Automatic farmland extraction from multi-temporal landsat TM data based on artificial neural network

It is an important method of the land use change dynamic monitoring to withdraw the land utilization information using remote sensing image accurately and quickly. However, most of them seemed to be immature enough. This paper aims to use the prior knowledge which is established from one land cover map and remote sensing imagery to realize the automatic extraction of specific land cove class from other remote sensing imagery. The TM satellite imageries in Changyang District of Beijing are taken as an example, and the automatic extraction procession introduce various key technology including relative radiometric correction, feature selection and ANN. The results show that the classification accuracies between the mentioned approach and conventional statistical method (MLC) for individual remote sensing image are very close.