Research On Parallelization And Intelligent Of Remote Sensing Image Classification

High-precision remote sensing image classification is an important prerequisite for the application of remote sensing technology in other fields.With the rapid development of remote sensing technology,it has been a hot spot in the field of remote sensing to classify the surface features by remote sensing technology.In recent years,more and more scholars apply the machine learning algorithm in the field of pattern recognition.Among them,the shallow machine learning algorithms have achieved certain results in remote sensing image classification,but due to the limited ability of learning and the poor of generalization ability,the efficiency is not good enough to meet the needs of massive remote sensing image classification.The rise and rapid development of deep learning has opened up a new path for the research of massive remote sensing image classification.The distributed computing framework provides an effective solution for the efficient parallel classification of remote sensing images.This paper studies the classification of remote sensing image from three aspects: shallow machine learning,deep learning and parallel computing,the main research contents are as follows:(1)A cotton recognition model based on BP neural network is established.In order to simplify the process of using the remote sensing image to recognize the cotton,and improve the accuracy of cotton recognition,this paper proposes a cotton recognition method based on BP neural network.Multiple feature index such as the normalized difference vegetation index,the difference vegetation index and the ratio vegetation index were extracted from the high-resolution 8m resolution image,and then combined different feature index as the input of BP neural network to train the network.The experimental results show that the model can effectively improve the accuracy of cotton recognition in remote sensing images.(2)A classification model of remote sensing image based on convolution neural network is proposed.For shallow machine learning limitations in processing of massive remote sensing data,this paper will introduce the deep learning theory into remote sensing image classification,and put forward the model of remote sensing image classification based on convolution neural network.The model can directly use remote sensing images as input of network,which can be used to realize the automatic classification of remote sensing images through the network training.The experimental results show that the classification accuracy of the model can be improved by increasing the number of iterations and training samples,and the classification accuracy of remote sensing image based on convolution neural network reaches 96.35%,which is 8.90% and 4.78% higher than Support Vector Machines(SVM)and BP neural network.(3)The parallel method of remote sensing image classification based on Spark distributed memory computing platform.In order to solve the problem of high consumption of convolution neural network model in the classification of remote sensing images,this paper introduces the idea of parallel computing into the research of remote sensing image classification.With the help of Hadoop Distributed File System(HDFS),the remote sensing image is distributed and stored,and the efficient parallel implementation of remote sensing image classification is realized by Spark distributed memory computing platform.Implementation of parallel neural network model based on data parallel in distributed computing cluster.The performance of the parallel convolution neural network model is verified by three experiments.The experimental results show that the parallel convolution neural network model can improve the efficiency of grassland classification while ensuring the accuracy of remote sensing image classification.