CN111898423A - Morphology-based multisource remote sensing image ground object fine classification method - Google Patents
Morphology-based multisource remote sensing image ground object fine classification method Download PDFInfo
- Publication number
- CN111898423A CN111898423A CN202010566343.XA CN202010566343A CN111898423A CN 111898423 A CN111898423 A CN 111898423A CN 202010566343 A CN202010566343 A CN 202010566343A CN 111898423 A CN111898423 A CN 111898423A
- Authority
- CN
- China
- Prior art keywords
- image
- images
- tree
- hyperspectral
- remote sensing
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2137—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on criteria of topology preservation, e.g. multidimensional scaling or self-organising maps
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
Abstract
The invention discloses a morphology-based method for finely classifying ground objects of multi-source remote sensing images. Performing dimensionality reduction on the hyperspectral image; carrying out infrared image and dimensionality reduction on the hyperspectral image; fusing morphological characteristics; and (6) classifying. The method utilizes the difference and complementarity between the hyperspectral image and the infrared image, extracts and fuses the features between the two types of images by using a morphological method, and improves the ground feature classification precision of the hyperspectral image.
Description
Technical Field
The invention provides a method for fusing a hyperspectral image and an infrared image for collaborative classification based on morphological characteristics, which combines the spectrum unification characteristic of hyperspectrum and the advantage that the infrared image is not easily influenced by shadows and weather conditions, makes full use of the difference and complementarity between the two images, and improves the precision of hyperspectral ground object classification.
Background
With the continuous improvement of remote sensing imaging technology, software and hardware integration technology system is continuously mature, and the carrying mode is continuously broken through. The spectral resolution and the spatial resolution of the existing remote sensing image are greatly improved. The remote sensing image can capture rich ground feature information, and better ground feature semantic or ground feature logical interpretation is realized, so that the remote sensing image has the advantage of a data source in the field of pattern recognition. However, due to different imaging principles of the sensor and other limitations, the image of any single source cannot fully describe the characteristics of the target object, i.e., the geometric, spatial, spectral, etc. aspects of single-source differences and limitations exist. The remote sensing image can be divided into multispectral remote sensing and hyperspectral remote sensing according to the spectral resolution; the remote sensing image can be divided into a visible light remote sensing image, an infrared image and a radar image according to the range of the covered wavelength. The invention mainly considers hyperspectral images and infrared images.
The hyperspectral image has the advantage of map integration, can distinguish different pixels and identify the categories of the pixels according to the spectral curves of the pixels, can perform semantic interpretation and ground object identification according to the spatial information of the shape blocks, and has wide application in the fields of agriculture, oceans, military, food, medicine, meteorology and the like. However, hyperspectral images still have two types of limitations: firstly, the hyperspectral technology can effectively identify ground object materials with different reflectances by means of abundant spectral information, but the expected effect on the identification effect of the materials with similar reflectances is difficult to achieve; secondly, the hyperspectral space information representation force is limited, and the identification effect on materials with the same shape but different reflectivities is not good. In addition, the hyperspectral images are susceptible to factors such as shadows and weather conditions, so that even if the classification effect of the hyperspectral images is improved to a certain extent by comprehensively utilizing the spectral information and the spatial information, the hyperspectral images are still classified only by means of the hyperspectral images. Infrared imaging is a passive imaging technology, achieves the imaging effect by capturing infrared rays radiated inside an object, can generate radiation with different intensities according to different temperatures of a background and a target, contains abundant spectrum and spatial characteristics, and can well identify the target, but information contained in an infrared image is single and is not enough to solve the problem of complex ground object classification. Therefore, in a complex ground object classification task, the single remote sensing data is very limited, and if the accurate acquisition of ground object space information is to be realized so as to achieve the expected classification effect, the classification performance of the hyperspectral images needs to be improved by utilizing the redundancy, the cooperation and the complementarity among multisource images.
