Depth belief network-based neighborhood weighted averaging hyperspectral image classification method

A deep belief network, hyperspectral image technology, applied in instrument, character and pattern recognition, scene recognition and other directions, can solve the problem of increasing pre-training and fine-tuning computing time, achieve low computer performance requirements, fast running speed, and optimize data The effect of dimension

Active Publication Date: 2017-02-15
HARBIN INST OF TECH
View PDF4 Cites 9 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to solve the shortcomings of the existing classification methods that increase the calculation time of pre-training and fine-tuning compared with pure spectral data, and propose a neighborhood weighted average hyperspectral image classification method based on deep belief networks

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
  • Depth belief network-based neighborhood weighted averaging hyperspectral image classification method
  • Depth belief network-based neighborhood weighted averaging hyperspectral image classification method
  • Depth belief network-based neighborhood weighted averaging hyperspectral image classification method

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach 1

[0023] Specific implementation mode one: combine figure 1 To illustrate this embodiment, a specific process of a neighborhood weighted average hyperspectral image classification method based on a depth belief network (Depth Belief Networks, DBN) in this embodiment is as follows:

[0024] Step 1, extracting the spatial information of the hyperspectral raw data to obtain the spatial information of the hyperspectral raw data;

[0025] Step 2: Neighborhood weighting is performed on the spatial information of the hyperspectral raw data to obtain the processed hyperspectral data, namely the matrix M;

[0026] Step 3. Using the matrix M as the training set, the deep belief network is used for classification training.

specific Embodiment approach 2

[0027] Embodiment 2: The difference between this embodiment and Embodiment 1 is that in the first step, the hyperspectral raw data is extracted with spatial information to obtain the spatial information of the hyperspectral raw data; the specific process is:

[0028] First of all, because the PCA transformation will lose part of the nonlinear information, the PCA transformation is not performed on the hyperspectral image here, and the small square neighborhood is directly extracted for each pixel.

[0029] When the pixels to be classified in the hyperspectral raw data are not located at the edge of the hyperspectral remote sensing image, that is, there is a 5*5 square neighborhood around the pixels to be classified in the hyperspectral raw data, the pixels to be classified in the hyperspectral raw data Extract the spatial information of the pixels in the neighborhood of a 5*5 square, and obtain the spatial information of the original hyperspectral data;

[0030] When the pixel...

specific Embodiment approach 3

[0032] Specific embodiment 3: The difference between this embodiment and specific embodiments 1 or 2 is that in the step 2, the spatial information of the hyperspectral raw data is subjected to neighborhood weighting to obtain the processed hyperspectral data, namely the matrix M; specifically The process is:

[0033] The so-called neighborhood weighting method takes the pixel to be classified as the center point, and performs weighted average according to the distance between the pixel in the square neighborhood and the center point and the number of pixels;

[0034] Step 21, grouping pixels in a square neighborhood of a size of 5*5 around the pixel to be classified according to the pixel distance from the center point in a square neighborhood of a size of 5*5 around the pixel to be classified;

[0035] Step 22, determining the weight coefficient and the overall weight coefficient of each group of pixels;

[0036] Step two and three, after summing the internal pixels of each...

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 present invention relates to a depth belief network-based neighborhood weighted averaging hyperspectral image classification method. The objective of the invention is to the problem of the increase of operation time of pre-training and fine-tuning of an existing classification classification compared with pure spectral data. The depth belief network-based neighborhood weighted averaging hyperspectral image classification method includes the following steps that: step 1, spatial information extraction is performed on original hyperspectral data, so that the spatial information of the original hyperspectral data can be obtained; step 2, neighborhood weighting is performed on the spatial information of the original hyperspectral data, so that processed hyperspectral data can be obtained, namely, a matrix M is obtained; and step 3, with the matrix M adopted as a training set, a depth belief network is adopted to carry out classification training. The method of the present invention is applied to the image classification field.

Description

technical field [0001] The invention relates to a neighborhood weighted average hyperspectral image classification method based on a deep belief network. Background technique [0002] With the continuous development of remote sensing technology, the spectral information of hyperspectral images is very rich, and each pixel has information of hundreds of wavelengths. But it was found that traditional methods for multispectral images are not suitable for hyperspectral images. The main contradiction is that traditional methods cannot handle such high dimension of spectral data well. In order to solve this problem, in the past ten years, new methods have been introduced into hyperspectral image classification. Among them, the training cost of traditional methods is high, and the classification accuracy often decreases with the increase of feature dimension. Hughes phenomenon, or the curse of dimensionality. With the rapid development of deep learning software and hardware tech...

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): G06K9/62G06K9/00
CPCG06V20/13G06F18/24137G06F18/214
Inventor 林连雷杨京礼宋欣益董弘健
Owner HARBIN INST OF TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap