CN109829511B - Texture classification-based method for detecting cloud layer area in downward-looking infrared image - Google Patents
Texture classification-based method for detecting cloud layer area in downward-looking infrared image Download PDFInfo
- Publication number
- CN109829511B CN109829511B CN201910148702.7A CN201910148702A CN109829511B CN 109829511 B CN109829511 B CN 109829511B CN 201910148702 A CN201910148702 A CN 201910148702A CN 109829511 B CN109829511 B CN 109829511B
- Authority
- CN
- China
- Prior art keywords
- image
- sub
- cloud layer
- pixel
- gray level
- 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.)
- Active
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to a texture classification-based method for detecting a cloud layer area in an downward-looking infrared image. The method takes the detection problem of the cloud layer region in the downward-looking infrared image as a texture classification problem between the cloud layer region and the underlying surface ground object region, firstly, a median filtering and contrast enhancement algorithm is adopted to preprocess an input image, eliminate noise interference and improve image contrast; secondly, extracting image characteristics, in order to adapt to the complexity of cloud layer shapes and improve the accuracy of judgment, the invention divides an input image into self-adaptive sub-images, processes each sub-image independently, and extracts the gray level co-occurrence matrix related characteristics of the image and a uniform local binary pattern with unchanged rotation; and finally, judging the type of the image subgraph by adopting a pre-trained support vector machine according to the extracted image characteristics, and detecting the cloud layer area in the image.
Description
Technical Field
The invention mainly relates to the field of computer vision and machine learning, in particular to a method for realizing cloud layer area detection by utilizing texture characteristic difference between a cloud layer area and an underlying surface ground object area in an underlying infrared image.
Technical Field
In the infrared terminal guidance system, when the infrared terminal guidance system is in a working state, if a cloud layer exists between a bomb body and a target, a ground target area is shielded to form a blind area, and matching errors are easily caused, so that a terminal guidance task fails. Therefore, in the terminal guidance process, it is necessary to introduce cloud layer detection, and a terminal guidance workflow is formulated according to the distribution of cloud layers in an image, so as to minimize adverse effects of the cloud layers on the terminal guidance process, for example, if most regions in the image are covered by the cloud layers and the remaining effective regions are not sufficient to support stable matching positioning, matching processing is not performed on current data, guidance is performed according to other modes, if only a few regions in the image are covered by the cloud layers, the cloud layer covered regions can be shielded, and matching processing is performed by using regions not polluted by the cloud layers. How to accurately and efficiently detect the cloud layer area in the downward-looking infrared image is an important problem to be solved in an infrared terminal guidance system, and has important significance for improving the performance of the guidance system.
The invention aims at detecting the cloud layer area in the downward-looking infrared image acquired in the infrared terminal guidance, and has fewer related research results. The cloud detection technology which is widely researched and mature at present is focused on the fields of remote sensing and weather. The remote sensing field mainly focuses on detecting and removing a cloud layer area and a shadow area corresponding to the cloud layer area in a remote sensing image, eliminates the shielding of the cloud layer and the shadow on ground objects, and provides a foundation for subsequent analysis and application; cloud detection in the meteorological field is mainly used for distinguishing meteorological relevant characteristics such as cloud amount, cloud layer type and the like, and provides basis for weather analysis, weather forecast and the like, image data is mainly acquired by departments such as meteorological stations and the like, and certainly, the image data acquired by remote sensing satellites is also applied to the meteorological field. The detection of the cloud layer region in the downward-looking infrared image, which is referred to herein, can be essentially regarded as a texture classification problem, i.e., a sub-image region in the image is distinguished as an underlying ground scene or a cloud layer according to texture attributes, which is quite similar to the cloud layer detection idea based on image characteristics in the fields of remote sensing and weather. Cloud layer detection in the field of remote sensing essentially means that a cloud layer part in an image is determined according to the texture characteristic difference between a cloud layer area and a ground object area; the task of the cloud layer detection processing module in the meteorological field comprises the judgment of cloud amount and cloud type, firstly, a cloud layer area is positioned in an image, namely, the cloud layer area and a non-cloud layer area in the image are distinguished, and then the cloud layer type is judged according to the texture characteristics of the cloud layer area.