The remote sensing image contains rich information such as space, texture and the like, and the morphology is used as a nonlinear and irreversible image analysis technical means based on theories such as integral geometry, geometric operation, geometric probability and the like, so that unimportant information in a target image can be removed through morphological transformation in the image, and important information in the image is reserved. The characteristic information contained in different remote sensing images is different, for example, a hyperspectral image has abundant spectral information and spatial information; the infrared image has abundant background information and context information, and a proper mathematical morphology algorithm needs to be selected to carry out data processing so as to realize the improvement of image classification and recognition effects. In recent years, more and more mathematical Morphology methods are applied to the field of image processing and analysis, and various novel remote sensing image processing technologies, such as Morphological Profile (MP), Attribute Profile (AP), Extinction Profile (EP), and other morphological algorithms, are promoted. Compared with other two algorithms, the EP can effectively extract the spatial information in the hyperspectral image, and the classification performance of the hyperspectrum is improved through the collaborative classification of the spatial-spectral information in the hyperspectral image. On one hand, the EP is based on the max/min-tree which is constructed according to the pixel values in the image, so that the EP algorithm is easily influenced by external factors such as noise, cloud and fog shielding and the like, and the stability of the algorithm is influenced; on the other hand, the processing of the image by the EP needs the expansion and erosion operations, and the image itself and the complementary set thereof need to be subjected to the feature extraction operation, so that the feature dimension obtained by the EP is high, and the Hughes phenomenon is easily caused in the actual classification problem, so that the classification by using the EP extraction features has certain defects. In view of the problems of EP, a new morphological algorithm, namely Local Contour (LCP), has been proposed by the scholars, and the LCP is improved on the EP algorithm level. First, unlike EP, LCP is constructed based on a topology tree, which is constructed based on a containment relationship, for a max/min-tree, where images captured at the same location, when the illumination is different, the constructed tree is different, but the topology tree does not change; and secondly, the LCP only carries out expansion processing on one image, so that compared with the feature dimension obtained by the EP, the LCP has half less characteristic dimension, reduces information redundancy and reduces calculation complexity.
Disclosure of Invention
Aiming at the problems, the invention provides a morphological method-based multisource remote sensing image ground object classification method which mainly uses hyperspectral images. The morphological characteristics of the hyperspectral image and the infrared image are respectively extracted, then a specific fusion method is used for fusing the extracted spatial characteristics of the two data sources, and the fused spatial characteristics are sent to a classifier (such as a Support Vector Machine (SVM)) for classification, so that a final classification result image and classification precision are obtained. The method comprises the following specific steps:
and S1, performing dimensionality reduction processing on the hyperspectral image. The invention adopts Principal Component Analysis (PCA) to reduce the dimension of a hyperspectral image. Different from a conventional optical image, a hyperspectral image contains rich spectral information, the number of wave bands of the hyperspectral image is dozens or even hundreds, but when a morphological method is used for extracting features, only a single-wave-band image can be processed, so that the dimension reduction processing is carried out on the hyperspectral image in the first step, and each wave band of the hyperspectral image after dimension reduction is respectively stored, namely the hyperspectral image is stored as a single-wave-band image. For the infrared image, because the number of the wave bands is 1, the dimension reduction processing is not needed;
and S2, morphological characteristics of the infrared image and the hyperspectral image after dimension reduction (the single-waveband image stored in S1) are respectively extracted. Morphological characteristics of multi-source remote sensing images (infrared images and dimension-reduced hyperspectral images) are extracted by using a morphological algorithm LCP;
further, step 2 specifically includes:
s21, constructing a tree. Whether the hyperspectral image or the infrared image is used, a topology tree needs to be constructed before feature extraction is carried out by using LCP (certainly, when the hyperspectral image is constructed, a single-waveband image stored in S1 is operated, and the infrared image is only required to be operated on an original image), wherein the topology tree is constructed according to the inclusion relation of connected regions in the image and can represent the neighborhood relation (such as inclusion, adjacency, phase separation and the like) of the connected regions in the image. Each node in the tree represents a connected region in the graph, each connected region corresponds to only a single node (i.e., a composite node), the root node represents the widest connected region in the global region, and the leaf node represents the smallest connected region in the local region.
And S22, extinction filtering. After the tree construction is completed, a cut operation needs to be performed on the tree. The pruning operation of LCP is based on extinction filtering. The extinction filtering principle is to keep the branches where the leaf nodes meeting the requirements in the topology tree constructed in S21 through a set threshold (i.e., the number of extinction values), and cut the branches where the leaf nodes not meeting the requirements are located.
