Prediction Of Growth Parameters For Individual Trees Using Fuzzy Neural Networks: A Research Study

Airborne lidar technology has widespread applications in the acquisition of forest parameters for large-scale forest resources surveys,national timber strategic reserves,carbon sink assessments,and related needs.In the field of forestry investigation,forest parameters are categorized into forest scale and single wood scale.The former reflects the overall distribution state of forest parameters in the study site,while the latter represents specific parameters of individual trees.This study focuses on various rubber tree plantations in Hainan,where different rubber tree varieties make it difficult to excavate forest scale parameters.While the single-wood structural parameters can be acquired directly from airborne lidar when the point cloud density is sufficient,complex forest parameters,such as chest diameter and lumber volume,are difficult to measure due to the angle of view shielding.Real-time acquisition of forest growth parameters is beneficial for the cultivation of large-scale plantation forests.In this study,the latest artificial intelligence technology is applied to the cloud of onboard laser point in woodland.The main objectives of this study are as follows:1.This research establishes a pre-processing process,which includes ground point cloud separation,point cloud denoising,and data enhancement,to address the impact of acquisition angle and natural environment on the obtained point cloud from airborne lidar,thus laying the foundation for this research.2.In this study,an improved single-wood segmentation method is proposed by combining the canopy height model and point cloud clustering.The elevation information is enhanced based on distance measurement and coronal variability to improve the correct rate of single-wood vertex recognition.Additionally,the algorithm run time is shortened by pre-partitioning the point cloud based on single wood vertices.The final single-wood segmentation is completed through the calculation method of bilateral distance judgment value.3.This study extracted forest parameters from single-wood point clouds,which directly provide characteristics such as tree height,east-west crown,north-south crown,crown product,and point cloud density.To calculate the slice volume by convex packet edge contour fitting,a point cloud adaptive slicing method is proposed.Furthermore,this paper analyzes the correlation between complex forest parameters,such as bra-diameter lumber,and the characteristics of single-wood point clouds.The analysis is verified by various regression models.4.This research proposes an optimized fuzzy neural network model that combines adaptive algorithms to determine the model structure and constructs an attention mechanism to identify dead trees.A multi-parameter autonomous optimization module is added based on the pigeon swarm optimization algorithm to establish an accurate prediction model for breast diameter and volume.Experimental results show that compared to classical backpropagation and radial basis neural networks,the proposed model has a 4-6% higher correlation in the inversion of tree parameters.This provides quantitative decision-making and data support for the cultivation and growth of rubber trees of different varieties.