CN109034274B - Method, device and equipment for improving hyperspectral image classification precision and storage medium - Google Patents
Method, device and equipment for improving hyperspectral image classification precision and storage medium Download PDFInfo
- Publication number
- CN109034274B CN109034274B CN201811001464.9A CN201811001464A CN109034274B CN 109034274 B CN109034274 B CN 109034274B CN 201811001464 A CN201811001464 A CN 201811001464A CN 109034274 B CN109034274 B CN 109034274B
- Authority
- CN
- China
- Prior art keywords
- tensor
- order
- hyperspectral image
- local area
- preset
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/758—Involving statistics of pixels or of feature values, e.g. histogram matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/58—Extraction of image or video features relating to hyperspectral data
Abstract
The invention discloses a method, a device, equipment and a storage medium for improving the classification precision of hyperspectral images, wherein the method comprises the following steps: establishing a third-order tensor of a preset local area of the hyperspectral image, and carrying out tensor decomposition on the third-order tensor to obtain a third-order nuclear tensor corresponding to the preset local area; traversing each front slice in the third-order kernel tensor, and establishing a feature statistical histogram of each front slice by using a preset coding function; and connecting the characteristic statistical histograms of the front slices to obtain a statistical histogram of a preset local area, and classifying the hyperspectral images by using the statistical histogram of the preset local area. Compared with the prior art, the statistical histogram obtained by the embodiment of the invention not only can reflect the information on the hyperspectral image space domain, but also can reflect the information of the spectral domain, so that the classification precision can be effectively improved in the hyperspectral image classification.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method, a device, equipment and a storage medium for improving the classification precision of hyperspectral images.
Background
The hyperspectral image not only contains space structure information reflecting morphological characteristics of the ground objects, but also contains spectral structure information reflecting category attributes and state attributes of the ground objects, and is called as 'map integration'. The 'map-in-one' technology greatly improves the human sensory ability and is widely applied in the fields of agricultural science, materials, earth science, food industry, biomedical application, military and the like.
An LBP (Local Binary Pattern) algorithm is a simple and effective image texture description algorithm with low computational complexity, can deeply mine spatial information contained in an image, and is widely used for tasks related to classification, such as texture analysis, target detection, target identification and the like. However, the conventional LBP algorithm applied to the hyperspectral image focuses on the description of the spatial information, and ignores the spectral information, so that the classification accuracy of the hyperspectral image is low.
Disclosure of Invention
The invention mainly aims to provide a method, a device, equipment and a storage medium for improving the classification precision of hyperspectral images, and aims to solve the technical problem of low classification precision when an LBP algorithm is used for classifying hyperspectral images in the prior art.
In order to achieve the above object, a first aspect of the present invention provides a method for improving the classification accuracy of hyperspectral images, the method comprising:
establishing a third-order tensor of a preset local area of a hyperspectral image, and carrying out tensor decomposition on the third-order tensor to obtain a third-order nuclear tensor corresponding to the preset local area;
traversing each front slice in the third-order kernel tensor, and establishing a feature statistical histogram of each front slice by using a preset coding function;
and connecting the characteristic statistical histograms of the front slices to obtain a statistical histogram of the preset local area, and classifying the hyperspectral images by using the statistical histogram of the preset local area.
Optionally, the step of traversing each front slice in the third-order kernel tensor and establishing a feature statistical histogram of each front slice by using a preset coding function includes:
determining domain feature information of tensor elements in the traversed front slice;
quantizing the domain feature information to obtain a spatial-spectral domain joint coding value of the tensor element;
and determining a characteristic statistical histogram of the traversed front slice by using the spatial-spectral domain joint coding value.
Optionally, the step of quantizing the domain feature information to obtain the spatio-spectral domain joint encoded value of the tensor element includes:
adding a preset binomial coefficient to the domain feature information;
and converting the domain characteristic information carrying the binomial coefficient into a decimal number to obtain a space-spectral domain joint coding value of the tensor element.
Optionally, the step of joining the feature statistical histograms of the front slices includes:
and connecting the feature statistical histograms of the front slices according to the position sequence of the front slices in the third-order nuclear tensor.