And S23, image reconstruction. After the tree is subjected to extinction filtering, a topological tree constructed according to an original multi-source remote sensing image (a single-waveband image and an infrared image stored in S1) is changed, namely, nodes on a plurality of branches are cut, a corresponding connected region is covered by a connected region corresponding to a parent node, the connected region corresponding to each node in the tree is changed, and the connected region represented by one parent node is a combined region of the connected region corresponding to the original parent node and the connected regions of a plurality of child nodes along with the shadow. The images need to be reconstructed according to the changed topology tree, for hyperspectrum, the reconstructed images correspond to the morphological features of the single-band images stored in S1, and for infrared images, the reconstructed images are features of the infrared images. These features retain useful ground feature information.
And S3, fusing morphological characteristics. After morphological characteristics of the multi-source remote sensing image are obtained, the characteristics are required to be fused at a characteristic level, a characteristic stacking method is used, namely, morphological characteristics obtained by subjecting each single-waveband image subjected to hyperspectral dimension reduction to an LCP algorithm and morphological characteristics obtained by subjecting an infrared image to the LCP algorithm are stacked in a 3 rd dimension, and fused morphological characteristics of a multi-source image are obtained;
and S4, classifying. And inputting the morphological characteristics of the fused multi-source remote sensing image into the SVM for classification to obtain a final classification precision and classification result graph.
The method utilizes the difference and complementarity between the hyperspectral image and the infrared image, extracts and fuses the features between the two types of images by using a morphological method, and improves the ground feature classification precision of the hyperspectral image.
Drawings
FIG. 1 is a collaborative classification framework of the present invention;
FIG. 2 is a topology tree construction diagram;
FIG. 3 is a flowchart of LCP extraction features;
FIG. 4 is a pseudo-color image of a hyperspectral image used in an experiment;
FIG. 5 is a thermal infrared image used for the experiment;
Detailed Description
In the block diagram shown in fig. 1, firstly, a hyperspectral image needs to be subjected to dimensionality reduction, the invention uses a Principal Component Analysis (PCA) to perform dimensionality reduction operation, reduces an original image to 2 dimensions, respectively performs morphological feature extraction on the two wave bands and an infrared image, performs feature fusion, and finally inputs the two wave bands and the infrared image into an SVM for classification.
The specific process of extracting features by LCP is shown in fig. 3, and for a single-band image, first, a topology tree is constructed, then, an attribute value and an extinction value of a corresponding node are calculated, extinction filtering is performed according to a given condition, and finally, for the filtered topology tree, image reconstruction is performed, and the whole process only needs to perform expansion operation on the image of each band.
When the extinction filtering in the feature is extracted, an attribute value of each node needs to be calculated, wherein 7 attributes are involved, namely area, height, volume, diagonal line of a bounding box, standard deviation, elongation and compactness. Setting of thresholds in extinction filtering follows mn(n-0, 1,2,3 … s-1) rule, m-3, s-5, i.e. there are 5 thresholds (1,3,9,27,81), so there are 35 LCP features per single band.
Finally, the final classification results obtained by the SVM classifier are given in table 1. The data used in the experiment is the data of a high-score five-numbered satellite (GF5) in China, which is the first high-spectrum comprehensive observation satellite in China. The data is a region collected by GF5 in a certain province, and mainly comprises two groups of data, wherein one group of data is visible light hyperspectral data, the size of the visible light hyperspectral data is 1220 × 973 × 150, and the spatial resolution is 30 meters; one set was thermal infrared data with a size of 1220 x 973 and a spatial resolution of 40 meters. The hyperspectral image is shown in fig. 4, and the thermal infrared image is shown in fig. 5.
As can be seen from table 1, for LCP, the Overall Accuracy (OA), the Average Accuracy (AA), and the Kappa coefficient after fusing the hyperspectral image and the infrared image are all higher than those using the hyperspectral image or the infrared image alone. Besides, the classification accuracy of various types after data fusion can be improved compared with the accuracy of using a hyperspectral image and an infrared image independently, because the hyperspectral data contains abundant spectrums and spatial features, the hyperspectrum has the same shape and the same reflectivity, the material cannot be effectively identified, the thermal infrared image can be used for identifying the target well according to the radiation with different intensities formed by the target and the background surface temperature, and the target can be effectively identified, so that the hyperspectral image can be effectively supplemented. Theoretical and practical results fully prove that in ground feature classification, remote sensing data of two different sources can be mutually supplemented and perfected, so that better ground feature fine identification and classification are realized.
