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

Self-learning based hyperspectral image and visible image fusion classification method

A hyperspectral image and classification method technology, applied in image enhancement, image data processing, character and pattern recognition, etc., can solve time-consuming and labor-intensive problems

Active Publication Date: 2015-11-11
HARBIN INST OF TECH
View PDF3 Cites 9 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to solve the problem that the existing active learning algorithm ignores the determination process of the candidate sample set, and the time-consuming and labor-intensive problem of manually marking the sample with the largest amount of information in the candidate sample set, and proposes a self-learning based Fusion classification method of hyperspectral image and visible light image

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
  • Self-learning based hyperspectral image and visible image fusion classification method
  • Self-learning based hyperspectral image and visible image fusion classification method
  • Self-learning based hyperspectral image and visible image fusion classification method

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach 1

[0035] The fusion classification method of hyperspectral images and visible light images based on self-learning in this embodiment, as for the self-learning method based on high-resolution image segmentation, high-resolution images provide fine spatial and structural information of ground objects, making the traditional pixel-based The classification method of spectral features cannot obtain satisfactory results. With the improvement of resolution, the continuity and uniformity of ground objects have changed, and the same ground objects often reflect different spectral features, that is, the phenomenon of "same object with different spectrum". This phenomenon causes the noise phenomenon of spots and holes in the classification cartography, so an object-oriented classification method is formed. Object-oriented is based on image segmentation, which divides the high-resolution image into multiple regions of different sizes composed of the same ground objects, and each region is c...

specific Embodiment approach 2

[0056] The difference from Embodiment 1 is that in this embodiment, based on the self-learning hyperspectral image and visible light image fusion classification method, step 8 eliminates the minimum Euclidean distance The process of the corresponding candidate samples is, the candidate sample set X C The smallest Euclidean distance between the middle and the given threshold δ Candidate samples corresponding to the same candidate samples are grouped into a set which is: X ^ C = X C ∩ { x i U | d i j min ≤ δ } - - - ( 5 ) ,

specific Embodiment approach 3

[0057] The difference from the first or second specific embodiment is that, in the self-learning-based hyperspectral image and visible light image fusion classification method of this embodiment, step ten uses the active learning model to select from the optimized candidate sample set The process of selecting the sample with the largest amount of information in is,

[0058] Hyperspectral images have the property of "map-spectrum integration", so they are widely used in the research of remote sensing ground object classification, etc. However, hyperspectral images generally have the characteristics of large data volume, high redundancy between bands, and serious mixed pixels. The classification method can achieve better classification results only when there are enough training samples. However, the labeling of training samples usually requires a lot of cost to obtain. Therefore, the active learning classification method is used to iteratively increase the training samples in...

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

A self-learning based hyperspectral image and visible image fusion classification method belongs to the field of hyperspectral image small sample classification. Existing active learning algorithm ignores a determination process of a candidate sample set, and a sample with the maximum information content in a manually labeled candidate sample set has the time-consuming and labor-consuming problem. The self-learning based hyperspectral image and visible image fusion classification method comprises the following steps: firstly, object tags of lots of unlabeled samples are obtained; then, a candidate sample set is determined according to the object tags and category tags. Thus, tags of samples are obtained while candidate samples are determined, and manual labeling process is avoided. The method has an advantage of raising classification accuracy.

Description

technical field [0001] The invention relates to a fusion classification method of hyperspectral images and visible light images based on self-learning. Background technique [0002] Active learning has been extensively studied for its promising performance in the few-shot classification problem of hyperspectral images. However, the research focus of the existing active learning algorithms is how to select the samples with the largest amount of information from the determined candidate sample set for manual labeling and add them to the training set, ignoring the determination of the candidate sample set, and the process of manually labeling candidate samples is time-consuming strenuous. In addition, classic active learning algorithms cannot achieve the collaborative classification of hyperspectral images and visible images well. Contents of the invention [0003] The purpose of the present invention is to solve the problem that the existing active learning algorithm ignor...

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
IPC IPC(8): G06T5/50G06K9/62
Inventor 张钧萍陆小辰李彤
Owner HARBIN INST OF TECH