Remote Sensing Classification Study Of Wetland Based On Artificial Neural Network

Classification of remote sensing image using computer is an importantapplication of remote sensing. It recognizes objects by data analysis of remote sensingimage pixels. There are mainly two methods: non-supervised classification, andsupervised classification. Non-supervised classification is a clustering process, whilesupervised classification is a study and training process, and it needs preliminaryknowledge.At present, the remote sensing classification is indispensable to the wetlandresource survey and supervision. The accuracy of the classification will directly effectremote sensing data's application and utility. Therefore, the solution of image'sidentification with accuracy is an important problem and has a great significance ofthe remote sensing image study.In the last few years, artificial neural network technology has increasinglybecome an effective tool in remote sensing image classification. Artificial neuralnetwork is a network connected by lots of processing cells (nerve cells), which is thesome abstract, generalization, simulation of human brains. Now, the neural networkhas a lot of models in the remote sensing image classification processing. Among themodels, the multi-layer perceptron's (MLP) back propagation is the most-widely used.Based on the world wetland remote sensing classification analysis, this study hasapplied the ETM+ remote sensing data and assist data, used BP neural network tostudy part wetland in Dong ying city. Also this study has compared its accuracy withthe maximum likelihood classification's.According to error matrix analysis, BP neural network method improved useraccuracy, producer accuracy, overall accuracy, Kappa coefficient, Compared withmaximum likelihood classification's . Using BP neural network method, the totalaccuracy is 82.128%,Kappa coefficient is 0.772,improved 12% and 15%. Thisindicates that BP neural network method is an effective classification method, and itcan improve the classification accuracy of remote sensing image.