In order to achieve the above object, a second aspect of the present invention provides an apparatus for improving the classification accuracy of hyperspectral images, the apparatus comprising:
the system comprises an establishing module, a calculating module and a calculating module, wherein the establishing module is used for establishing a third-order tensor of a preset local area of a hyperspectral image, and carrying out tensor decomposition on the third-order tensor to obtain a third-order nuclear tensor corresponding to the preset local area;
the operation module is used for traversing each front slice in the third-order kernel tensor and establishing a feature statistical histogram of each front slice by using a preset coding function;
and the classification module is used for connecting the characteristic statistical histograms of the front slices to obtain a statistical histogram of the preset local area, and classifying the hyperspectral image by using the statistical histogram of the preset local area.
Optionally, the operation module includes:
the first determination module is used for determining domain feature information of tensor elements in the traversed front slice;
the quantization module is used for quantizing the domain characteristic information to obtain a spatial-spectral domain joint coding value of the tensor element;
and the second determining module is used for determining the feature statistical histogram of the traversed front slice by using the spatial-spectral domain joint coding value.
Optionally, the quantization module is specifically configured to:
and adding a preset binomial coefficient into the domain characteristic information, and converting the domain characteristic information carrying the binomial coefficient into decimal numbers to obtain a space-spectral domain joint coding value of the tensor element. Optionally, the classification module includes:
and the connection module is used for connecting the feature statistical histograms of the front slices according to the position sequence of the front slices in the third-order nuclear tensor.
In order to achieve the above object, a third aspect of the present invention provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor executes the computer program to implement the steps in the method for improving the classification accuracy of hyperspectral images according to the first aspect of the present invention.
To achieve the above object, a fourth aspect of the present invention provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the computer program implements the steps in the method for improving the classification accuracy of hyperspectral images according to the first aspect of the present invention.
Compared with the prior art, the method for improving the classification precision of the hyperspectral image comprises the steps of firstly establishing a third-order tensor of a preset local area of the hyperspectral image, carrying out tensor decomposition on the third-order tensor to obtain a third-order kernel tensor corresponding to the preset local area, then establishing a feature statistical histogram of each front section in the third-order kernel tensor, finally connecting the feature statistical histograms of the front sections to obtain a statistical histogram of the preset local area, and classifying the hyperspectral image by utilizing the statistical histogram. Because each front slice can reflect the joint distribution condition and the spatial distribution information of the spectral values of different wave bands in the hyperspectral image, the information on the spectral domain and the spatial domain of the hyperspectral image can be simultaneously represented through the characteristic statistical histogram of each front slice, namely the hyperspectral image is classified by using the statistical histogram, and the classification precision can be effectively improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic diagram of a third order tensor model of a hyperspectral image in an embodiment of the invention;
FIG. 2 is a schematic flow chart illustrating steps of a method for improving the classification accuracy of hyperspectral images in the embodiment of the invention;
FIG. 3 is an exploded view of a mode 3 fiber of the third order tensor T according to the embodiment of the present invention;
FIG. 4 is a schematic diagram of a virtual program module of an apparatus for improving the classification accuracy of hyperspectral images according to an embodiment of the invention;
FIG. 5 is a block diagram illustrating a detailed procedure of the operation module 402 according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of an electronic device provided in an embodiment of the present invention.
Detailed Description
In order to make the objects, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In an embodiment of the present invention, the hyperspectral image may be represented as I (x, y, λ), where (x, y) is a spatial domain coordinate and λ is a spectral domain coordinate. The hyperspectral image can be modeled into a third-order tensor, namely the third-order tensor T belongs to RX×Y×ΛRepresents I. Wherein, X and Y are the size of the hyperspectral image space domain, namely the value range of (X, Y); lambda is the spectral domain of the hyperspectral imageThe size (number of bands) of (a), that is, the value interval of (a); each pixel I (x, y, lambda) of the hyperspectral image corresponds to each element T of the third-order tensor Txyλ. Specifically, referring to fig. 1, fig. 1 is a schematic diagram of a third-order tensor model of a hyperspectral image.
