Unmanned aerial vehicle hyperspectral pine wood nematode disease rapid detection method based on decoupling risk estimation

Abstract

The invention relates to an unmanned aerial vehicle hyperspectral pine wood nematode disease rapid detection method based on decoupling risk estimation. According to the method, through risk estimation of category decoupling, a threshold determination problem is converted into a risk estimation problem, and a threshold adjustment step is avoided; meanwhile, a full convolutional neural network is introduced into a single classification framework, and a dependency relationship between pixels with relatively long distances in an image is captured through utilization of global space information; asalt and pepper noise phenomenon frequently occurring in an unmanned aerial vehicle image detection result is relieved; in addition, compared with a single classification method based on image blocks, the method provided by the invention has the advantage that the reasoning speed is higher. The method can be used for detecting the pine wood nematode disease without manual intervention.

Description

technical field [0001] The invention belongs to the field of UAV remote sensing image technology processing, and in particular relates to a method for rapid detection of UAV hyperspectral pine wood nematode disease based on decoupling risk estimation. Background technique [0002] Pine wood nematode disease is called the "cancer" of pine trees. Pine trees will show symptoms of withering about 40 days after being infected with the disease, and will die within 2 to 3 months, and will cause large-scale deforestation within 3 to 5 years. disaster. Finding the location of dead pine trees and cutting down diseased pine trees quickly is one of the effective means to stop the spread of pine wood nematode. However, manual monitoring of the location of dead pine trees is time-consuming and laborious, and it is difficult to quickly extract the location of a large area of ​​dead pine trees. The change in the canopy reflectance of diseased pine trees provides the possibility of hyperspe...

AI Technical Summary

Problems solved by technology

[0003] At the same time, the detection of pine wood nematode dead pine trees based on UAV hyperspectral remote sensing images is also a difficult task: First, the traditional hyperspectral image detection algorithm cannot directly obtain the location of dead pine trees, and it is necessary to determine the threshold after additional , the category of the image pixel can be obtained, and the location of the dead pine tree can be obtained directly without threshold setting, which is essentially a more difficult problem of single classification of remote sensing images

Second, traditional remote sensing image detection methods are based on manual features, and the ability to identify diseased pine trees in complex scenes is limited

Third, the spatial heterogeneity and spectral variability brought about by the high spatial resolution of UAV data lead to the intensification of the phenomenon of the same object with different spectra and the same spectrum with different objects, and the detection method that only uses local spatial information leads to the existence of Obvious "salt and pepper noise"

The above problems limit the application of hyperspectral image detection algorithm in the detection of pine wood nematode disease

Images