Hyperspectral image classification method based on data augmentation and classifier fusion
A hyperspectral image and classification method technology, applied in the field of hyperspectral image classification, can solve problems such as limiting the accuracy of hyperspectral image classification, avoid local minimum and overfitting problems, achieve good classification results, and reduce the impact of randomness Effect
Pending Publication Date: 2019-07-05
NORTHWESTERN POLYTECHNICAL UNIV
View PDF9 Cites 8 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
The above two methods are only considered from a single direction, and are not jointly designed from the perspective of data augmentation and classifier fusion, thus limiting the accuracy of hyperspectral image classification
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0017] The present invention will be further described with reference to the following examples. The present invention includes but is not limited to the following examples.
[0018] The present invention provides a hyperspectral image classification method based on data augmentation and classifier fusion. The specific process is as follows:
[0019] 1. Data preprocessing
[0020] Generally speaking, hyperspectral 3D image data can be expressed as Among them, r represents the number of rows, c represents the number of columns, and b represents the spectral dimension. For the convenience of presentation, χ can also be transposed into a two-dimensional matrix, that is, the hyperspectral image data is expressed as among them, Indicates the i-th sample data, i=1, 2,...,n, n=r×c is the total number of samples. At the same time, the category label of image data X is expressed as Is the one-hot label of the i-th sample data, i=1, 2,...,n, and L represents the total number of categories...
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
Login to View More Abstract
The invention provides a hyperspectral image classification method based on data augmentation and classifier fusion. The method comprises the following steps of firstly, preprocessing the hyperspectral image data, and randomly selecting to obtain an initial training sample; then establishing a data mixing model to expand the training samples with labels exponentially; then, training a convolutional neural network-based classifier by using the expanded data set; and finally, fusing the results of the plurality of classifiers by using a voting strategy to obtain a final classification result. According to the method, the problems of the local minimization and the over-fitting in convolutional neural network training can be avoided, the influence of randomness is reduced, and a better classification effect can be obtained under the condition of the limited labeled training samples.
Description
Technical field [0001] The invention belongs to the technical field of hyperspectral image processing, and specifically relates to a hyperspectral image classification method based on data augmentation and classifier fusion. Background technique [0002] At present, hyperspectral images have been widely used in the fields of environmental monitoring, resource exploration and feature recognition. Hyperspectral image classification is one of the important applications. In practical applications, annotating hyperspectral images requires the experience of geologists, which usually requires a lot of time and is costly. Therefore, labeled samples in hyperspectral images are often limited. The method based on deep neural network has been widely used in hyperspectral image classification, and has achieved good classification results. But when there are fewer labeled training samples, the deep neural network-based method will become over-fitting, resulting in a decline in the 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
Login to View More IPC IPC(8): G06K9/62G06N3/04G06N3/08
CPCG06N3/084G06N3/045G06F18/24G06F18/214
Inventor 魏巍张艳宁王聪张磊
Owner NORTHWESTERN POLYTECHNICAL UNIV