Model of characteristic parameter for forest plantation with data obtained by light small aerial remote sensing system

Abstract: Airborne lidar and digital aerial photography, in a light small aerial remote sensing system, can obtain three-dimensional coordinates to the quantitative estimate of forest parameters, and in particular have unique advantages in terms of tree height and forest spatial structure estimation. Even though China mainly uses foreign aerial photography system, this study, based on Chinese self-developed high-precision small aerial remote sensing system, established a model between remote sensing data and the ground forestry stand value, and evaluated the accuracy of the model and the feasibility of the aviation system in forestry. The Chinese pine plantation in Shangcheng City, Henan Province was chosen for the study area, and a standard single tree was chosen in the 40 sample plots. The tree height and tree diameter at breast height (DBH) measured by traditional methods were treated as the reference values. A photographic image obtained by the aerial digital photography system was transformed to an orthophoto through mosaic, matching, and stitching processes. With the adjacent pixel-comparison method, the tree crown width was extracted from the orthophoto based on the object-oriented fuzzy algorithm. After noise removal, point cloud data obtained by airborne LIDAR (light detection and ranging) generated a digital elevation model (DEM) and digital surface model (DSM) through an interpolation algorithm. Thus the tree height model is obtained by subtraction. In this paper, based on the 30 sample trees, the linear regression model for tree height was built between LIDAR data and field survey data with model correlation coefficient R2 of 0.895. The relationship is remarkable. The linear regression model for DBH was built by the average tree crown width extracted by aerial images and field survey DBH data, and R is 0.876, also a remarkable result. Based on the other 10 sample trees, the accuracies of tree height model and DBH model were estimated. The height model's overall relative error RS was 0.8%, the average relative error was 0.71%, and the estimated precision P was 97.5%. Therefore, the forecast accuracy is high and can achieve the forestry production requirement standard error of less than 5%. The DBH model's overall relative error RS is -1.9%, the average relative error is -2.0%, and forecast precision P is 91.6%.