CN112561883A - Method for reconstructing hyperspectral image from crop RGB image - Google Patents

Method for reconstructing hyperspectral image from crop RGB image Download PDF

Info

Publication number
CN112561883A
CN112561883A CN202011494670.5A CN202011494670A CN112561883A CN 112561883 A CN112561883 A CN 112561883A CN 202011494670 A CN202011494670 A CN 202011494670A CN 112561883 A CN112561883 A CN 112561883A
Authority
CN
China
Prior art keywords
image
hyperspectral
pixel
matrix
rgb
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
Application number
CN202011494670.5A
Other languages
Chinese (zh)
Other versions
CN112561883B (en
Inventor
易强
王政
于洪志
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chengdu Asionstar Technology Co ltd
Original Assignee
Chengdu Asionstar Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chengdu Asionstar Technology Co ltd filed Critical Chengdu Asionstar Technology Co ltd
Priority to CN202011494670.5A priority Critical patent/CN112561883B/en
Publication of CN112561883A publication Critical patent/CN112561883A/en
Application granted granted Critical
Publication of CN112561883B publication Critical patent/CN112561883B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
    • G06T3/4007—Scaling of whole images or parts thereof, e.g. expanding or contracting based on interpolation, e.g. bilinear interpolation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30181—Earth observation
    • G06T2207/30188—Vegetation; Agriculture
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A40/00—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production
    • Y02A40/10—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in agriculture

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Quality & Reliability (AREA)
  • Image Processing (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a method for reconstructing a hyperspectral image from an RGB (red, green and blue) image of a crop, which comprises the steps of utilizing a common camera and a hyperspectral camera to collect images in the same environment, and utilizing mark points to convert coordinate systems, so that the coordinate systems of two types of images are converted into the same coordinate system; through data cleaning, the sizes of the two types of images are consistent, and pixel points are in one-to-one correspondence; establishing a matrix equation by using data acquired by the two types of images, and calculating a conversion matrix from the RGB image of the common camera to the hyperspectral image; converting the RGB image into the hyperspectral image by using the conversion matrix; and further adopting a Hermite piecewise interpolation algorithm to carry out polynomial interpolation, so that the spectrum curve is subjected to expansion fitting to obtain a spectrum image on the hyperspectral expansion frequency point. The invention can greatly reduce the cost of detecting crop diseases by utilizing an image processing technology, makes up the defect that a hyperspectral camera cannot be used for collecting spectral images with specific frequencies, and is suitable for detecting novel diseases on an extended frequency point.

Description

Method for reconstructing hyperspectral image from crop RGB image
Technical Field
The invention relates to a method for reconstructing a hyperspectral image from a crop RGB image, and belongs to the technical field of intelligent monitoring of crops.
Background
From the current development of agriculture, wisdom agriculture can lead the agricultural field to intelligent, high-quality development with science and technology. As a concrete expression of an intelligent economic form in agriculture, the intelligent agriculture applies computer technology and Internet of things technology to traditional agriculture, so that the constraints of 'eating in the sky' and 'planting in the market force' are gradually overcome in agricultural production. In the greenhouse, proper air temperature and humidity, soil humidity, carbon dioxide concentration and various crop growth environment parameters can be detected and adjusted through a computer screen; the sowing and harvesting can be finished in a 'quick, good and economic' way by using advanced agricultural machinery in a wide field. The intelligent agriculture promotes the high-quality development of agriculture, and lays a solid foundation for the development of the agricultural industry.
Crop diseases refer to the pathological changes of the form, physiology and biochemistry of crops under the influence of biological or non-biological factors, which can hinder the normal growth, development and fruiting process of plants. The fine management is the inevitable trend of the global agricultural development, the technical basis is the acquisition of farmland and crop information, and how to quickly acquire crop disease information in real time is a key problem for realizing the fine management of the agriculture and improving the yield and the quality of crops.
The crop image-based disease detection technology mainly comprises the technical means of early detection of crop leaf surface diseases based on vein features, crop leaf surface disease detection based on image texture features, crop leaf surface disease detection based on spectrum features and the like, wherein the spectral features are utilized to detect the crop leaf surface diseases by adopting images with specified spectral frequencies shot by a hyperspectral camera, and the detection method is the mainstream detection means at present.
In the actual agricultural production process, the problems of high cost, large volume, heavy weight and the like of a hyperspectral camera exist, especially when the unmanned aerial vehicle is used for shooting, the load of the unmanned aerial vehicle is increased, and the cost of agricultural production is further increased; on the other hand, in a hyperspectral camera, according to the specific design of optical devices and electronic devices in the camera, a certain spectral frequency band is usually preset, and imaging can be performed only on the series of spectral frequency bands. When a specific disease needs to be analyzed, the hyperspectral camera cannot acquire spectral characteristics on non-preset frequency points, and spectral imaging and corresponding disease analysis on specified frequency cannot be realized.
Therefore, how to reconstruct a spectrum image shot by the hyperspectral camera by using an RGB image shot by a common camera and realize spectrum continuity meets the requirement of specific frequency spectrum imaging becomes one of important technologies for detecting crop diseases by using spectral characteristics. The method for reconstructing the hyperspectral image from the crop RGB image can greatly reduce the cost of disease detection, and can be popularized and applied to the intelligent agricultural production practice in a large area.
Disclosure of Invention
The invention aims to overcome the problems in the prior art and provides a method for reconstructing a hyperspectral image from an RGB (red, green and blue) image of a crop. The invention can greatly reduce the cost of crop disease detection by using an image technology, makes up the defect that a hyperspectral camera cannot be used for collecting specific frequency spectrum images, and can be used for detecting novel diseases on an extended frequency point.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a method for reconstructing a hyperspectral image from an RGB image of a crop is characterized by comprising the following steps:
a. crop image data acquisition:
respectively collecting images of crops by using a common camera and a hyperspectral camera under the same environment;
b. image pixel coordinate system conversion:
calculating a coordinate system conversion parameter by adopting a Morojinsky parameter model to convert the coordinate systems of the two pictures; c. Data cleaning:
through data cleaning, the sizes of the two types of pictures are consistent, and pixel points are in one-to-one correspondence;
d. conversion of RGB image to hyperspectral image:
establishing an RGB (red, green and blue) acquisition matrix of a common camera and an intensity matrix of a hyperspectral camera, and calculating a conversion matrix under the condition of obtaining the minimum mean square error by using a minimum quadratic method to convert an RGB image into a hyperspectral image;
e. fitting by high spectral curve extension:
and performing polynomial interpolation by adopting a Hermite piecewise interpolation algorithm, so that the spectrum curve is subjected to expansion fitting to obtain a spectrum image on the hyperspectral expansion frequency point.
In the step a, 4 marking points Q1, Q2, Q3 and Q4 are deployed around the collected crops and are used for positioning images and converting a camera coordinate system; the collected images are RGB images of a common camera and a plurality of images of a hyperspectral camera on different frequency bands; the image collected by the ordinary camera is A, the image collected by the hyperspectral camera is B, the frequency range number of the hyperspectral camera is n, and then the image collected by the n frequency ranges of the hyperspectral camera is Bi(i＝1,2...n)。
In the step b, the pixel coordinates of the marking points Q1, Q2, Q3 and Q4 are recorded as X in the image A acquired by the common cameraA＝(xA,i,yA,i) Wherein i is 1,2,3, 4 is the serial number of 4 marking points; image B acquired at hyperspectral camerai(i 1,2.. n), one image is arbitrarily selected, and the pixel coordinates of the mark points Q1, Q2, Q3 and Q4 are recorded as XB＝(xB,i,yB,i) Wherein, i is 1,2,3, 4 are the serial numbers of 4 marked points.
The coordinate transformation formula of the Morojinskyseven parameter model is XB＝Xp+(1+α)R(XA-Xp) + dX, wherein XpFor the coordinates of the transition points, the geometric gravity center is obtained through Q1, Q2, Q3 and Q4 points, and can be seen as a known value in an equation; xA、XBThe coordinate vectors of two coordinate systems are acquired by points Q1, Q2, Q3 and Q4 on the image, dX is the translation amount of three coordinates, alpha is a scaling parameter, and R is a rotation matrix generated by the rotation angle of three coordinate axes; three rotation angles ω of translation amount dX, scaling amount α, and rotation selection amount Rx、ωy、ωzI.e. seven parameters to be solved using the morokins model, the 7 parameter vectors to be solved can be represented as Y ═ { dx, dy, dz, ωx,ωy,ωzα }; wherein the rotation matrix can be represented as the product of three sub-matrices:
Figure RE-GDA0002944146630000021
it follows that the matrix equation XB＝X0+(1+α)R(XA-X0) + dX is a nonlinear equation, and the nonlinear equation is solved by using a gauss-newton method and an iterative algorithm, and finally 7 parameter vectors Y ═ { dX, dy, dz, ω are solvedx,ωy,ωzAnd alpha, completing the coordinate transformation.
