Research On The Classification Of Remote Sensing Image And Information Issuance Technique

Auto-classification of remote sensing image is the concrete application of patternrecognition technique in remote sensing technique territory. Compared with conventionalstatistic classifier, the artificial neural network (ANN) shows tremendous advantage. ANNdoesn't need suppose parameterized distribution of sample space in advance. It has goodparallel processing performance and adapts to the necessaries of multi-band. ANN hascomplicated mapping capability. The back propagation neural network modal (BP modal) isoften been used. An auto-classification model of remote-sensing image using BP neuralnetwork based on Simulated Annealing is put forward in this paper. Utilizing the SimulatedAnnealing idea this network model unites the respective preponderances of the method ofgradient descent with momentum and the standard BP neural network to adjust themomentum parameter and the weights, which is able to make the network escape from thelocal minimum spots and converge steadily. Experimental results of network training andclassification of remote sensing image show that the improved network converges easily, itsperformance is steady, the classification accuracy is comparatively high and it has practicalapplication value.According to the actual issuance of remote-sensing information and combining with theuniversal Web service object's encapsulation technique, a solution is brought forward andachieved, which issues remote-sensing information with XML Web service technique tosupport multi-platform and seamless integration. Application across platform and differentlanguage are realized. At the same time, remote-sensing information e-commercializationand the remote-sensing economy development are accelerated.