TABLE 1 comparison of LCP-based classification results
Claims (2)
1. A morphology-based method for finely classifying land features of multi-source remote sensing images is characterized by comprising the following steps: the method comprises the following specific steps:
s1, performing dimensionality reduction on the hyperspectral image by adopting a principal component analysis method, and storing each wave band of the hyperspectral image after dimensionality reduction as a single-wave-band image; the infrared image does not need to be subjected to dimension reduction processing;
s2, morphological characteristics of the infrared image and the hyperspectral image after dimensionality reduction are respectively extracted; morphological characteristics of the multi-source remote sensing image are extracted by using a morphological algorithm LCP;
s3, fusing morphological characteristics; after morphological characteristics of the multi-source remote sensing images are obtained, the characteristics are fused at a characteristic level, and the morphological characteristics of each single-waveband image subjected to hyperspectral dimension reduction and obtained through an LCP algorithm and the morphological characteristics of infrared images and obtained through the LCP algorithm are stacked in a 3 rd dimension by using a characteristic stacking method to obtain fused morphological characteristics of the multi-source images;
s4, classifying; and inputting the morphological characteristics of the fused multi-source remote sensing image into the SVM for classification to obtain a final classification precision and classification result graph.
2. The morphology-based multisource remote sensing image ground object fine classification method according to claim 1, characterized in that: the step 2 specifically comprises the following steps:
s21, constructing a tree; whether the image is a hyperspectral image or an infrared image, a topological tree needs to be constructed before feature extraction is carried out by utilizing LCP, and the topological tree is constructed according to the inclusion relation of connected regions in the image and represents the neighborhood relation of the connected regions in the image; each node in the tree represents a connected region in the graph, each connected region only corresponds to a single node, the root node represents the widest connected region in the global region, and the leaf node represents the smallest connected region in the local region;
s22, extinction filtering; after the tree construction is completed, performing cutting operation on the tree; the pruning operation of LCP is based on extinction filtering; through a set threshold, the branches where the leaf nodes meeting the requirements in the topology tree constructed in S21 are located are reserved, and the branches where the leaf nodes not meeting the requirements are located are cut off;
s23, image reconstruction; after the tree is subjected to extinction filtering, a topological tree constructed according to an original multi-source remote sensing image changes, namely nodes on a plurality of branches are cut, a corresponding connected region is also covered by a connected region corresponding to a parent node, the connected region corresponding to each node in the tree changes, and the connected region represented by one parent node is a combined region of the connected region corresponding to the original parent node and the connected regions of a plurality of child nodes along with the shadow; and reconstructing the images according to the changed topology tree, wherein the reconstructed images correspond to the morphological characteristics of the single-waveband images stored in S1 for the hyperspectral images, and the reconstructed images are the characteristics of the infrared images for the infrared images.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010566343.XA CN111898423B (en) | 2020-06-19 | 2020-06-19 | Morphology-based multisource remote sensing image ground object fine classification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010566343.XA CN111898423B (en) | 2020-06-19 | 2020-06-19 | Morphology-based multisource remote sensing image ground object fine classification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111898423A true CN111898423A (en) | 2020-11-06 |
| CN111898423B CN111898423B (en) | 2022-12-23 |
Family
ID=73207504
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010566343.XA Active CN111898423B (en) | 2020-06-19 | 2020-06-19 | Morphology-based multisource remote sensing image ground object fine classification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111898423B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112396125A (en) * | 2020-12-01 | 2021-02-23 | 中国第一汽车股份有限公司 | Classification method, device, equipment and storage medium for positioning test scenes |
| CN112906528A (en) * | 2021-02-05 | 2021-06-04 | 北京观微科技有限公司 | Urban building material classification method fusing multi-source satellite remote sensing data |
| CN113256682A (en) * | 2021-06-04 | 2021-08-13 | 北京理工大学 | Infrared small target detection tracking method based on morphology and pipeline filtering |