The coordinate transformation specifically comprises the following steps:
1. giving an initial value of Y0；
2. From the functional form f (X) X- [ X ] of the Morogkins matrix equationp+(1+α)R(XA-Xp)+dX]Obtaining a Jacobian matrix expression J through differential calculation of f (X); substitution into Y0Calculating J (Y)0)；
3. Calculate H (Y)0)＝JT(Y0)*J(Y0)，B(Y0)＝-JT(Y0)*f(Y0) Wherein f (X) is ═ X- [ Xp+(1+α)R(XA-Xp)+dX]；
4. Solving an equation H, delta Y and B to obtain delta Y;
5. if the delta Y is smaller than the set threshold, stopping iteration, wherein the Y value at the moment is seven parameter values to be solved; otherwise, setting Y as Y0And (5) repeating the steps 2,3 and 4, and repeating the iterative calculation.
7 parameters Y ═ { dx, dy, dz, ω) obtained by calculation to coordinate conversionx,ωy,ωzAlpha, when the actual coordinate is converted and calculated, the matrix equation X is usedB＝X0+(1+α)R(XA-X0) And + dX, high-precision conversion of coordinates between the two images can be carried out.
The step c comprises the following specific steps:
1. selecting a basic image: selecting an image A as a basic image;
2. image cutting: on the A image, according to the range of the target crop, selecting the pixel coordinate range to be analyzed as X on the imageA，min、XA，maxForming a rectangular frame, and performing the following pixel coordinate conversion and pixel convergence in the rectangular frame; mixing XA，min、XA，maxRespectively substituted into coordinate conversion formula XB＝X0+(1+α)R(XA-X0) + dX, calculating to obtain XB，min、XB，maxThereby forming a rectangular frame on picture B; a, B pixel data within a rectangular frame on two images are acquired separately, forming A, B image dataset PA＝{R，G，B，XA},PB＝{I，XBWhere R, G, B is the pixel three channel value, I is the spectral intensity value, XA、XBCoordinate vectors under respective coordinate systems;
3. and (3) pixel coordinate conversion: for PB＝{I，XBFirstly, the matrix is converted into the formula XB＝X0+(1+α)R(XA-X0) + dX conversion to XA＝X0+R-1(XB-dX-X0) V (1+ α), using this formula, P isBCollective coordinate XBConverting into coordinates on the A image coordinate system to form a B image sampling data set P after coordinate conversionB->A＝{I，XB->A}；
4. Image pixel convergence: will PB->A＝{I，XB->AX coordinates of } XB->ARounding, namely, taking the average value of the data of the same integer coordinate value as the data value of the integer pixel point;
through the steps, A, B two types of image acquisition data correspond to each other according to the pixel coordinates, and the two types of images correspond to the same physical position on the same pixel coordinate point; b, converting the pixel sampling values of the image set to form an image set D after coordinate conversioni(i＝1,2...n)。
The step d specifically comprises:
1. for coordinate transformed and data scrubbed image C, Di(i 1,2.. n), the pixel positions and the coordinates of which correspond one to one, and the sampling data on the same position point is marked; for image C, all pixels are traversed,establishing RGB acquisition matrix
Figure RE-GDA0002944146630000041
R, G, B is a sampling value on a pixel point of 3 channels of RGB on the pixel, and m is the total number of the pixel; for a set of images Di(i 1,2.. n), extracting the intensity value on each image by using the same pixel traversal mode of the image C to form a hyperspectral camera intensity sampling value matrix
Figure RE-GDA0002944146630000042
Wherein I is a sampling value of the strength on the pixel point, m is the total number of the pixels, n is the frequency band number of the hyperspectral camera, and Ii,jIn the formula, i is a pixel serial number, and j is a frequency band serial number; further establishing a conversion matrix equation of the S, I matrix as I ═ S × T, wherein T is a 3 × n conversion matrix to be solved; the equation is subjected to matrix transformation to form a new matrix equation with the form of (S)TS)-1ST*I*IT＝T*ITWherein, superscript T is the matrix transposition, and superscript-1 is the inverse of the matrix; let R ═ STS)-1ST*I*IT，P＝ITThe matrix equation is converted into a standard matrix equation R ═ T × P, wherein R, P is obtained by calculating a sampling value matrix S, I, and T is a conversion matrix for solving the tape; aiming at the matrix equation R ═ T ═ P, a least square method is utilized to calculate the transformation matrix T ═ P (P) under the condition of minimum mean square errorTP)-1PTR；
2. Calculating intensity values on different frequency bands by calculating the obtained conversion matrix T, using RGB channel numerical values of pixel points on the RGB image and I ═ S × T, and completing fitting calculation from the RGB image to the high spectral frequency band intensity;
3. traversing each pixel point on the RGB image, calculating I (S) T, and obtaining the intensity value of n hyperspectral frequency bands on each pixel point; and (4) arranging pixel points according to the hyperspectral frequency bands, reconstructing a spectral image aiming at each frequency band, and finishing spectral reconstruction images of n frequency bands.
