Research On The Method Of Improving The Accuracy Of Land-cover Classification And Extraction Based On Airborne LiDAR Data

Airborne Li DAR(Light Detection and Ranging,laser scanning ranging)system,as an important branch of active remote sensing measurement,has the characteristics of high detection data accuracy,convenient and flexible working methods and so on.Using airborne Li DAR data for land-cover classification and object extraction is the main research content of airborne Li DAR data processing methods and an important prerequisite for urban scale construction planning and vegetation growth situation monitoring.However,due to the differences in the characteristics of different types of data and the limitations of feature identification capabilities,the accuracy of land-cover classification and object extraction is low,and it is difficult to meet the high-precision requirements of multiple scenarios,which seriously restricts the research on airborne Li DAR data algorithms' development and application.Therefore,in view of the low accuracy of airborne Li DAR data land-cover classification and object extraction,this paper starts with the performance analysis and extraction of different types of data features,and studies the methods for improving the classification and extraction accuracy of land-cover objects.The main research contents are as follows:(1)Feature analysis of airborne Li DAR data: On the basis of understanding the working principle of the airborne Li DAR system and the data acquisition process,the possibility of extracting direct features and indirect features from the point cloud data and image data of the airborne Li DAR system and the effectiveness of the different feature identification ability about different Land-cover objects is analyzed,and the suitable data feature types for two different application requirements of land-cover classification and single object extraction are studied.(2)Research on land-cover classification methods based on multi-source data feature fusion: Firstly,combining the complementary advantages of classification performance between different features,then analyze the land-cover feature attributes represented and extract relevant applicable features by airborne Li DAR point cloud data and remote sensing image data;Secondly,aiming at the issue of uncertain information that affects the classification accuracy of land-cover objects caused by heterologous difference features in feature fusion,by constructing different feature classification performance subsets and fuzzy trust allocation functions,a land-cover classification method based on DS evidence theory is proposed;Finally,the experimental results of the land-cover classification method are analyzed and evaluated on different regional experimental data sets to verify the effectiveness of our land-cover classification method.(3)Research on land-cover object extraction method based on multi-feature hierarchical strategy: Aiming at the issue of insufficient feature recognition in land-cover object extraction,this paper directly analyzes and extracts features from airborne Li DAR point cloud data.Firstly,the idea of point cloud filtering is introduced to solve the issue of low algorithm efficiency caused by the huge amount of point cloud data,so as to realize the removal of ground points irrelevant to the extraction of certain object;Secondly,through the analysis of the attributes of the type of land-cover objects,a combination of geometric and statistical characteristics of the spatial structure suitable for the extraction of a certain type of land-cover objects is constructed on the three-dimensional structure of the land-cover object point cloud;Finally,a hierarchical strategy is introduced to achieve accurate extraction of a certain type of object through multifeature threshold segmentation,In the experimental part,two typical land-cover objects of vegetation and buildings are taken as examples,quantitative indicators were used to evaluate the experimental results and compare and analyze the experimental errors to verify the effectiveness of our land-cover object extraction method.