| CN115984635A (en) * | 2023-03-21 | 2023-04-18 | 自然资源部第一海洋研究所 | Multi-source remote sensing data classification model training method, classification method and electronic equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180260657A1 (en) * | 2016-11-01 | 2018-09-13 | Shenzhen University | Method and system for extracting and classifying features of hyperspectral remote sensing image |
| CN108985360A (en) * | 2018-06-29 | 2018-12-11 | 西安电子科技大学 | Hyperspectral classification method based on expanding morphology and Active Learning |
| CN109583479A (en) * | 2018-11-07 | 2019-04-05 | 北京市遥感信息研究所 | One kind being based on morphologic high spectrum image feature extracting method |
| CN110569884A (en) * | 2019-08-16 | 2019-12-13 | 河海大学 | Hyperspectral remote sensing image classification method based on deep learning and morphology |
-
2020
- 2020-06-19 CN CN202010566343.XA patent/CN111898423B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180260657A1 (en) * | 2016-11-01 | 2018-09-13 | Shenzhen University | Method and system for extracting and classifying features of hyperspectral remote sensing image |
| CN108985360A (en) * | 2018-06-29 | 2018-12-11 | 西安电子科技大学 | Hyperspectral classification method based on expanding morphology and Active Learning |
| CN109583479A (en) * | 2018-11-07 | 2019-04-05 | 北京市遥感信息研究所 | One kind being based on morphologic high spectrum image feature extracting method |
| CN110569884A (en) * | 2019-08-16 | 2019-12-13 | 河海大学 | Hyperspectral remote sensing image classification method based on deep learning and morphology |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112396125A (en) * | 2020-12-01 | 2021-02-23 | 中国第一汽车股份有限公司 | Classification method, device, equipment and storage medium for positioning test scenes |
| CN112396125B (en) * | 2020-12-01 | 2022-11-18 | 中国第一汽车股份有限公司 | Classification method, device, equipment and storage medium for positioning test scenes |
| CN112906528A (en) * | 2021-02-05 | 2021-06-04 | 北京观微科技有限公司 | Urban building material classification method fusing multi-source satellite remote sensing data |
| CN112906528B (en) * | 2021-02-05 | 2021-10-26 | 北京观微科技有限公司 | Urban building material classification method fusing multi-source satellite remote sensing data |
| CN113256682A (en) * | 2021-06-04 | 2021-08-13 | 北京理工大学 | Infrared small target detection tracking method based on morphology and pipeline filtering |
| CN115984635A (en) * | 2023-03-21 | 2023-04-18 | 自然资源部第一海洋研究所 | Multi-source remote sensing data classification model training method, classification method and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111898423B (en) | 2022-12-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111898423B (en) | Morphology-based multisource remote sensing image ground object fine classification method | |
| Tu et al. | KNN-based representation of superpixels for hyperspectral image classification | |
| Matsuki et al. | Hyperspectral tree species classification of Japanese complex mixed forest with the aid of LiDAR data | |
| Gómez-Chova et al. | Multimodal classification of remote sensing images: A review and future directions | |
| Wang et al. | Nearest neighbor-based contrastive learning for hyperspectral and LiDAR data classification | |
| Jahan et al. | Fusion of hyperspectral and LiDAR data using discriminant correlation analysis for land cover classification | |
| Xi et al. | Evaluation of dimensionality reduction methods for individual tree crown delineation using instance segmentation network and UAV multispectral imagery in urban forest | |
| Pham et al. | Recent developments from attribute profiles for remote sensing image classification | |
| Hou et al. | Spectral–spatial classification of hyperspectral data using 3-D morphological profile | |
| Bulatov et al. | Automatic tree-crown detection in challenging scenarios | |
| Li et al. | Adaptive material matching for hyperspectral imagery destriping | |
| Zhu et al. | HCNNet: A hybrid convolutional neural network for spatiotemporal image fusion | |
| Zhao et al. | Research on camouflage recognition in simulated operational environment based on hyperspectral imaging technology | |
| Seydi et al. | A quadratic morphological deep neural network fusing radar and optical data for the mapping of burned areas | |
| CN112861802A (en) | Full-automatic crop classification method based on space-time deep learning fusion technology | |
| CN104268535A (en) | Method for extracting features of two-dimensional image | |
| Ouerghemmi et al. | Urban vegetation mapping by airborne hyperspetral imagery; feasibility and limitations | |
| Zhang | Pan-sharpening for improved information extraction | |
| Tawade et al. | Remote sensing image fusion using machine learning and deep learning: a systematic review | |
| Zhang et al. | Advanced processing techniques for remotely sensed imagery | |
| Li et al. | Modified extinction profiles for hyperspectral image classification | |
| Patel et al. | Survey on image fusion: Hand designed to deep learning algorithms | |
| Sumarsono et al. | Hyperspectral image segmentation with low-rank representation and spectral clustering | |
| Widyaningrum et al. | Skeleton-based automatic road network extraction from an orthophoto colored point cloud | |
| Shibi et al. | A methodology for unsupervised feature learning in hyperspectral imagery using deep belief network |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |