Study On Simultaneous Estimation Of Forest Resource Survey Factors Based On Airborne Lidar Measurement Data

Human interest in forests is increasing with the development of the times, laser radar technology provides a new technical support to obtain spatial information for forest, Its distinctive data features and access to massive data provide practical application of great potential. This paper studied focuses on the theme of using airborne LIDAR data to acquire forest resources investigating factors in order to promote the developing of laser radar technology application in the forestry field.In this paper, airborne LIDAR datas, measured ground positioning datas and stand datas were obtained from Qinghai spruce water conversation forest of Dayoukou Ecological Station in Gansu province. DEM filtering of LIDAR datas was carried out firstly, and next part of the measurement factors (individual tree height, crown width and number of trees) were inverted from airborne LIDAR datas,and then LIDAR datas and the inversion of the measurement factors were used to construct variables, forest parameters were re-estimated by application of multiple linear regression model and simultaneous equations,this fully reflects the capacity of LIDAR data inversion of forest parameters The results showed that determine coefficient of estimation model based on LIDAR data about these major forest parameters such as the biomass, number of trees, basal area mean diameter, tree height, crown width of was 0.829,0.837,0.866,0.897,and 0.886.Main contents and conclusions are:1.Putting forward ideas and methods to obtain DEM by way of discriminant analysis Forest land with obvious open space was selected as supplementary sample plot , The single-and multiple-echo airborne LiDAR data of the sample plot were obtained and analyzed in terms of three-dimensional coordinates distribution,reflection intensity distribution and vertical distribution. The two initial groups of ground and vegetation point set was separated from airborne LiDAR data, then presents a DEM extraction method of forest land based on discriminant analysis. Forest with a large number of control points was selected for model validation. A total of 1472 control points were measured. Results showed that it is feasible to extract DEM in the complex environment of the mountain forest using discriminant analysis method with variance of 0.05.Its accuracy can reach mean elevation difference 0.236m. Data with elevation difference between -0.5m and 0.5m amounted for 93.85% of the total. This method avoids determining parameters or threshold values artificial and help to improve the automation of data processing.2.Determing the identification number of individual trees and establishing the empirical models of variable window sizesThe identification number of individual tree was determined by matching the CHM local highest point of the window with the measured single-tree. Window was divided into round and square templates, fixed and variable forms. It was obtained by comparing that the circular window is better than the square window, the variable.window is better than the fixed window. Variable window diamete(rW = 0.003164*h^2 + 0.1311*h + 1.108)was constructed according to the correspondent average crown of the classification height , and with the error adjustment. After comparing the vertical and horizontal error of identified individual tree in the different adjusted windows and the smallest window,the number of recognition plant and matching rate, the identification number of individual tree was determined.Further testing by the change of height and distance error of remaining identified individual tree removed part of single tree (too small identification crown) , secondary verification was carried out by the measured number of identified individual tree above the major height of crown closure.Eventually it was reached that such mode of window size construction is feasible, the number of identified individual trees in the variable window adjusted by 3 times standard deviation (ie Cw-0.6) was the best.3. Analyzing the elevation-frequency curve shape of CHM scattered data on forestThe elevation-frequency curve shape of CHM scattered data on forest have typical characteristics: With the elevation decrease, the frequency is always increased to a maximum value, then the first decline. The frequency of the maximum means stand branch and leaf area of the height in horizontal space to the maximum level of expression. The height can be seen the main canopy height, this feature is highly stable. Based on the characteristics -height, the Identification of individual tree was divided into large trees and medium trees(see 5.1.2). In the later study, the height is a significant indicator to respond major forest parameters. 4. Establishing the single estimate model group of main forest parameters based on airborne laser radar datasIn this paper, a series of variables was constructed with the individual tree datas identified from laser datas, classification datas of the identified individual trees, the first echo datas of the surface of vegetation and point cloud datas of vegetation. with which and the main forest parameters multivariate linear regression models were built .After variable selection, a total of 14 single major forest parameters estimation model were obtained,viewing from the model regression effect, determine coefficient of these major forest arameter estimation model such as biomass,plant trees,basal area mean diameter,tree height and crown width reached 0.829,0.837,0.866,0.897and 0.886. Airborne LIDAR data have the the better estimation ability of main forest parameters. Conducted a meaningful exploration on LIDAR explanatory variables in significant response of forest parameters,and pointed out a number of research ideas for future work.5.Establishing simultaneous equations modelSingle model can not reflect the inherent relations among forest parameters, especially in the case of the forest is not destroyed, within the parameters of forest there are often fixed mathematical relationship or a wider range of biological links Through the intervention of simultaneous equations model and obtaining its parameters, strengthened the linkages between the individual models. This not only provided the probability of improving the regression effect of part model but also more extended suitable ability of the model.As the constraints of the model relationship, the practical application of a single model has been challenged. In this paper,through the simultaneous application of a single model and parameter solution, the relationship between the model make possible to embody.the effect of simultaneous solution is better than that of the direct derivation between the model.