The fibers and the slices of the tensor are one of basic elements forming the tensor, and the extraction of vectors or matrixes in the high-dimensional geometric body can be realized through the fibers and the slices of the tensor, so that the calculation complexity is effectively reduced. The fibers of the tensor are defined as indexes of any one order of change and indexes of other orders of fixation. The fibers of the third-order tensor T of the hyperspectral image are divided into rows (mode 1 fibers), columns (mode 2 fibers) and tube fibers (mode 3 fibers) which are respectively marked as T：yλ、Tx：λAnd Txy：. The pixel with spatial coordinates (x, y) can be mapped into an n-dimensional vector by the model
The representation is carried out, and the hyperspectral image is analyzed by combining the mode 3 fiber model, so that the utilization efficiency of the information of the spectral domain can be effectively improved.
Where a slice is a two-dimensional portion of a tensor, it is implemented by fixing the indices of any two orders, which is typically used for matrix extraction operations. The slices of the third order tensor T of the hyperspectral image are defined as horizontal slices, side slices and front slices which are respectively expressed as Tx::、T:y:And T::λAnd the front slice can effectively represent the joint distribution of the spectral values of different wave bands in the hyperspectral image. By performing matrix extraction on tensor elements in three different directions in the T, information of a spatial domain and information of a spectral domain can be effectively combined, and the efficiency of spatial-spectral-domain coding is improved.
In an embodiment of the present invention, on the basis of a third-order tensor model of a hyperspectral image, a spatial-spectral domain joint probability distribution is adopted to represent characteristics of the hyperspectral image, and a method for improving the classification precision of the hyperspectral image is provided, specifically referring to fig. 2, where fig. 2 is a schematic flow chart illustrating steps of the method for improving the classification precision of the hyperspectral image in the embodiment of the present invention, and in the embodiment of the present invention, the method for improving the classification precision of the hyperspectral image includes:
Firstly, due to the fact that the spectral imaging interval of the hyperspectral imager is small, image correlation of adjacent wave bands is high, serious redundant data exist among the wave bands, in order to reduce redundant information of the hyperspectral image, extract more refined useful information and improve efficiency of feature coding, the redundant information with large correlation needs to be removed, and data volume is compressed. As known from tensor mathematics, the modulo n multiplication of tensor refers to tensor
And an array
Multiplication, the product is still a tensor whose modulo size is:
I1×I2×…In-1×J×In-1×…×IN
when J is less than InThe modulo n multiplication of the tensor can be regarded as a dimension reduction process.
Tensor decomposition based on modulo n multiplication is an important method for performing compressed representation on an original tensor, and the tensor can be represented in a form of multiplying a nuclear tensor by a projection matrix on each order, wherein the nuclear tensor is the compressed representation of the original tensor. In order to fully consider the correlation among different wave bands and eliminate the spectrum redundancy, the embodiment of the invention utilizes tensor decomposition to decompose the hyperspectral image on the spectrum dimension on the basis of a three-order tensor model of the hyperspectral image, and can effectively reduce the spectrum redundancy information of the hyperspectral image. The hyperspectral image is reconstructed into a third-order nuclear tensor through tensor decomposition to be expressed, so that the space spectral domain code has the invariance of spectral resolution.
Assuming that T represents a tensor model of the hyperspectral image, tensor decomposition is performed on T as shown in the following formula:
T＝T′×3LT
wherein L isTIs the principal component on the mode 3 fiber of T, is the principal component3And expressing the modular 3 fiber multiplication of the tensor, wherein T ' belongs to RX multiplied by Y multiplied by Lambda ' is the nuclear tensor of T after tensor decomposition, X, Y and Lambda ' respectively express the sizes of the modular 1 fiber, the modular 2 fiber and the modular 3 fiber of T ', and Lambda ' is less than or equal to Lambda. In the above formula, T passes through the transpose matrix L of its projection matrixTThe corresponding kernel tensor T 'is obtained through compression, the hyperspectral image with the wave band number of lambda is mapped into the tensor with the modulo 3 fiber degree of lambda', the correlation and the redundant information among the wave bands can be effectively reduced, and the spectrum curve of each pixel faces to more refined ground feature type information. This can not only significantly improve the information content and efficiency of the spatial-spectral domain LBP, but also enhance the invariance of the characteristic against the spectral domain scale change, specifically refer to fig. 3, where fig. 3 is a schematic diagram of the modulo-3 fiber decomposition of the third-order tensor T in the embodiment of the present invention.