Disclosure of Invention
The patent discloses a cloud layer area detection method based on textural feature classification, aiming at the problem of cloud layer area detection in an downward-looking infrared image, which comprises the following steps: and finishing the detection of the cloud layer area by utilizing the texture difference between the cloud layer area and the underlying surface ground object area in the downward-looking infrared image. In order to adapt to the complexity of the cloud layer shape and improve the accuracy of judgment, the invention performs adaptive sub-image division on an input image and independently processes each sub-image block; describing texture features by using gray level co-occurrence matrix correlation features and rotation-invariant uniform local binary pattern features; and judging whether the sub-image blocks are in the cloud layer region or not based on a pre-trained support vector machine model according to the extracted textural feature description of each sub-image block, thereby completing the detection of the cloud layer region in the image.
1. Implementation process of cloud layer region detection method in downward-looking infrared image based on texture classification
The implementation process of the invention is shown in the attached figure 2, and concretely comprises the following steps:
(1) image preprocessing: the method comprises the steps of preprocessing an input image by adopting a median filtering and contrast enhancement algorithm, eliminating noise interference, improving image contrast, and carrying out sub-image division on the input image to adapt to the complexity of cloud layer shapes and improve the accuracy of judgment, wherein sub-image blocks are used as elements for subsequent processing;
(2) extracting texture features: processing each sub-image block independently, extracting the relevant characteristics of the gray level co-occurrence matrix corresponding to the sub-image block and a uniform local binary pattern with unchanged rotation as the description of the texture characteristics of the sub-image block;
(3) and (4) classification and judgment: and aiming at each sub-image block, judging each sub-image block type by adopting a pre-trained support vector machine model according to the extracted textural feature description, and further detecting the cloud layer area in the image.
2. The invention discloses a texture classification-based method for detecting cloud layer areas in an downward-looking infrared image
(1) Image pre-processing
The image preprocessing comprises filtering processing, image dynamic range adjustment and sub-image block self-adaptive division, and introduction is performed one by one as follows:
in comprehensive consideration of the aspects of noise characteristics, filtering performance, efficiency and the like in the downward-looking infrared image in the application background, the method selects a median filtering algorithm to filter the downward-looking infrared image. The median filtering is a nonlinear filter, as shown in formula (1), when a certain pixel point in an image is median filtered, the pixels in the mask and subjected to filtering and the pixels in the neighborhood are sorted according to the pixel value, the median is found and assigned to the pixel point,
wherein I is an input image, f is a filtered image, SuvIs a mask area corresponding to the pixel point (u, v) to be filtered;
s and t are respectively the pixel points at SuvRelative abscissa and ordinate of (1);
in order to eliminate the adverse interference caused by the dynamic range difference, the invention adjusts the dynamic range of the input image. The dynamic range of the image is adjusted by adopting a gray level normalization operation, and the calculation formula is as follows:
wherein f is the median filtered image,
and σ is the mean and mean square error of the gray scale of the image f, and equation (2) maps the input image to
The mean and variance of the gray scale values are 128 and
k is a constant and is usually selected in the value range of [3,5 ]]；
The shape of the cloud layer is not regular and varies variously, so that the cloud layer is more accurateThe invention can detect the cloud layer area in the image, and the invention can process self-adapting division operation to the sub-image block, and uses the sub-image as the basic processing unit to judge whether each sub-image belongs to the cloud layer area. In image partitioning, the size of a sub-image needs to be determined according to the resolution of the image, the accuracy of a result cannot be guaranteed if the size of the sub-image is too large, and a stable feature extraction result cannot be obtained if the size of the sub-image is too small and the number of pixels is small. For an input image of size M x N, the sub-image size Mb*NbThe following relationship exists with the classification accuracy d% that needs to be achieved:
in the formula
Is a round-up operation. For an input image with a size of 320 × 256, if the classification accuracy is 95% and the sub-images have the same length and width, the total number of sub-images is 20 and the partitioning result is shown in fig. 2 if the sub-image size is 64 × 64.
