CN114821293A - Prototype spectral set generation method based on super-pixels - Google Patents
Prototype spectral set generation method based on super-pixels Download PDFInfo
- Publication number
- CN114821293A CN114821293A CN202210220828.2A CN202210220828A CN114821293A CN 114821293 A CN114821293 A CN 114821293A CN 202210220828 A CN202210220828 A CN 202210220828A CN 114821293 A CN114821293 A CN 114821293A
- Authority
- CN
- China
- Prior art keywords
- image
- wave bands
- wave
- images
- pixel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
Landscapes
- Image Processing (AREA)
Abstract
The invention discloses a prototype spectrum set generation method based on superpixels, which belongs to the field of hyperspectral image processing and comprises the following specific steps: firstly, collecting spectral bands for a certain area, and taking each band as an image to form an image set X; then, randomly selecting 40% of wave bands, and averaging pixel points in the image corresponding to each wave band to form an image Y; secondly, calculating the SSIM structural indexes of the images in the image set X and the images Y one by one to serve as scores of the images, and arranging the scores in a descending order; selecting the previous 40% wave band to regenerate the image Y again, and repeatedly calculating the score of each image in the image set X until the wave band is stable; and finally, selecting the first three wave bands with the highest score, inputting the wave bands into a superpixel segmentation algorithm to extract superpixel small blocks at fixed intervals, taking the average spectrum of each superpixel small block as an initial clustering center of a k-means clustering algorithm, and clustering to obtain a prototype spectrum set of the region. The invention reduces the calculation amount in the clustering process and stabilizes the clustering result.
Description
Technical Field
The invention belongs to the field of hyperspectral image processing, and particularly relates to a prototype spectrum set generation method based on superpixels.
Background
In recent years, with the wide application of hyperspectral sensors, hyperspectral image processing technology is developed, and the hyperspectral image processing technology plays an increasingly important role in the fields of military, agriculture, forestry, environmental protection and industry.
In the hyperspectral remote sensing technology, a ground object standard spectrum is an important basis of the hyperspectral remote sensing technology, and a spectrum curve is a main basis of waveband selection and target classification. However, the standard spectrum of the ground objects is difficult to be effectively extracted due to various factors such as large influence of the hyperspectral image under the illumination condition, less labeled hyperspectral image, slight difference of the standard spectrum of the ground objects under different environments and the like.
The super-pixels are areas formed by pixel points with adjacent positions and similar characteristics such as color, brightness and the like, and the areas generally cannot damage boundary information of objects in the image. In the hyperspectral image, the superpixel small blocks mainly refer to homogeneous areas with relatively close spectral information in the hyperspectral image. The superpixel segmentation algorithm is introduced into a component of a prototype spectrum set, so that on one hand, the whole image can be effectively segmented into nearly pure superpixel small blocks only containing similar pixels according to the homogeneity of pixel points; on the other hand, the super-pixel small blocks replace the original pixels to perform subsequent calculation, so that the calculation complexity of a subsequent clustering algorithm can be greatly reduced.
Shijin Li and Jianbian Qiu put forward a prototype spectrum generation method in 2013, and a corresponding prototype spectrum set is generated by clustering hyperspectral images; on one hand, the method has large calculation amount, and on the other hand, the clustering result is unstable and easily falls into the local optimal condition, so that an effective prototype spectrum set cannot be obtained.
The common region segmentation algorithm capable of constructing the superpixels is mainly a k-means clustering method, and is used for directly carrying out k-means clustering on the hyperspectral images and taking stable clustering centers obtained by clustering as prototype spectrum sets of the hyperspectral images; but this leads to a high computational time complexity; a clustering center generated by k-means clustering is unstable, and the experimental results are inconsistent and poor in robustness; the number of clustering centers is not determined by a definite criterion, and multiple experiments are needed to obtain the best result.
Disclosure of Invention
In order to solve the problem that the ground feature standard spectrum is lacked or difficult to extract, the invention provides a prototype spectrum set generation method based on superpixels.