Specifically, D e R is used for setting the third-order tensor of the preset local area of the hyperspectral imageM×N×UExpressing that the third-order nuclear tensor after tensor decomposition is D' epsilon RM×N×U′And D 'is a subset of the third order core tensor T'. Wherein, M and N are the sizes of the mode 1 fiber and the mode 2 fiber of D ', and U ' is the size of the mode 3 fiber of D '.
Specifically, the step 202 includes:
step one, determining domain feature information of tensor elements in a traversed front slice;
quantizing the domain feature information to obtain a space-spectral domain joint coding value of the tensor element;
and thirdly, determining a characteristic statistical histogram of the traversed front section by using the spatial-spectral domain joint coding value.
Wherein, it is provided with
Is a positive slice of T' and is provided with
And is
Then, by using the LBP thought of the gray level image, the tensor elements in the front section of the third-order nuclear tensor are sliced
Is defined as
The joint probability distribution of tensor element values in the spatial spectral domain local neighborhood of (a) is as follows:
wherein S isk(k-0, 1, …, P-1) is cut on the front side
To move upwards
Tensor element values S of P sampling points sampled at equal intervals in circular neighborhood with radius R (R is more than 0) as circle centerkThe coordinate of (k-0, 1, …, P-1) is (x)c+Rsin(2πi/P),yc-Rcos(2πi/P),λc′),i∈{1,2,…,P}。Sk(k ═ P, P +1, …, P + J-1) is
Form 3 fiber
The tensor element values of the J sampling points have the coordinate of (x)c,yc,λc′+j),j∈{1,2,…,J}。
Wherein, in order to realize illumination invariance, tensor element value S of each sampling point is usedkAnd
subtracting to obtain:
wherein due to the difference
And
the values of (a) are independent of each other, so the above formula can be factored into:
in an actual hyperspectral image, the difference cannot be guaranteed
And
are independent of each other, so the above equation is an approximation of the joint probability distribution of the values of the tensor elements in the spatial domain. Since most of the information of the joint probability distribution is contained in the differential distribution, the above formula can also be expressed as:
illumination invariance of joint differential distribution is realized by only counting symbol information of pixels in a local neighborhood of a space spectral domain of a central point, rather than actual spectral values of the pixels, as follows:
wherein the content of the first and second substances,
further, the step of quantizing the domain feature information in the second step to obtain the spatial-spectral domain joint encoded value of the tensor element includes:
step a, adding a preset binomial coefficient into the domain characteristic information;
and b, converting the domain characteristic information carrying the binomial coefficient into a decimal number to obtain a space-spectral domain joint coding value of the tensor element.
Wherein, by for each symbol
Given a binomial coefficient of 2kConverting V into a complete and efficient description
Spatial-spectral domain joint coding value TSSLBP in a centered local neighborhoodP,J,RAs shown in the following equation:
from the above formula, TSSLBPP,J,RWill generate 2P+JA different output value, i.e. TSSLBPP,J,RThe number of coded binary patterns is 2P+J。
Due to the particularity of the hyperspectral image, the rotation operation only occurs on the pixel of the spatial domain, and the pixel of the spectral domain does not change, so that the spatial spectral domain coding value with rotation invariance is provided
Is defined as follows:
in the above formula, min (·) represents the minimum value, and ROR (x, r) represents the binary value obtained by circularly right-shifting a P-bit binary number x by r times. And finally, carrying out histogram statistics, and taking the obtained feature vector as the spatial spectral domain feature description of the local area of the hyperspectral image.
Wherein, for D', let the feature statistical histogram of the u-th slice be huWherein the value range of U is 1,2,3, …, U', and the statistical histogram h isuEach column of (1) is usedu(b) Expressed, it can be calculated from:
wherein the content of the first and second substances,
TSSLBP (m, n, u) represents the TSSLBP code value of the pixel with the spatial position (m, n) and modulo 3 fiber u. b represents any binary pattern of TSSLBP coding, and b is within [0,2 ]P+J]And b is an integer.
And 203, connecting the characteristic statistical histograms of the front slices to obtain a statistical histogram of the preset local area, and classifying the hyperspectral images by using the statistical histogram of the preset local area.
In the embodiment of the present invention, after the local feature statistical histogram of each front slice is obtained, the statistical histograms of U 'front slices need to be connected to be used as the feature statistical histogram h of the local region D'.