(2) Texture feature extraction
The invention adopts gray level co-occurrence matrix correlation characteristics and rotation invariant uniform local binary pattern characteristics to extract texture characteristics:
the gray level co-occurrence matrix is defined based on the joint probability density of the gray level values of the pixels at two positions in the image, the gray level co-occurrence matrix describes the distribution characteristic of the gray level and the position distribution characteristic between the pixels with similar gray level values, and belongs to the second-order statistical characteristic for describing the gray level change in the image. Let the gray level co-occurrence matrix be P ═ P (i, j | d, θ)]L×LIn the direction θ, the distance between the pixel pairs is d, and the probability p (i, j) of the occurrence of the pixel pairs with gray levels i and j is described, where i, j is 0,1,2, L, and L-1 are the gray levels corresponding to the pixels, and L is the gray level number of the image. The direction θ is typically taken to be four directions, 0 °, 45 °, 90 ° and 135 °, as shown in fig. 3. Each direction corresponds to a matrix, and the four directions are totally fourFor the input image I, calculating the gray level co-occurrence matrix elements in all directions as shown in formula (5);
wherein, (k, l), (m, n) are respectively the abscissa and ordinate of the corresponding position of the two pixels to be compared.
During texture feature analysis, the gray level co-occurrence matrix obtained through calculation is not directly used, but some statistic quantity is obtained through calculation based on the gray level co-occurrence matrix and is used as feature quantity of the texture, entropy, energy, homogeneity, contrast, clustering tendency, clustering form, correlation, information measurement correlation 1, information measurement correlation 2 and maximum probability are used, and 10 statistic quantities are calculated;
the basic principle of the local binary pattern is to compare the threshold value of the pixel in the image with the pixel in the neighborhood to obtain a binary coding sequence, and to take the decimal number corresponding to the binary coding sequence as the local binary pattern value LBP of the pointP,RAs shown in the formula (6),
wherein g iscIs the central pixel (u)c,vc) Gray value of gpIs the gray value of its neighborhood pixels, p is the pixel in the neighborhood, s (-) is a sign function,
fig. 4 shows an example of local binary pattern extraction in the 3 × 3 domain. In order to adapt to rotation, the binary coding sequence obtained by the circular neighborhood is connected end to form a cyclic coding sequence, a series of corresponding local binary pattern values are obtained by rotating the circular neighborhood, and the minimum value is used as the local binary pattern characteristic of the neighborhood
See the formula (8),
ri denotes rotation invariant, ROR (LBP)P,RL) denotes the P-bit number LBPP,RRotating l bits clockwise, P represents the number of sampling points, and R represents the radius of the circular neighborhood.
The transition U between 0 and 1 in the binary representation of the local binary pattern is defined as follows
g0Gray value g representing pixel No. 0 in circular neighborhood of center pixelp-1Gray value g representing the (p-1) th pixel in the circular neighborhood of the center pixelP-1And expressing the gray value of the (P-1) th pixel point in the circular neighborhood of the central pixel point.