The prototype spectral set generation method based on the super-pixels comprises the following specific steps:
firstly, collecting wave bands of a certain area by using a hyperspectral instrument, and forming a wave band image set X by independently using each wave band as an image;
X＝{x 1 ,x 2 ,...,x n ,...,x N }; n is the total number of wave bands;
step two, randomly selecting 40% of wave bands, summing pixel points in the image corresponding to each wave band, dividing the sum by the number of the wave bands to obtain the average pixel number to form an image Y;
selecting images in the image set X one by one, and calculating SSIM structural indexes with the images Y respectively to serve as corresponding scores of the images;
for image x n The formula for calculating the SSIM structural index is as follows:
wherein mu X Is the current image x n Mean value of middle pixel points, μ Y Is the mean, σ, of the pixels in the image Y X Is an image x n Standard deviation, sigma, of the middle pixel Y Is the standard deviation, C, of the pixels in the image Y 1 ,C 2 ,C 3 Represents a constant to maintain stability.
And step four, arranging the image scores in a descending order, selecting the first 40% wave bands corresponding to the scores, regenerating the image Y by using the average values of all the pixel points, and returning to the step for repeatedly calculating the scores of all the images in the image set X until the currently selected wave band is the same as the last wave band.
Selecting the first three wave bands with the highest SSIM structure index score from the current stable wave bands, inputting the wave bands into a superpixel segmentation algorithm, and extracting superpixel small blocks with fixed intervals from the segmentation result;
the fixed interval is calculated as: dividing the total number of the super-pixel small blocks by the actual number of the ground objects in the region for acquiring the image;
and step six, taking the average spectrum of the super-pixel small blocks at fixed intervals as an initial clustering center of a k-means clustering algorithm, and clustering the super-pixel small blocks segmented by all wave bands in the region to obtain a prototype spectrum set of the region.
The invention has the advantages that:
1) a prototype spectrum set generation method based on super pixels replaces a ground object standard spectrum with a hyperspectral prototype spectrum set under the unsupervised condition, and solves the problem that the ground object standard spectrum and actual spectrum information are not matched due to the fact that the hyperspectrum is greatly influenced by illumination conditions, the ground object standard spectrum is slightly different in different environments and the like.
2) The prototype spectrum set generation method based on the super-pixels replaces original spectra of tens of thousands of single pixel points with average spectra of small super-pixel blocks, and reduces the calculated amount in the clustering process.
3) The traditional kmeans algorithm is unstable and is easy to fall into a local optimal solution; the invention distributes the clustering centers on the whole image uniformly, the clustering result is stable, and the clustering results basically tend to the optimal solution.
Drawings
FIG. 1 is a flow chart of a method for generating a prototype spectral set based on superpixels in accordance with the present invention;
FIG. 2 is a schematic diagram comparing a prototype spectral curve calculated by the present invention with an existing hyperspectral dataset Indian _ pines for two different surface features;
FIG. 3 is a diagram illustrating the result of the super-pixel tile segmentation of the present invention.
Detailed Description
The following is a more detailed explanation of the present invention.
The invention provides a flowchart of a prototype spectral set generation method based on superpixels, and as the input requirement of a superpixel segmentation algorithm is consistent with the format of a visible image, namely, images of three wave bands are only required, and a hyperspectral image has hundreds of wave bands, the invention provides a method for extracting three wave bands with the best structure by using SSIM structural indexes to improve the effect of the superpixel segmentation algorithm; aiming at the problem that a clustering center generated by k-means clustering is unstable, the method combines the label format of the SLIC superpixel block and provides a measure of adopting an average spectrum of the superpixel blocks with fixed intervals as the clustering center, so that a clustering result is kept stable, and the condition of falling into local optimum is avoided.
The prototype spectral set generation method based on the super-pixels comprises the following specific steps as shown in fig. 1:
firstly, collecting wave bands of a certain area by using a hyperspectral instrument, and forming a wave band image set X by independently using each wave band as an image;
X＝{x 1 ,x 2 ,...,x n ,...,x N }; n is the total number of wave bands;
randomly selecting 40% of wave bands, summing pixel points in the image corresponding to each wave band, dividing the sum by the number of the wave bands to obtain an average pixel number to form an image Y;
selecting images in the image set X one by one, and calculating SSIM structural indexes with the images Y respectively to serve as corresponding scores of the images;
due to the special imaging principle of the hyperspectral image, namely that different wave bands are actually images generated on different narrow bands in the same region, the quality of the wave bands can be determined through the SSIM structural index.