Wherein the step of joining the feature statistical histograms of the front slices in step 203 includes:
and connecting the feature statistical histograms of the front slices according to the position sequence of the front slices in the third-order nuclear tensor.
Specifically, the calculation method of h is shown by the following formula:
h＝[h1,h2,…hu,…,hU′]
further, in order to better understand the embodiment of the present invention, in the embodiment of the present invention, a pseudo code for implementing the method for improving the classification accuracy of the hyperspectral image according to the present invention is provided:
inputting: a hyperspectral image I;
establishing a third-order tensor T epsilon R of a hyperspectral image IX×Y×Λ；
Carrying out tensor decomposition on the third-order tensor T to obtain a third-order core tensor T' belonging to the third-order tensor TX×Y×Λ′；
for the u th frontal slice; in local area D' is belonged to RM×N×U′；
The position of the for space is an image element of (m, n); in the u-th frontal slice;
calculating TSSLBP codes by using a characteristic coding function;
End for；
establishing a feature statistical histogram h of the u-th front sliceu；
End for；
for the u th frontal slice; in local area D' is belonged to RM×N×U′；
h＝[h,hu](ii) a V/joining the feature statistical histogram of each front slice;
End for；
and (3) outputting: local region D' is belonged to RM×N×U′Is calculated.
Compared with the prior art, the method for improving the classification precision of the hyperspectral image comprises the steps of firstly establishing a third-order tensor of a preset local area of the hyperspectral image, carrying out tensor decomposition on the third-order tensor to obtain a third-order nuclear tensor corresponding to the preset local area, then establishing a feature statistical histogram of each front section in the third-order nuclear tensor, finally connecting the feature statistical histograms of the front sections to obtain a statistical histogram of the preset local area, and classifying the hyperspectral image by using the statistical histogram. The front slices can reflect the joint distribution condition and the spatial distribution information of the spectral values of different wave bands in the hyperspectral image, so that the information on the spectral domain and the spatial domain of the hyperspectral image can be simultaneously represented through the characteristic statistical histogram of each front slice, and the hyperspectral image is classified by using the statistical histogram, so that the classification precision can be effectively improved.
Further, an embodiment of the present invention further provides a device for improving the precision of hyperspectral image classification, referring to fig. 4, fig. 4 is a schematic diagram of a virtual program module of the device for improving the precision of hyperspectral image classification in the embodiment of the present invention, where the device includes:
the establishing module 401 is configured to establish a third-order tensor of a preset local area of the hyperspectral image, and perform tensor decomposition on the third-order tensor to obtain a third-order nuclear tensor corresponding to the preset local area.
An operation module 402, configured to traverse each front slice in the third-order kernel tensor, and establish a feature statistical histogram of each front slice by using a preset coding function.
The classification module 403 is configured to connect the feature statistical histograms of the front slices to obtain a statistical histogram of the preset local area, and classify the hyperspectral image by using the statistical histogram of the preset local area.
The device for improving the hyperspectral image classification precision provided by the embodiment of the invention can realize that: establishing a third-order tensor of a preset local area of the hyperspectral image, and carrying out tensor decomposition on the third-order tensor to obtain a third-order nuclear tensor corresponding to the preset local area; traversing each front slice in the third-order kernel tensor, and establishing a feature statistical histogram of each front slice by using a preset coding function; and connecting the characteristic statistical histograms of the front slices to obtain a statistical histogram of a preset local area, and classifying the hyperspectral images by using the statistical histogram of the preset local area. Compared with the prior art, the statistical histogram obtained by the embodiment of the invention not only can reflect the information on the hyperspectral image space domain, but also can reflect the information of the spectral domain, so that the classification precision can be effectively improved in the hyperspectral image classification.
Further, referring to fig. 5, fig. 5 is a detailed program module diagram of the operation module 402 in the embodiment of the present invention, where the operation module 402 includes:
a first determining module 501, configured to determine domain feature information of tensor elements in the traversed front slice.
A quantization module 502, configured to quantize the domain feature information to obtain a spatial-spectral domain joint encoded value of the tensor element.
A second determining module 503, configured to determine a feature statistical histogram of the traversed front slice by using the spatial-spectral domain joint encoding value.
