Forest Resource Classification And Volume Inversion Based On SPOT5

Forest Resources Survey investigates forestland, forest fauna and flora, and their habitation conditions. The purpose of the survey is to understand the dynamic laws of the quality, quantity, growth and death of forest fauna and flora and their relationships with natural environment, economy, and exploitation. This is done to formulate and adjust forestry policies, to design forestry plans, and to evaluate results accordingly in order to make sure that forest resources are fully used in a way that maximize their usefulness for the domestic economy. The traditional investigation method, which is both time consuming and cost inefficient, is dependent upon manual labor. Remote Sensing Technology (RST), with features such as micro, macro, high efficiency and dynamic monitoring, used along with GIS and GPS has become the new standard in forestry research. RST has become a hot topic in domestic and foreign forestry studies.The data used in this thesis was abstracted from SPOT 5 Remote Sensing Image Fusion of Zhongshan Mountain 2003, and 1: 10000 the relief maps, along with the actual survey figures from 2002 Zhongshan Mountain Survey. First, the standards on Remote Sensing Image, Data Fusion, and Vegetation Index abstraction are set. Then the variables of each band are analyzed and band sets for Remote Sensing Image Fusion classification through past experiences are sorted out. Finally the classification of Object-Oriented Image, Object-Pixl (CART Decision tree and Maximum Likelihood), and evaluation of these classifications are discussed.Image Segmentation Technology is used in the classification process of Object-Oriented Image, and to some extent reduces the noise level created during the classifying process. Both spectral information and geospatial information are used during the classifying process. This classifying method is generally easy to understand and follow, as a result has created better classifying results. The classification of Object-Oriented Image is a superior classification process compared to other methods in Re Remote Sensing Image with high resolution. Volume of vegetation is used to create model, the results of Object-Oriented classify Image, and Remote Sensing spectral information is used for predictions, the actual Volume figures of the 38 samples from field work are used as target factors. Nonlinear BP neural networks model is used to develop remote sensing information, classification and volume models. When working on the superior plants in the Broad-Leaved Forest of the Zhongshan Mountain , spectral information of remote sensing is used to make predictions. The measured volume of the 28 samples taken from the Broad-Leaved Forest is being used as target factors for developing BP neural networks models.The Neural Network System is composed of several neurons. Each neuron can receive multiple signals, process them, and properly relay them. Due to the complex linking relationships between neurons in the neural network and the nonlinear signal transmission methods, a various relationships can be developed within the receiving and sending signals. The various relationships demonstrate conditions that cannot be expressed Mechanic Law, but actually exist in the receiving and sending process. The network models are used to predict vegetation volume of different locations. The gathered data was then used to design volume prediction maps and record image fusion maps for the Image Fusion Vegetation and for the Broad-Leaved Forest of Zhongsan Mountain.Using this method, a volume of 22.6mm~3 for Vegetation and 18.3mm~3 for Broad-Leaved forest of Zhongsan Volume are measured.