For image x n The formula for calculating the SSIM structural index is as follows:
wherein mu X Is the current image x n Mean value of middle pixel, mu Y Is the mean, σ, of the pixels in the image Y X Is an image x n Standard deviation, sigma, of the middle pixel Y Is the standard deviation, C, of the pixels in the image Y 1 ,C 2 ,C 3 The constant is represented in order to avoid the denominator being 0 and maintain the stability of the formula. Usually take C 1 ＝(K 1 ×L) 2 ，C 2 ＝(K 2 ×L) 2 ，C 3 ＝C 2 And/2 constant to maintain stability.
K 1 ＝0.01，K 2 0.03 percent; l is the image x n The pixel range of (2).
And step four, arranging the image scores in a descending order, selecting the first 40% wave bands corresponding to the scores, regenerating the image Y by using the average values of all the pixel points, and returning to the step for repeatedly calculating the scores of all the images in the image set X until the currently selected wave band is the same as the last wave band.
The step is mainly to propose a noise band and reduce the influence of the noise band on the screening result.
Selecting the first three wave bands with the highest SSIM structural index score from the current stable wave bands, inputting the wave bands into a superpixel segmentation SLIC algorithm to obtain corresponding superpixel segmentation results, and extracting superpixel small blocks with fixed intervals from the segmentation results;
the fixed interval is calculated as: dividing the total number of the super-pixel small blocks by the actual number of the ground objects in the region for acquiring the image;
and step six, taking the average spectrum of the super-pixel small blocks at fixed intervals as an initial clustering center of a k-means clustering algorithm, and clustering the super-pixel small blocks segmented by all wave bands in the region to obtain a prototype spectrum set of the region.
Aiming at the problem that a clustering center generated by k-means clustering is unstable; due to the general rule of ground feature distribution and the super-pixel small blocks generated by the SLIC algorithm, the super-pixel small blocks are relatively regular, and the sequence of the labels is fixed. Therefore, the average spectrum of the super-pixel small blocks at fixed intervals is used as the initial clustering center of the k-means algorithm, and the problem of unstable clustering results is effectively solved by fixing the number of the clustering centers.
Comparing the prototype spectrum curve in the prototype spectrum set calculated by the application with a published hyperspectral data set Indian _ pines, as shown in fig. 2, wherein a is an average spectrum of a ground feature with the ground feature type of 7 in the Indian _ pines data set and a prototype spectrum curve closest to the ground feature, b is an average spectrum of a ground feature with the ground feature type of 13 in the Indian _ pines data set and a prototype spectrum curve closest to the ground feature, a solid line is a real ground feature spectrum curve obtained according to a real ground feature label provided in data, and a dotted line is the prototype spectrum curve of the application; where graph a is the most different and graph b is the least different. It can be obviously seen that the method and the device can obtain a good prototype spectrum set which is extremely close to the spectrum of the real ground object without the help of the label of the real ground object. (the real ground object label is difficult to label, or the label is time-consuming and labor-consuming, and the real ground object label cannot be automatically generated in the actual acquisition process.)
The result of the super-pixel small block segmentation is shown in fig. 3, (an effect graph after the part of the ground object without labels is removed), and it can be seen that the super-pixel small blocks with regular and compact arrangement and well described ground object boundary can be effectively obtained through the method and the device.
Claims (3)
1. A prototype spectral set generation method based on superpixels is characterized by comprising the following specific steps:
firstly, collecting wave bands of a certain area by using a hyperspectral instrument, and forming a wave band image set X by independently using each wave band as an image;
X＝{x 1 ,x 2 ,...,x n ,...,x N }; n is the total number of wave bands;
then, randomly selecting 40% of wave bands, summing pixel points in the image corresponding to each wave band, dividing the sum by the number of the wave bands to obtain an average pixel number to form an image Y;
then, selecting the images in the image set X one by one, and calculating SSIM structural indexes with the images Y respectively to serve as corresponding scores of all the images; arranging according to the image scores in a descending order, selecting the first 40% wave bands corresponding to the scores, regenerating the image Y by using the average values of all the pixel points, and repeatedly calculating the scores of all the images in the image set X until the currently selected wave band is the same as the last wave band;
finally, selecting the first three wave bands with the highest SSIM structure index score from the current stable wave bands, inputting the wave bands into a superpixel segmentation algorithm, and extracting superpixel small blocks with fixed intervals from the segmentation result; and (3) taking the respective average spectrum of the super-pixel small blocks at fixed intervals as an initial clustering center of a k-means clustering algorithm, and clustering the super-pixel small blocks segmented by all wave bands in the region to obtain a prototype spectrum set of the region.
2. A method of generating a prototype spectral set based on superpixels, according to claim 1, wherein: the image score is calculated by the following process:
for image x n The formula for calculating the SSIM structural index is as follows:
wherein mu X Is the current image x n Mean value of middle pixel, mu Y Is an imageMean, σ, of pixels in Y X Is an image x n Standard deviation, sigma, of the middle pixel Y Is the standard deviation, C, of the pixels in the image Y 1 ,C 2 ,C 3 Represents a constant to maintain stability.
3. A method of generating a prototype spectral set based on superpixels, according to claim 1, wherein: the fixed interval is calculated as: the total number of super-pixel patches is divided by the actual number of surface features in the area where the image was acquired.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210220828.2A CN114821293A (en) | 2022-03-08 | 2022-03-08 | Prototype spectral set generation method based on super-pixels |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210220828.2A CN114821293A (en) | 2022-03-08 | 2022-03-08 | Prototype spectral set generation method based on super-pixels |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114821293A true CN114821293A (en) | 2022-07-29 |
Family
ID=82529678
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210220828.2A Pending CN114821293A (en) | 2022-03-08 | 2022-03-08 | Prototype spectral set generation method based on super-pixels |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114821293A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117746079A (en) * | 2023-11-15 | 2024-03-22 | 中国地质大学（武汉） | Clustering prediction method, system, storage medium and equipment for hyperspectral image |
| CN118655098A (en) * | 2024-08-19 | 2024-09-17 | 陕西省环境监测中心站 | Water quality parameter classification inversion method and system based on spectral reflection |
-
2022
- 2022-03-08 CN CN202210220828.2A patent/CN114821293A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117746079A (en) * | 2023-11-15 | 2024-03-22 | 中国地质大学（武汉） | Clustering prediction method, system, storage medium and equipment for hyperspectral image |
| CN117746079B (en) * | 2023-11-15 | 2024-05-14 | 中国地质大学（武汉） | Clustering prediction method, system, storage medium and equipment for hyperspectral image |
| CN118655098A (en) * | 2024-08-19 | 2024-09-17 | 陕西省环境监测中心站 | Water quality parameter classification inversion method and system based on spectral reflection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Dian et al. | Hyperspectral image super-resolution via non-local sparse tensor factorization | |
| CN110399909B (en) | Hyperspectral image classification method based on label constraint elastic network graph model | |
| CN110717354A (en) | Superpixel classification method based on semi-supervised K-SVD and multi-scale sparse representation | |
| Huang et al. | Local binary patterns and superpixel-based multiple kernels for hyperspectral image classification | |
| Ramaiah et al. | De-duplication of photograph images using histogram refinement | |
| CN114821293A (en) | Prototype spectral set generation method based on super-pixels | |
| CN112766291A (en) | Matching method of specific target object in scene image | |
| CN106097252B (en) | High spectrum image superpixel segmentation method based on figure Graph model | |
| CN111798526B (en) | Method and system for rapidly extracting dominant colors of color images based on clustering space mapping | |
| CN116842210B (en) | Textile printing texture intelligent retrieval method based on texture features | |
| CN113052859A (en) | Super-pixel segmentation method based on self-adaptive seed point density clustering | |
| CN110070545B (en) | Method for automatically extracting urban built-up area by urban texture feature density | |
| CN110569859A (en) | Color feature extraction method for clothing image | |
| KR20010011701A (en) | A method for extracting characterized vectors in color image and a retrieval method using the characterized vectors | |
| Yadav et al. | Improved color-based K-mean algorithm for clustering of satellite image | |
| CN108960276A (en) | The sample for promoting spectrum picture supervised classification performance expands and consistency discrimination method | |
| Kuzovkin et al. | Descriptor-based image colorization and regularization | |
| CN113850748A (en) | Point cloud quality evaluation system and method | |
| CN117746079B (en) | Clustering prediction method, system, storage medium and equipment for hyperspectral image | |
| CN118230166A (en) | Corn canopy organ identification method and canopy phenotype detection method based on improved Mask2YOLO network | |
| CN110599518B (en) | Target tracking method based on visual saliency and super-pixel segmentation and condition number blocking | |
| CN113344047A (en) | Platen state identification method based on improved K-means algorithm | |
| CN109840914B (en) | Texture segmentation method based on user interaction | |
| CN109753912B (en) | Multispectral palm print matching method based on tensor | |
| CN111401485A (en) | Practical texture classification method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |