Application Research On The Particle Swarm Neural Network In The Remote Sensing Image Classification

With the continuous development of remote sensing technology, people can quickly get a wide range of real-time remote sensing data, the data for scientific research, the national development and national security has an important role. Classification of remote sensing data is also one of the key technology of remote sensing application. This makes a lot of researchers to study the automatic classification algorithm, emerged many supervised and non-supervised classification algorithm. These algorithms are continuously improved to improve the accuracy of remote sensing image classification, however, the improved algorithm also need further research to more accurate realization of remote sensing image classification.This article is started from the remote sensing image classification research status at home and abroad, this paper introduces the main data source for remote sensing classification and their respective characteristics, this paper introduces the remote sensing image classification as the main characteristic of the study and the precision evaluation of classification. In the third chapter, this paper introduces the traditional remote sensing image classification method and support vector machine(SVM) method and object-oriented technology in the application of remote sensing image classification, and some problems in the process of classification, etc. A later chapter of this article mainly introduces two kinds of algorithm, particle swarm and neural network. Because particle swarm can get global optimal solution, BP neural network put the global optimal solution as itself initial weights and thresholds. In the end, the BP neural network model get local optimal solution by training, the training results will be more accurate classification of remote sensing images.On this basis, using particle swarm- BP neural network for remote sensing image classification, the particle swarm- BP neural network model is established. Experiments of remote sensing image classification were conducted. At last, evaluate accuracy of the classification results. The classification result is compared with the traditional classification. Experimental results show that particle swarm- BP neural network classification result is better than the k-means and ISODATA classification result. The classification result overall accuracy is higher than the overall accuracy of minimum distance classification 0.124. The classification result Kappa coefficient is higher than the Kappa coefficient of minimum distance model classification 0.157.