When the jump number between 0 and 1 contained in the cyclic binary code corresponding to the local binary pattern is not more than 2, the binary code corresponding to the local binary pattern is marked as an equivalent pattern class, and the patterns except the equivalent pattern class are classified as another class, which is called a mixed mode class. Improvement of two aspects of combining rotation invariance and 'equivalent mode', rotation invariance uniform local binary mode
The definition is shown in the formula (10),
(3) classification and discrimination
Based on the obtained texture feature description of the sub-image block, the method adopts a support vector machine as a classifier to judge the type of the sub-image. The linear support vector machine is defined as
y＝g(wTx+b) (11)
Where y ∈ { -1,1} outputs the class label, x is the input feature vector, w is the weight vector, b is the offset, and the threshold function g (·) outputs either-1 (· < 0) or 1(· > 0). The linear support vector machine determines the best classification plane by maximizing the classification interval, as shown in fig. 5. Based on the labeled training sample, parameters w and b are realized by adopting a Lagrange multiplier method and a sequential minimum optimization algorithm;
in order to process the nonlinear classification problem, a kernel function method is introduced to popularize a linear support vector machine to the nonlinear condition, and a radial basis kernel function is adopted in the invention
Where σ is the radial basis function standard deviation parameter, x is the input feature vector, xzA relaxation variable xi is introduced on the basis of a basic support vector machine for a characteristic vector corresponding to the z-th support vector and adapting to the interference of noisezNot less than 0, and simultaneously optimizing and solving w, b and xiz。
Compared with the prior art, the invention has the advantages that:
(1) according to the method, the cloud layer region detection is realized by utilizing the difference of the texture of the cloud layer region and the texture of the ground object region of the underlying surface in the downward-looking infrared image based on the idea of feature classification, the algorithm steps are simple and clear, the calculation complexity is low, and the method is easy to apply;
(2) in order to adapt to the complexity of the cloud layer shape and improve the accuracy and efficiency of judgment, the invention performs adaptive sub-map division on an input image and takes a sub-map block as a primitive for subsequent processing.
Drawings
FIG. 1 is a flow chart of a method for detecting cloud layer regions in an undersight infrared image based on texture classification;
FIG. 2 is a schematic diagram of image sub-graph partitioning;
FIG. 3 is a schematic view of gray level co-occurrence matrix angles;
figure 43 x 3 neighborhood local binary pattern extraction example;
FIG. 5 is a linear support vector machine classification schematic.
Detailed Description
The embodiments of the present invention are described in further detail below:
(1) image pre-processing
Preprocessing an input image, and sequentially performing median filtering, dynamic range adjustment and blocking processing;
the median filtering is a nonlinear filter, as shown in formula (1), when a certain pixel point in an image is median filtered, the pixels in the mask and subjected to filtering and the pixels in the neighborhood are sorted according to the pixel value, the median is found and assigned to the pixel point,
wherein I is an input image, f is a filtered image, SuvA mask region corresponding to the pixel (u, v) to be filtered; s and t are respectively the pixel points at SuvRelative abscissa and ordinate of (1);
in order to eliminate the adverse interference caused by the dynamic range difference, the invention adjusts the dynamic range of the input image. The dynamic range of the image is adjusted by adopting a gray level normalization operation, and the calculation formula is as follows:
wherein f is the median filtered image,
and σ is the mean and mean square error of the gray scale of the image f, and equation (2) maps the input image to
The mean and variance of the gray scale values are 128 and
k can range from [3,5 ]]；
In the blocking, the size of a sub-image needs to be determined according to the resolution of the image, the accuracy of the result cannot be ensured when the size of the sub-image is too large, and a stable feature extraction result cannot be obtained when the number of pixels is less when the size of the sub-image is too small. For an input image of size M x N, the sub-image size Mb*NbThe following relationship exists with the classification accuracy d% that needs to be achieved:
in the formula
Is a round-up operation. For an input image with a size of 320 × 256, if the classification accuracy is 95% and the sub-images have the same length and width, the total number of sub-images is 20 and the partitioning result is shown in fig. 2 if the sub-image size is 64 × 64.