The quantization module 502 is specifically configured to:
and adding a preset binomial coefficient into the domain characteristic information, and converting the domain characteristic information carrying the binomial coefficient into decimal numbers to obtain a space-spectral domain joint coding value of the tensor elements.
In addition, the classification module 403 includes:
and the connection module is used for connecting the feature statistical histograms of the front slices according to the position sequence of the front slices in the third-order nuclear tensor.
Compared with the prior art, the device for improving the hyperspectral image classification precision provided by the embodiment of the invention can realize the following steps: the method comprises the steps of establishing a third-order tensor of a preset local area of the hyperspectral image, carrying out tensor decomposition on the third-order tensor to obtain a third-order nuclear tensor corresponding to the preset local area, then establishing a feature statistical histogram of each front section in the third-order nuclear tensor, finally connecting the feature statistical histograms of the front sections to obtain a statistical histogram of the preset local area, and classifying the hyperspectral image by using the statistical histogram. The front slices can reflect the joint distribution condition and the spatial distribution information of the spectral values of different wave bands in the hyperspectral image, so that the information on the spectral domain and the spatial domain of the hyperspectral image can be simultaneously represented through the characteristic statistical histogram of each front slice, and the hyperspectral image is classified by using the statistical histogram, so that the classification precision can be effectively improved.
The embodiment of the invention also provides electronic equipment which comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, and when the processor executes the computer program, the method for improving the classification precision of the hyperspectral image corresponds to each step in each embodiment.
The embodiment of the invention also provides a readable storage medium which is a computer readable storage medium and is stored with a computer program, and when the computer program is executed by a processor, the method for improving the classification precision of the hyperspectral image corresponds to each step in each embodiment.
For better understanding of the present invention, referring to fig. 6, fig. 6 is a schematic structural diagram of an electronic device provided in an embodiment of the present invention. As shown in fig. 6, the electronic apparatus 06 of this embodiment mainly includes: a processor 60, a memory 61 and a computer program 62 stored in the memory 61 and executable on the processor 60, such as a program to improve the accuracy of the classification of hyperspectral images. The processor 60, when executing the computer program 62, implements the steps in the embodiments of the method for improving the classification accuracy of hyperspectral images, such as the steps shown in any of the examples in fig. 1 to 3. Alternatively, the processor 60, when executing the computer program 62, implements the functions of the modules/units in the above-described device embodiments, such as the functions of the modules shown in fig. 4.
The computer program 62 may be divided into one or more modules/units, which are stored in the memory 61 and executed by the processor 60 to implement the present invention. One or more of the modules/units may be a series of computer program instruction segments capable of performing certain functions, which are used to describe the execution of the computer program 62 in the computing device 06. For example, the computer program 62 may be divided into functions of a building module 401, an operation module 402, and a classification module 403 (a module in a virtual device).
The computing device 06 may include, but is not limited to, a processor 60, a memory 61. Those skilled in the art will appreciate that fig. 6 is merely an example of the computing device 06, and does not constitute a limitation of the computing device 06, and may include more or less components than illustrated, or combine certain components, or different components, e.g., the computing device may also include input-output devices, network access devices, buses, etc.
The Processor 60 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory 61 may be an internal storage unit of the computing device 06, such as a hard disk or a memory of the computing device 06. The memory 61 may also be an external storage device of the computing device 06, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) Card, Flash memory Card (Flash Card), etc. provided on the computing device 06. Further, the memory 61 may also include both internal storage units of the computing device 06 and external storage devices. The memory 61 is used to store computer programs and other programs and data required by the computing device. The memory 61 may also be used to temporarily store data that has been output or is to be output.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is merely a logical division, and in actual implementation, there may be other divisions, for example, multiple modules or components may be combined or integrated into another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or modules, and may be in an electrical, mechanical or other form.
The modules described as separate parts may or may not be physically separate, and parts displayed as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.
In addition, functional modules in the embodiments of the present invention may be integrated into one processing module, or each of the modules may exist alone physically, or two or more modules are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode.
