Research On Land Use Information Extraction Method Based On Airborne LiDAR Date And Hyperspectral Data

Land is an important resource of our country,and it is the foundation of the national economic development.Real and accurate land use information is the main basis for the rational exploitation and utilization of land resources.And the rapid and accurate acquisition of land use information is of great significance for timely planning and management of land resources.remote sensing images have been more and more widely used for the extraction of land information.However,with the continuous development of urbanization,the number of buildings has increased,and the types of features have become more and more complex.The development of remote sensing image classification technology has not significantly improved the accuracy of land use classification,so it has shifted from a single remote sensing data source to a multi-source remote sensing data source Fusion.Hyperspectral data has rich spectral information and a large number of bands,and it can extract more complex land use information.And the airborne Li DAR data has unique high-precision three-dimensional space information,but it lacks spectral information.Therefore,extraction of Principal Component Analysis,Minimum Noise Fraction Rotation,Gray-Level Co-occurrence Matrix,Normalized Difference Vegetation Index,Normalized Difference Build-up Index from hyperspectral data,and normalized Digital Surface Model and Intensity are extracted from airborne Li DAR data for parameter Fusion.In order to obtain more accurate land use information.In recent years,object-oriented classification methods have been able to overcome the pepper and salt effect based on pixel classification methods to a certain extent,and have played an important role in improving the accuracy of remote sensing classification of land use information.At the same time,it analyzes the classification results obtained by using the object-oriented threshold classification method and the traditional supervision classification method.Therefore,Zhangye in Gan Su province of China were taken as research area,andIV combined with hyperspectral data and airborne Li DAR data fusion operations to extract land use information.The main innovations and conclusions are as follows:(1)It is based on the most adjacent taxonomy in software e Cognition,combine the parameters obtained from the analysis of hyperspectral data transformation,which was combined with the parameters of Airborne Li DAR data,in order to avoid the same accident,the combination of all the neighboring classification,the classification results show that,the classification results of hyperspectral data and Airborne Li DAR data fusion are more significant than those with single hyperspectral data.Among all the parameters,the best classification effect is the combination of hyperspectral principal component analysis parameters and altitude parameters and intensity parameters of Airborne Li DAR data,and the overall classification accuracy is 92.06%.Compared with the hyperspectral data only,the overall classification accuracy increased by 7.93%.The classification accuracy of the classification results of different parameters is slightly lower,but the classification accuracy of single ground objects is found to be different.So,when requirements for the extraction of different individual ground objects,it can according to the different parameters fusion to obtain the best classification result.(2)In view of the results of the fusion of hyperspectral data parameters and Airborne Li DAR data parameters,the object oriented classification method,the traditional supervised classification method similar to the object oriented nearest neighbor classification,and the object oriented threshold classification based on object classification are compared respectively.The following conclusions are drawn: in view of the fusion of hyperspectral data parameters and Airborne Li DAR data parameters,the land use information extraction of the object oriented nearest neighbor classification method is the best,and the overall classification accuracy is 8.56% higher than the maximum likelihood method of supervised classification.and the classification accuracy of the object oriented threshold classification method is improved by 5.66%,and a variety of feature classification accuracy are improved.In the traditional supervision and classification results,many buildings and road are similar to the material of the two,and the spectral features are very similar.Thus,the same spectrum foreign object phenomenon is produced,and the edge of the building is misclassified into the road.After the fusion of hyperspectral data and airborne Li DAR data,the classification resultsare compared with the object oriented nearest classification method.The nearest neighbor classification method effectively avoids the misclassification of buildings.And feature values are selected as the basis of classification in the process of object oriented nearest neighbor classification,which is more accurate than traditional supervised classification.In addition,in the object-oriented threshold classification method,the effect of eliminating the building images by setting a threshold is remarkable.Thus,a single ground object can be extracted by setting a threshold,but it makes the classification process more cumbersome when setting the threshold,and is not advantageous to the extraction of large area of ground objects,and it is applicable to a single ground.(3)This paper focused on the method of fusion of hyperspectral data parameters and Airborne Li DAR data parameters,and in view of the combination of hyperspectral data and Airborne Li DAR data,the limitations of the object oriented nearest classification,threshold classification and supervised classification are discussed and analyzed,sought the best effect of land use information extraction through various parameters combination.At the same time,it is found that for the most serious area of building shadow,we can set thresholds in the threshold classification to remove the shadow,and then do related research.Therefore,this paper makes full use of the advantages of hyperspectral data and Airborne Li DAR data,it is of practical significance to improve the accuracy of land use information extraction by using many kinds of remote sensing image information.