(2) Texture feature extraction
Extracting relevant features of a gray level co-occurrence matrix and rotation invariant uniform binary pattern features for each image sub-block for perhaps classification judgment;
the parameters involved in the gray level co-occurrence matrix calculation include a direction θ and a step length d, and the direction θ generally takes four directions of 0 °, 45 °, 90 ° and 135 °, as shown in fig. 3. Each direction corresponds to one matrix, four matrixes are obtained in all four directions, and for an input image I, the calculation of gray level co-occurrence matrix elements in all directions is shown in a formula (5);
calculating the relevant statistics of the gray level co-occurrence matrix: entropy, energy, homogeneity, contrast, clustering tendency, clustering morphology, correlation, information measure correlation 1, information measure correlation 2, and maximum probability;
the rotation-invariant uniform binary pattern is defined as shown in equation (10),
(3) classification and discrimination
And (3) judging the category of each image subblock based on the pre-trained support vector machine model according to the texture feature description extracted in the step (2), and further completing the detection of the cloud layer region in the image.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may be made by those skilled in the art without departing from the principle of the invention.
Claims (3)
1. A method for detecting a cloud layer region in an downward-looking infrared image based on texture classification utilizes texture difference between the cloud layer region in the downward-looking infrared image and an underlying surface ground object region to complete cloud layer region detection, and is characterized in that:
performing adaptive sub-map division on an input image, and processing each sub-map block independently; describing texture features by using gray level co-occurrence matrix correlation features and rotation-invariant uniform local binary pattern features; according to the extracted textural feature description of each sub-image block, based on a pre-trained support vector machine model, judging whether the sub-image block is a cloud layer region or not, and further completing the detection of the cloud layer region in the image, wherein the steps comprise:
(1) image preprocessing: comprises filtering processing, image dynamic range adjustment and sub-image block self-adaptive division,
filtering the downward-looking infrared image by a median filtering algorithm, sorting pixels in a mask and subjected to filtering processing and pixels in the neighborhood thereof according to pixel values when performing median filtering on a certain pixel point in an image, finding out a median value, and assigning the median value to the pixel point,
wherein I is an input image, f is a filtered image, SuvIs a mask area corresponding to the pixel point (u, v) to be filtered; s and t are respectively the pixel points at SuvRelative abscissa and ordinate of (1);
in order to eliminate the interference caused by the difference of the dynamic ranges, the dynamic range of the input image is adjusted, the dynamic range of the image is adjusted by adopting a gray level normalization operation, and the calculation formula is as follows:
wherein f is the median filtered image,
and σ is the mean and mean square error of the gray scale of the image f, and equation (2) maps the input image to
The mean and variance of the gray scale values are 128 and
k is a constant and is usually selected in the value range of [3,5 ]]；
Performing adaptive partition operation on the sub-image blocks, and judging by using the sub-image as a basic processing unitWhether each subgraph belongs to the cloud layer area or not, in image blocking, if the subgraph size is too large, the accuracy of a result cannot be ensured, if the subgraph size is too small, the number of pixels is less, a stable feature extraction result cannot be obtained, and for an input image with the size of M x N, the subgraph size Mb*NbThe following relationship exists with the classification accuracy d% that needs to be achieved:
in the formula
The operation of rounding up is carried out;
(2) extracting texture features: processing each sub-image block independently, extracting the relevant characteristics of the gray level co-occurrence matrix corresponding to the sub-image block and a uniform local binary pattern with unchanged rotation as the description of the texture characteristics of the sub-image block;
(3) and (4) classification and judgment: and aiming at each sub-image block, judging each sub-image block type by adopting a pre-trained support vector machine model according to the extracted textural feature description, and further detecting the cloud layer area in the image.