The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
It should be noted that, for the sake of simplicity, the above-mentioned method embodiments are described as a series of acts or combinations, but those skilled in the art should understand that the present invention is not limited by the described order of acts, as some steps may be performed in other orders or simultaneously according to the present invention. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred and that no acts or modules are necessarily required of the invention.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
In the above description, for a person skilled in the art, according to the idea of the embodiment of the present invention, there may be changes in the specific implementation manner and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (4)
1. A method for improving the classification precision of hyperspectral images is characterized by comprising the following steps: establishing a third-order tensor of a preset local area of the hyperspectral image, carrying out tensor decomposition on the third-order tensor,
obtaining a third-order nuclear tensor corresponding to the preset local area;
traversing each front slice in the third-order kernel tensor, and establishing a feature statistical histogram of each front slice by using a preset coding function;
connecting the feature statistical histograms of the front slices to obtain a statistical histogram of the preset local area, and classifying the hyperspectral images by using the statistical histogram of the preset local area, wherein,
the step of traversing each front slice in the third-order kernel tensor and establishing a feature statistical histogram of each front slice by using a preset coding function includes:
determining domain feature information of tensor elements in the traversed front slice;
quantizing the domain feature information to obtain a spatial-spectral domain joint coding value of the tensor element; determining a feature statistical histogram of the traversed front slice by using the spatial-spectral domain joint coding value;
the step of quantizing the domain feature information to obtain the spatial-spectral domain joint encoded value of the tensor element includes:
adding a preset binomial coefficient to the domain feature information;
converting the domain characteristic information carrying the binomial coefficient into a decimal number to obtain a space-spectral domain joint coding value of the tensor element;
the step of concatenating the feature statistical histograms of the respective front slices includes:
and connecting the feature statistical histograms of the front slices according to the position sequence of the front slices in the third-order nuclear tensor.
2. The utility model provides a promote device of hyperspectral image classification precision which characterized in that, the device includes: the establishing module is used for establishing a third order tensor of a preset local area of the hyperspectral image and carrying out adjustment on the third order tensor
Carrying out tensor decomposition to obtain a third-order nuclear tensor corresponding to the preset local area;
the operation module is used for traversing each front slice in the third-order kernel tensor and establishing a feature statistical histogram of each front slice by using a preset coding function;
a classification module for connecting the feature statistical histograms of the front slices to obtain a statistical histogram of the preset local area and classifying the hyperspectral image by using the statistical histogram of the preset local area, wherein,
the operation module comprises: the first determination module is used for determining domain feature information of tensor elements in the traversed front slice; the quantization module is used for quantizing the domain characteristic information to obtain a spatial-spectral domain joint coding value of the tensor element;
a second determining module, configured to determine a feature statistical histogram of the traversed front slice by using the spatial-spectral domain joint encoding value;
the quantization module is specifically configured to: adding a preset binomial coefficient into the domain characteristic information, and converting the domain characteristic information carrying the binomial coefficient into decimal numbers to obtain a space-spectral domain joint coding value of the tensor element;
the classification module comprises: and the connection module is used for connecting the feature statistical histograms of the front slices according to the position sequence of the front slices in the third-order nuclear tensor.
3. An electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method for improving the accuracy of hyperspectral image classification as claimed in claim 1 when executing the computer program.
