Abstract: In this paper we have studied testing the use image information of the multi-stage remote sensing, by sampling of the multistage unequal probability (sampling with probability proportional to size), to do the forest resources inventory, in Xishuang Bana, Yunnan. As to main procedures, there are a few programs as follows:1. Using the LANDSAT image to interpret preliminary datum of percentage of forest land on primary stage unit. 2. Random unequal probability extraction for the sample unit of primary stage. 3. Using the aerophotographs to interpret preliminary datum of percentage of forest volume on secondary stage unit. 4. Random unequal probability extraction of sample unit of secondary stage. 5. Extraction of the sample unit of third stage. 6. Measuring the ground plot. 7. Estimating the population forest volume. The advantages of this method lie in the fact that the aeronautical and spacial remote sensing information is combined together, and so are the remote sensing and sampling. Thus, it fully develops the potential abilities of remote sensing information. By this experiment, the result of calculation of the forest resources in Xishuang Bana Prefecture of Yunnan Province of China is 70, 751, 400 m³, while reliability is 85%(t=1.44), estimated accuracy is 81%. It is verified that this method is efficient in the forest resources inventory of large areas. Besides, to monitor the change of the forest resources, it is also a worthwhile technical method to recommend.