Forest crown breadth recognition method applied to onboard field hyperspectral image

Abstract

The invention discloses a forest crown breadth recognition method applied to an onboard field hyperspectral image and belongs to the field of forestry remote sensing data processing and information extraction. The method is technically characterized in that on the basis of traditional support vector machine classification based on a spectrum, guided filtering is introduced on the basis of completing support vector machine classification and forming various ground object initial probability graphs, the guided filtering is utilized to perform texture and edge information optimization on a classification initial probability graph, categories of all pixel points after optimization are divided according to the maximum probability principle, categories corresponding to trees are extracted after final classification is completed, forest edge vectors are formed, and precise recognition of forest crown breadth is completed.

Description

1. Technical field [0001] The invention relates to a forest tree crown width identification method based on spatial spectrum classification, in particular to a forest tree crown width identification method applied to airborne near-ground hyperspectral image data, which is suitable for forest stands with complex tree crown width structures and belongs to forestry The field of hyperspectral remote sensing data processing technology. 2. Technical background [0002] Airborne near-Earth imaging hyperspectral data is a remote sensing technology that uses a UAV equipped with a hyperspectral camera to obtain hyperspectral images of ground objects. Airborne near-Earth imaging hyperspectral data has the characteristics of high spatial resolution and high spectral resolution at the same time, and the data acquisition time is flexible, and high temporal resolution data can be acquired in time according to requirements. hot spot. In addition, with the continuous development of remote ...

AI Technical Summary

Problems solved by technology

[0005] (1) In the case of complex forest structure, the existing classification methods cannot make full use of the effective information of the near-earth hyperspectral data, and the phenomenon of misclassification and too roughness is easy to occur when extracting the range of forest crown width

[0006] (2) Airborne near-Earth hyperspectral imagery has a large amount of data, and the existing classification algorithms are complex. The classification process runs slowly and takes a long time. The requirements for basic equipment are high, and small-scale data processing takes a lot of time.

Images