Unlock instant, AI-driven research and patent intelligence for your innovation.

Rotating forest hyperspectral image classification method based on weighting

A technology of hyperspectral images and classification methods, applied in the fields of instruments, character and pattern recognition, computer components, etc., can solve the problems of ignoring potential, increasing the computational complexity of algorithms, and reducing the integration performance of algorithms, so as to improve the integration performance, improve the The effect of classification accuracy

Pending Publication Date: 2021-02-02
XIDIAN UNIV
View PDF8 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, since the RoF algorithm assigns the same weight to all training samples, it ignores the potential of providing important information samples
In addition, these algorithms generate base classifiers independently, some of which not only increase the computational complexity of the algorithm, but also reduce the integration performance of the algorithm

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Rotating forest hyperspectral image classification method based on weighting
  • Rotating forest hyperspectral image classification method based on weighting
  • Rotating forest hyperspectral image classification method based on weighting

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0020] Hyperspectral images contain rich spectral information and can effectively reflect the information of imaging targets. Therefore, they are widely used in precision agriculture, environmental monitoring, military reconnaissance and other fields. In these applications, hyperspectral image classification is one of the important links, and the ultimate goal of classification is to accurately give each pixel in the image a unique category identifier. The RoF algorithm is one of many classification algorithms. The RoF algorithm generates a sparse rotation matrix by using a feature extraction algorithm, and projects the original image into a different coordinate system, so that the constructed base classifier has a strong difference. Compared with algorithms such as bagging, AdaBoost and RFs, the RoF algorithm can improve the classification accuracy of hyperspectral image samples. However, in the RoF algorithm, all training samples are given the same weight, ignoring the pote...

Embodiment 2

[0032] Based on weighted rotating forest hyperspectral image classification method with embodiment 1, the establishment in step (4) is based on weighted rotating forest model, see figure 2 , including the following steps:

[0033] (4a) Initialize the decision tree base classifier: introduce the rotation forest model, the basic structure of the introduction rotation forest model is to assume that the weighted rotation forest model is composed of T decision tree base classifiers, and set the sequence number of the decision tree base classifier as t, t=1,2,...,T, T decision tree base classifiers are arranged in sequence, and trained according to the sequence, initialize the sequence number t=1 of the decision tree base classifier, and start the iteration of training the decision tree base classifier . After initializing the serial number of the decision tree base classifier, the diversity training sample set, the feature subset, the diversity training sample subset, and the rot...

Embodiment 3

[0046] Weighted rotation forest hyperspectral image classification method based on embodiment 1-2, update sample weight W(x described in step (4h) i ):

[0047]

[0048]

[0049] Among them, t represents the sequence number of the decision tree base classifier after current training, q represents the sequence number of the trained decision tree base classifier, q=1,2,...,t, ξ q (x i ) represents the qth trained decision tree base classifier ξ q For sample x i The classification result of Y t (x i ) represents the diversity training sample set S t Corresponding to sample x in i Tag of.

[0050] The RoF algorithm has the problem of treating all training samples as equal and ignoring the potential of samples containing important information. This invention provides an improved technical solution. By designing a dynamic weighting function, the potential of samples containing important information is tapped and the samples are weighted. . In the present invention, th...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a rotating forest hyperspectral image classification method based on weighting, and solves the problems of low hyperspectral image classification precision and low classification model integration performance. According to the scheme, the method includes dividing hyperspectral image samples into a training set and a test set; initializing a training set sample weight, and multiplying the training set sample weight by a training set corresponding sample to obtain a weighted training set; training a decision tree base classifier and obtaining a weighted training set classification result; establishing a rotating forest model based on weighting; and putting the test set into a weighting-based rotating forest model to obtain a final classification result of the hyperspectral image sample. According to the invention, samples containing important information are mined by designing the dynamic weighting function, the weighted training set classification result of the generated decision tree base classifier is substituted into the current decision tree base classifier to be trained, the classification precision and the model integration performance are improved, andthe method can be used for land classification of hyperspectral images.

Description

[0001] The invention belongs to the technical field of image processing, and mainly relates to remote sensing image processing, in particular to a weighted rotation forest hyperspectral image classification method. In particular, remote sensing classification methods that involve mining important samples can be used for land classification in hyperspectral images. Background technique [0002] Classification is one of the main tasks of remote sensing information processing. Classification of hyperspectral data is generally more difficult than other remote sensing images because of the high feature-to-sample ratio of hyperspectral data and the presence of redundant information in feature sets. Although most learning systems face an intractable problem known as the "curse of dimensionality", studies have demonstrated the successful application of classifier ensemble techniques for hyperspectral classification. Ensemble learning is an effective way to develop accurate classifica...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62
CPCG06F18/24G06F18/214
Inventor 冯伟董淑仙全英汇钟娴童莹萍
Owner XIDIAN UNIV