An efficient feature extraction and tree species recognition method oriented to a tree laser point cloud

Abstract

The invention discloses an effective feature extraction and tree species identification method oriented to a tree laser point cloud, which comprises the following steps of obtaining full-coverage point cloud data of a target tree; denoising the full coverage point cloud data of the target tree to remove anomalies; according to the point cloud data obtained after noise reduction, analyzing the treecharacteristics of three kinds of target trees respectively; extracting the optimal feature parameter group based on the relative clustering feature of trees; extracting the optimal feature parameters based on point cloud distribution features; extracting the optimal feature parameters based on apparent features of trees; combining and inputting the optimal feature parameters of three tree species to SVM classifier as variables to classify tree species. The invention achieves higher tree species classification accuracy, provides a powerful tool for obtaining more accurate forest tree speciesdistribution, reduces the high cost of field entity investigation, is time-consuming and laborious, and reduces the error caused by manual interpretation.

Description

technical field [0001] The invention relates to the technical field of tree species classification, in particular to an effective feature extraction and tree species identification method for tree laser point clouds. Background technique [0002] As the main body of renewable natural resources and terrestrial ecosystems on the earth, forests provide abundant material resources for human survival and development, and play an important role in maintaining ecological processes and ecological balance. The correct identification of forest species is the basis and basis for utilizing and protecting forest resources. Over the past forty years, advances in remote sensing technology have enabled tree species classification within several sensor types. Most tree species classification methods at this stage rely on some high-cost, time-consuming, laborious field entity survey methods or low-accuracy aerial photos for manual interpretation. [0003] At present, many studies have explo...

AI Technical Summary

Problems solved by technology

Most of the tree species classification methods at this stage rely on some high-cost, time-consuming and laborious field entity survey methods or low-accuracy aerial photos for manual interpretation

[0003] At present, many studies have explored the problem of tree species classification from various aspects, such as using the internal signal characteristics of the canopy as additional features to evaluate the robustness of tree internal parameters and size; using the normalized difference vegetation index as a feature parameter in tree species diversity Classify tree species in relatively large cases; use spectral variable selection combined with PLS-DA to improve random tree species classification accuracy without classifier constraints; use average weighting algorithm to extract canopy pixel-weighted spectra in hyperspectral images And carry out effective analysis; propose a semi-supervised support vector machine classifier for tree species classification at the canopy level; detect differences in crown aggregation skewness attributes between tree species; formulate a tree species classification program based on static TLS; however, specific to the classification of specific tree species However, there are relatively few studies on LiDAR data extraction features based on single tree, and there are few reports on optimal feature extraction with a certain algorithm.

Images