ALS Point Cloud Regression Analysis
Lidar technology has a strong ability to obtain the three-dimensional structure of the forest, and the obtained canopy structure parameters have a strong correlation with the forest volume and above-ground biomass. It is a large-area forest photosynthesis capacity assessment, biomass and carbon storage. Estimation provides a good means. LiDAR360 provides four methods: linear regression, support vector machine, fast artificial neural network, and random forest. The regression model is constructed from the sample plot survey data and the variables obtained from the lidar point cloud to estimate the forest volume and the amount of the sample square or larger area. Parameters such as biomass.
Regression analysis using LiDAR360 roughly goes through the following steps: denoising, filtering, generating Digital Elevation Model (DEM), point cloud normalization, generating forest parameters, and regression analysis.
The input data for regression analysis is normalized point cloud data. The steps to generate a normalized point cloud are as follows:
-Click Data Management> Point Cloud Tools> Outlier Removal to denoise the point cloud data and remove the influence of noise.
-Click Classification> Classify Ground Points to classify the ground points from the point cloud to generate DEM.
-Click Terrain> DEM to generate a digital elevation model based on the site surface.
-Click Data Management> Point Cloud Tools> Normalization to generate normalized point cloud data.
-Click ALS Forest> Forest Parameters> Calculate Forest Metrics By Grid, Calculate Forest Metrics By Polygon or Calculate Forest Metrics by Forest Stands to generate the independent variable data set required for regression.
The linear regression steps are as follows:
-Click ALS Forest>Regression Analysis>Linear Regression to load the generated forest parameter independent variables into the function dialog box for linear regression analysis.
The support vector regression steps are as follows:
-Click ALS Forest> Regression Analysis> Support Vector Machine, load the generated forest parameter argument into the function dialog box, and perform support vector regression analysis.
The steps of fast artificial neural network regression are as follows:
-Click ALS Forest> Regression Analysis> Fast Artificial Neural Network to load the generated forest parameter independent variables into the function dialog box for quick Artificial neural network regression analysis.
The regression steps of the random forest network are as follows:
-Click ALS Forest> Regression Analysis> Random Forest to load the generated forest parameter independent variables into the function dialog box for fast artificial nerve Network regression analysis.
After the regression analysis is completed, an html file with the same name as the result file will be generated to view the accuracy of the regression analysis and the variables involved in the regression analysis.
The figure below is the support vector machine regression accuracy report. Degree, Gamma and K-Fold are input parameters, which are set by the user in the parameter interface. The accuracy evaluation of the regression model is determined by R, R Square and RMSE. R Square is the ratio of the sum of squares of the difference between the predicted data and the mean of the original data to the sum of the squares of the difference between the original data and the mean. The value range is [0 1]. The closer the value is to 1, the independent variable's ability to explain the dependent variable. The stronger. RMSE (Root Mean Square Error), this value is the mean value of the square root of the error between the predicted value and the measured value. Dependent and Independent Variable are the dependent and independent variables participating in the regression analysis. Only one dependent variable is used for each regression analysis.