functions, such as water conservation and air purification, even to affect the regional environment [2]. In order to accomplish the scientific management and effective utilization of forest resources, it is necessary to obtain a large amount of information of forest resources accurately, quickly and efficiently [3]. In early time, forest vegetation classification was often based on measured data adopting visual interpretation, which bring about a series of problems, such as heavy workload, time-consuming, poor timeliness and low accuracy. Remote sensing technology provides a possibility way for monitoring the forest resources with its dynamic, convenient, high-quality and efficiently characteristics [4]. Recent years, compared with the single-date images, time-series vegetation indices have a more efficiency tendency to improve the classification accuracy, such as the normalized-difference vegetation index (NDVI) [5]. However, due to the restriction of acquiring data with both high spatial resolution and high temporal resolution, it is advisable to fuse data from multiple sensors, thereby gaining the rich informative remote sensing data [6]. The spatio-temporal data fusion technique includes the spatial and temporal nonlocal filter-based fusion model (STNLFFM) [7], the spatial and temporal adaptive-reflectance fusion model (STARFM) [8] and so forth. The STARFM algorithm is most widely used. However, due to giving more attention on the reflectance changes in the multitemporal images, STNLFFM model performs a higher prediction accuracy.