2. The method for detecting the cloud layer region in the downward-looking infrared image based on the texture classification as claimed in claim 1, wherein the texture feature extraction is performed by using a gray level co-occurrence matrix correlation feature and a rotation-invariant uniform local binary pattern feature:
defining a gray level co-occurrence matrix based on joint probability densities of gray levels of pixels at two locations in the image, noting that the gray level co-occurrence matrix is P ═ P (i, j | d, theta)]L×LIt describes the probability p (i, j) of the occurrence of the pixel pair whose gray levels are i and j respectively at a distance step d in the direction θ, where i, j is 0,1,2, L-1 is the gray value corresponding to the pixel, L is the gray level number of the image, and the solving process can find that the gray level co-occurrence matrix is a symmetric matrix, the direction θ generally takes four directions of 0 °, 45 °, 90 ° and 135 °, and each direction isObtaining four matrixes in four directions corresponding to one matrix, and calculating gray level co-occurrence matrix elements in all directions of the input image I as shown in a formula (5);
wherein, (k, l), (m, n) are respectively the horizontal and vertical coordinates of the corresponding positions of the two pixels to be compared;
the basic principle of the local binary pattern is to compare the threshold value of the pixel in the image with the pixel in the neighborhood to obtain a binary coding sequence, and to take the decimal number corresponding to the binary coding sequence as the local binary pattern value LBP of the pointP,RAs shown in the formula (6),
wherein g iscIs the central pixel (u)c,vc) Gray value of gpIs the gray value of its neighborhood pixels, p is the pixel in the neighborhood, s (-) is a sign function,
in order to adapt to rotation, the binary coding sequence obtained by the circular neighborhood is connected end to form a cyclic coding sequence, a series of corresponding local binary pattern values are obtained by rotating the circular neighborhood, and the minimum value is used as the local binary pattern characteristic of the neighborhood
See the formula (8),
ri denotes rotation invariant, ROR (LBP)P,RL) denotes the P-bit number LBPP,RRotating l bits clockwise, P represents the number of sampling points, and R represents the radius of a circular neighborhood;
the transition U between 0 and 1 in the binary representation of the local binary pattern is defined as follows
g0Gray value g representing pixel No. 0 in circular neighborhood of center pixelp-1Gray value g representing the (p-1) th pixel in the circular neighborhood of the center pixelP-1Expressing the gray value of the No. (P-1) pixel point in the circular neighborhood of the central pixel point;
when the jump number between 0 and 1 contained in the cyclic binary code corresponding to the local binary pattern is not more than 2, the binary code corresponding to the local binary pattern is marked as an equivalent pattern class, the patterns except the equivalent pattern class are classified into another class called a mixed mode class, the improvement of two aspects of rotation invariance and equivalent pattern is synthesized, and the rotation invariance is uniform
The definition is shown in the formula (10),
3. the method of claim 1, wherein the classification decision uses a support vector machine as a classifier to decide the sub-image type, and the linear support vector machine is defined as
y＝g(wTx+b) (11)
In the formula, y belongs to { -1,1} output class labels, x is an input feature vector, w is a weight vector, b is an offset, a threshold function g (·) outputs-1 (· < 0) or 1(· > 0), a linear support vector machine determines an optimal classification surface through a maximized classification interval, and parameters w and b are realized by adopting a Lagrange multiplier method and a sequential minimum optimization algorithm based on labeled training samples;
in order to process the nonlinear classification problem, a kernel function method is introduced to popularize a linear support vector machine to the nonlinear condition, and a radial basis kernel function is adopted
Where σ is the radial basis function standard deviation parameter, x is the input feature vector, xzA relaxation variable xi is introduced on the basis of a basic support vector machine for a characteristic vector corresponding to the z-th support vector and adapting to the interference of noisezNot less than 0, and simultaneously optimizing and solving w, b and xiz。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910148702.7A CN109829511B (en) | 2019-02-28 | 2019-02-28 | Texture classification-based method for detecting cloud layer area in downward-looking infrared image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910148702.7A CN109829511B (en) | 2019-02-28 | 2019-02-28 | Texture classification-based method for detecting cloud layer area in downward-looking infrared image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109829511A CN109829511A (en) | 2019-05-31 |