4. A computer-readable storage medium, in which a computer program is stored, which, when being executed by a processor, carries out the steps of the method for improving the accuracy of hyperspectral image classification as claimed in claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811001464.9A CN109034274B (en) | 2018-08-30 | 2018-08-30 | Method, device and equipment for improving hyperspectral image classification precision and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811001464.9A CN109034274B (en) | 2018-08-30 | 2018-08-30 | Method, device and equipment for improving hyperspectral image classification precision and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109034274A CN109034274A (en) | 2018-12-18 |
| CN109034274B true CN109034274B (en) | 2021-01-22 |
Family
ID=64626265
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811001464.9A Active CN109034274B (en) | 2018-08-30 | 2018-08-30 | Method, device and equipment for improving hyperspectral image classification precision and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109034274B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110648275B (en) * | 2019-09-24 | 2023-05-09 | 广州大学 | Multichannel silhouette reconstruction method and device, storage medium and terminal equipment |
| CN113435265A (en) * | 2021-06-09 | 2021-09-24 | 深圳大学 | Hyperspectral image classification method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101151623A (en) * | 2005-01-27 | 2008-03-26 | 剑桥研究和仪器设备股份有限公司 | Classifying image features |
| WO2012136874A1 (en) * | 2011-04-06 | 2012-10-11 | Universitat Autonoma De Barcelona | Method for the characterisation and classification of kidney stones |
| WO2016112430A1 (en) * | 2015-01-14 | 2016-07-21 | The University Of Sydney | Hyperspectral imager method and apparatus |
| CN106022391A (en) * | 2016-05-31 | 2016-10-12 | 哈尔滨工业大学深圳研究生院 | Hyperspectral image characteristic parallel extraction and classification method |
| CN107133958A (en) * | 2017-04-26 | 2017-09-05 | 北方工业大学 | Optical remote sensing ship slice segmentation method based on block particle size pre-judging balance histogram |
-
2018
- 2018-08-30 CN CN201811001464.9A patent/CN109034274B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101151623A (en) * | 2005-01-27 | 2008-03-26 | 剑桥研究和仪器设备股份有限公司 | Classifying image features |
| WO2012136874A1 (en) * | 2011-04-06 | 2012-10-11 | Universitat Autonoma De Barcelona | Method for the characterisation and classification of kidney stones |
| WO2016112430A1 (en) * | 2015-01-14 | 2016-07-21 | The University Of Sydney | Hyperspectral imager method and apparatus |
| CN106022391A (en) * | 2016-05-31 | 2016-10-12 | 哈尔滨工业大学深圳研究生院 | Hyperspectral image characteristic parallel extraction and classification method |
| CN107133958A (en) * | 2017-04-26 | 2017-09-05 | 北方工业大学 | Optical remote sensing ship slice segmentation method based on block particle size pre-judging balance histogram |
Non-Patent Citations (2)
| Title |
|---|
| 《基于张量的高光谱遥感图像压缩研究》;王磊;《中国优秀硕士学位论文全文数据库 信息科技辑》;20160331;全文 * |
| 《基于数据融合的高光谱遥感图像分类研究》;朱志球;《中国优秀硕士学位论文全文数据库 信息科技辑》;20100630;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109034274A (en) | 2018-12-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Salmon et al. | Poisson noise reduction with non-local PCA | |
| CN107633216B (en) | Three-dimensional surface space spectrum combined feature coding method and device for hyperspectral remote sensing image | |
| Xu et al. | Fast tensor nuclear norm for structured low-rank visual inpainting | |
| CN109034274B (en) | Method, device and equipment for improving hyperspectral image classification precision and storage medium | |
| CN104392207B (en) | A kind of feature coding method recognized for digital image content | |
| US20230401756A1 (en) | Data Encoding Method and Related Device | |
| CN111783774A (en) | Image processing method, apparatus and storage medium | |
| Zhang et al. | A separation–aggregation network for image denoising | |
| Otair et al. | Improved near-lossless technique using the Huffman coding for enhancing the quality of image compression | |
| CN111488810A (en) | Face recognition method and device, terminal equipment and computer readable medium | |
| Sheppard et al. | A vector quantizer for image restoration | |
| Qi et al. | Morphology-based visible-infrared image fusion framework for smart city | |
| CN113435265A (en) | Hyperspectral image classification method and device, electronic equipment and storage medium | |
| WO2020042070A1 (en) | Method for improving classification accuracy of hyperspectral image, device, apparatus, and storage medium | |
| WO2023051335A1 (en) | Data encoding method, data decoding method, and data processing apparatus | |
| Wang et al. | An improved coupled dictionary and multi-norm constraint fusion method for CT/MR medical images | |
| Biswas et al. | MRI and CT image indexing and retrieval using steerable pyramid transform and local neighborhood difference pattern | |
| CN113628148A (en) | Infrared image noise reduction method and device | |
| CN114049491A (en) | Fingerprint segmentation model training method, fingerprint segmentation device, fingerprint segmentation equipment and fingerprint segmentation medium | |
| Zhang et al. | Randomized methods in lossless compression of hyperspectral data | |
| Chervyakov et al. | Rns-based image processing | |
| CN106874824B (en) | Frequency band selection method and device for palm print identification | |
| CN112927150B (en) | Water reflection area spectrum recovery method of hyperspectral image and terminal device | |
| Yang et al. | Learning compressive sampling via multiscale and steerable support value transform | |
| CN110689513A (en) | Color image fusion method and device and terminal equipment |
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 |