Research On Remote Sensing Extraction Of Crop Planting Structure Based On Machine Learning

Nowadays,with the promotion of remote sensing technology,remote sensing data show similar characteristics to big data,that is,large data volume,fast response,diverse data and low value density,this requires machine learning related knowledge to deal with these remote sensing data..How to improve the accuracy of crop planting structure remote sensing extraction is the focus of this study.In view of the extraction of crop planting structure in the resolution remote sensing image of GF-2 satellite 3.2m,Chengguan Town,Tai Qian County,the classification process of remote sensing image of BP neural network and the method of improving the accuracy of remote sensing image classification are emphatically studied.Therefore,this article has carried out the research work from the following aspects:(1)Research on the related theories of agricultural remote sensing,to explore the theoretical basis of various remote sensing image classification methods;(2)Describe the classification process of remote sensing image based on BP neural network in machine learning,and analysis the current situation and difficulties of Remote Sensing Extraction of crop planting structure;(3)Exploration of methods for improving classification accuracy of remote sensing images: a comprehensive classification method for improving the classification accuracy of agricultural remote sensing images is proposed through the comparison of various classification methods and classification algorithms of machine learning and the improvement of the operation methods in the experimental process.The research results show that the image classification method based on multi-source data can effectively improve the classification accuracy of low resolution remote sensing images compared with the image classification method based on single source data;compared to the ISODATA,K-Means methods of unsupervised classification,the BP neural network classification method can more accurately identify the target objects information in remote sensing images,compared with the SVM and maximum likelihood classification methods of the supervised classification,the BP neural network classification method is more responsive to the trend of intelligent classification;compared with the traditional classification methods,the precision of a variety of lifting classification precision will be improved obviously;the object oriented classification method can improve the classification efficiency and save more time,labor and cost than other classification methods.