| CN109829511B true CN109829511B (en) | 2022-04-08 |
Family
ID=66864721
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910148702.7A Active CN109829511B (en) | 2019-02-28 | 2019-02-28 | Texture classification-based method for detecting cloud layer area in downward-looking infrared image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109829511B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111753689A (en) * | 2020-06-15 | 2020-10-09 | 珠海格力电器股份有限公司 | Image texture feature extraction method and image identification method |
| CN114724188B (en) * | 2022-05-23 | 2022-08-16 | 北京圣点云信息技术有限公司 | Vein identification method and device based on gray level co-occurrence matrix |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103034858A (en) * | 2012-11-30 | 2013-04-10 | 宁波大学 | Secondary clustering segmentation method for satellite cloud picture |
| CN107886490A (en) * | 2018-01-14 | 2018-04-06 | 中国人民解放军国防科技大学 | Offshore sea area azimuth ambiguity removing method based on double-temporal SAR image |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8532360B2 (en) * | 2010-04-20 | 2013-09-10 | Atheropoint Llc | Imaging based symptomatic classification using a combination of trace transform, fuzzy technique and multitude of features |
-
2019
- 2019-02-28 CN CN201910148702.7A patent/CN109829511B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103034858A (en) * | 2012-11-30 | 2013-04-10 | 宁波大学 | Secondary clustering segmentation method for satellite cloud picture |
| CN107886490A (en) * | 2018-01-14 | 2018-04-06 | 中国人民解放军国防科技大学 | Offshore sea area azimuth ambiguity removing method based on double-temporal SAR image |
Non-Patent Citations (3)
| Title |
|---|
| LBP和GLCM融合的织物组织结构分类;景军锋等;《电子测量与仪器学报》;20150915(第09期);全文 * |
| Study of Enteromorpha Idendification Based on Machine Learning Technology;Xueying MA et al.;《2018 OCEANS-MTS/IEEE Kobe Techno-Oceans(OT0)》;20181206;全文 * |
| 基于可见光――红外图像信息融合的云状识别方法;张弛等;《气象与环境学报》;20180215(第01期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109829511A (en) | 2019-05-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108961235B (en) | Defective insulator identification method based on YOLOv3 network and particle filter algorithm | |
| CN108573276B (en) | Change detection method based on high-resolution remote sensing image | |
| CN113160192B (en) | Visual sense-based snow pressing vehicle appearance defect detection method and device under complex background | |
| CN109271991B (en) | License plate detection method based on deep learning | |
| CN103049763B (en) | Context-constraint-based target identification method | |
| CN111753828B (en) | Natural scene horizontal character detection method based on deep convolutional neural network | |
| CN110264448B (en) | Insulator fault detection method based on machine vision | |
| CN104182763B (en) | A kind of floristics identifying system based on flower feature | |
| CN107633226B (en) | Human body motion tracking feature processing method | |
| CN107038416B (en) | Pedestrian detection method based on binary image improved HOG characteristics | |
| CN109801305B (en) | SAR image change detection method based on deep capsule network | |
| CN111539330B (en) | Transformer substation digital display instrument identification method based on double-SVM multi-classifier | |
| CN112307919B (en) | Improved YOLOv 3-based digital information area identification method in document image | |
| CN110766016A (en) | Code spraying character recognition method based on probabilistic neural network | |
| CN109829511B (en) | Texture classification-based method for detecting cloud layer area in downward-looking infrared image | |
| CN111091071B (en) | Underground target detection method and system based on ground penetrating radar hyperbolic wave fitting | |
| CN115272225A (en) | Strip steel surface defect detection method and system based on countermeasure learning network | |
| CN111046838A (en) | Method and device for identifying wetland remote sensing information | |
| CN114494704A (en) | Method and system for extracting framework from binary image in anti-noise manner | |
| Shire et al. | A review paper on: agricultural plant leaf disease detection using image processing | |
| CN112101283A (en) | Intelligent identification method and system for traffic signs | |
| CN105512682B (en) | A kind of security level identification recognition methods based on Krawtchouk square and KNN-SMO classifier | |
| Günen et al. | A novel edge detection approach based on backtracking search optimization algorithm (BSA) clustering | |
| CN109785318B (en) | Remote sensing image change detection method based on facial line primitive association constraint | |
| CN115063679B (en) | Pavement quality assessment method based on deep learning |
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 |