Remote Sensing Image Classification Of Forest Vegetation Based On Artificial Neural Networks And Fuzzy Classification

Forest is an important component to the whole world ecosystem.The condition and changes of forest resource,not only affects the changes of environment,but also affects the sustainable development of the society economy.With the development of remote sensing technique,the classification of the forest remote sensing image has become an essential work in the investigation and monitoring of the forest resources.At present the classification technology of the forest remote sensing image is in a quite backward condition in the forestry.How to use the remote sensing data resources and other forestry data to study a new classification method to improve classification accuracy to satisfy the demand of the forestry department is the main work of this research.Based on the related research of the domestic and international,taking typical forest in the east of northeast China as experimental area,using Landat5 TM multi-spectral image as the remote sensing data,this research applies the BP neural network and the fuzzy C mean to classify the image.According to the analysis of the classification accuracy,and comparing the advantages and disadvantages of the two methods,some valuable conclusions can be drawn.According to the analysis of classification result,the primary conclusions as followed:(1) The classification accuracy of traditional classification methods to the remote sensing image of northeast forest is low.(2) The advantage of BPNN method is obvious to the traditional classification methods. Both the overall classification accuracy and the Kappa coefficient of BPNN method are higher than the traditional classification methods.And the classification accuracy of different vegetation types has increased by using BPNN method.(3) Compared with the traditional classification methods,the geographical assistance data,can be joined into the classification easily with the BPNN method,which can improve the classification accuracy of the remote sensing data using the rich geographical assistance data provided by the geographic information system.(4) The fuzzy C mean clustering method based on the fuzzy set theory is introduced into the remote sensing image classification.Although the classification result has not achieved the anticipated goal,the fuzzy C mean clustering method still has higher classification accuracy than the traditional classification methods.(5) It is indicated that the BPNN method has greater advantage than the fuzzy C mean clustering method by comparing the classification accuracy.That is,BPNN classification method is more suitable for remote sensing image classification of forest vegetation.