Forest Structure Parameter Extraction Based On UAV And TLS

Forest resource survey is the basis for dynamic resource monitoring and management.As the prerequisite,the survey is significant for the protection and utilization of forest resources.With the development of drone photogrammetry technology and ground-based lidar technology,its application in forestry is becoming more and more extensive.The high-resolution remote sensing image of drone and the high-density 3D point cloud of ground-based lidar provide a new possibility to extract forest information.Due to the complexity of the forest structure,the ability of UAV remote sensing and Terrestrial Laser Scanning to obtain data is limited.Combining UAV and TLS for forest structure parameter extraction can make up for the shortcomings of single measurement method.The combination can better exploit the potential of UAV and TLS point cloud in forest structure parameters extraction.Based on object-oriented classification,spatial analysis and regression analysis,UAV photogrammetry and lidar ranging technology are used to process image and point cloud data.Combining the measured values of the ground with the UAV and TLS point cloud to analyze the advantages of the fusion point in extracting forest structure parameters.The main conclusions are as follows:(1)The UAV and TLS point clouds are well-aligned by coarse registration and accurate registration based on the nearest point iterative ICP algorithm.The study selects the common points of the two point cloud models for fusion.The maximum error of common point matching is 0.4926,and the root mean square error is 0.3367.The results show that CloudCompare can achieve high-precision fusion of different data sources and point clouds of different density levels.(2)Based on Agisoft PhotoScan,the automatic splicing and 3D model reconstruction of UAV 2D images can be quickly realized,and the color point cloud data of the study area can be obtained.The high-precision digital orthonormal model(DOM)and digital elevation model(DEM)are used to extract the canopy closure degree,the east-west and north-south direction crown of single-wood,and the correlation coefficient was 0.9538,0.8672,0.8939 respectively.Meanwhile the accuracy of the north-south direction is slightly higher than the east-west direction due to the data acquisition method.(3)After pre-processing,the TLS point cloud generates seed points for point cloud segmentation based on regional growth combined with threshold judgment(PCS)algorithm,canopy height model(CHM),and layer stacking.The research finds that the point cloud segmentation based on PCS algorithm fits the actual situation.The relative error of the generated seed point is 2.885%,the estimation accuracy is 97.115%,and the average running time is 116 min.The results of single-wood parameter extraction showed that the TLS point cloud had higher precision for single-wood position,breast diameter and crown width,while the high extraction capacity of Metasequoia trees was limited,with a relative error of 5.8%.(4)The UAV and TLS fusion point cloud has obvious advantages in extraction of Metasequoia tree height,and the correlation coefficient with the measured values is 0.9993,which can make up for the deficiency of using only TLS point cloud extraction.The study also analyzed the abnormal data in the automatic extraction.After the abnormal value was removed by human-computer interaction,the correlation coefficient between the average crown amplitude and the breast diameter fitting measured value increased by 0.6 and 0.35,respectively.The results showed that the fusion point cloud could effectively extract the crown and DBH of the metasequoia,and the correlation coefficients with the measured values were 0.9706 and 0.9751,respectively.The method of extracting tree height by UAV and TLS fusion point cloud can be applied to the forest farm to accurately evaluate the value of forest assets.After the ground-based lidar scanning,the TLS point cloud model of the lower canopy of the canopy is obtained,and the UAV point cloud of the upper canopy can be collected by the drone every year.After registration and fusion,the model can be updated and the height of the tree can be automatically extracted.Through measuring the diameter of the breast,the annual growth of wood can be calculated.The method can realize the dynamic monitoring of the forest growth status,which provides scientific suggestions for the management of forest resources in small and medium-sized forest farms.