In the step e, a conversion matrix T is calculated by using the data of the n frequency bands generated by the hyperspectral camera and the RGB images shot in the same environment, and the hyperspectral images of the n frequency bands are reconstructed from the RGB images.
In the step e, for an RGB channel sampling value (R, G, B) of a certain pixel on the RGB image, the spectral intensity (I) at n frequency bands is calculated by using I ═ S × T1,I2...,In) The center frequency of n frequency bands is (f)1,f2...,fn) And constructing a dispersion curve I (f) of the I-f, interpolating by using a Hermite interpolation piecewise curve fitting method to generate a smooth curve I (f), substituting a corresponding frequency value f in the curve, calculating a spectral intensity value on the frequency, and completing fitting calculation of the spectral intensity from the RGB pixel point to the frequency f.
The fitting calculation steps of the spectrum intensity from the RGB pixel point to the frequency f are as follows:
for the piecewise fitting of the spectral intensity I, use is made of (I)1,I2...,In) For two adjacent data points Ii,Ii+1Performing two-point three-time Hermite segmented interpolation, wherein the interpolation function is a 3-time function Hi(f)＝a0+a1f+a2f2+a3f3Wherein f is a frequency variable; during Hermite piecewise interpolation, Hi(f) The function satisfies:
1、Ii、Ii+1in function Hi(f) The above step (1);
2. to satisfy the smoothness of the function, at the end point IiAbove, satisfy Hi(f) Derivative and Hi-1(f) Is equal at the endpoint Ii+1Above, satisfy Hi(f) Derivative and Hi+1(f) Are equal, so that the smoothness of the fitting function is satisfied;
substituting the obtained 4 known conditions into the 3-degree function Hi(f)＝a0+a1f+a2f2+a3f3And solving in sections to fit undetermined coefficient ai(I-0, 1,2,3) to obtain (I)1,I2)......(In-1,In) 3 rd order polynomial fitting between pointsCurve Hi(f)(i＝1,2,....n-1)；
For the frequency f needing fitting calculation, in the frequency set (f)1,f2...,fn) In, judgment of fk＜f≤fk+1H by piecewise fittingk(f) A function, calculating the intensity value at the frequency f to complete the fitting calculation of the intensity;
by using the method, spectral intensity fitting calculation of the frequency f is carried out on each pixel point in the RGB image, the conversion from the RGB image to the spectral image of the frequency f is completed, and the spectral image at the frequency f is reconstructed.
The invention has the advantages that:
the invention provides a method for reconstructing a hyperspectral image from a crop RGB image, which is suitable for reconstructing a hyperspectral image from a crop RGB image shot by a common camera by adopting an advanced image processing technology and an advanced algorithm, effectively reduces the cost of disease detection, is suitable for large-area detection of crop diseases, and has the characteristics of low cost, advancement, practicability, intellectualization, extensibility, high efficiency and the like.
Secondly, the coordinate system conversion of the ordinary camera and the hyperspectral camera is realized by utilizing the Molotkinsseven parameter model. The invention realizes the pixel coordinate conversion of different cameras by using the map conversion technology, has high calculation speed and high precision, and is an efficient and high-precision method for the pixel coordinate conversion of the image.
The collected image data of the ordinary camera and the hyperspectral camera are utilized, the conversion matrix is established through data cleaning and collection, the conversion equation is established, the least square method is adopted to solve the conversion matrix, the conversion from the RGB image of the ordinary camera to the spectrum image of the hyperspectral camera is efficiently realized, the cost of crop disease detection is effectively reduced, and the method can be widely popularized to the production practice of large-area disease detection of intelligent agriculture.
The discrete spectral intensity is further expanded into a continuous spectral intensity curve by using a Hermite segmented interpolation mathematical method, so that a spectral image on a hyperspectral expansion frequency point is obtained, the defect that the existing hyperspectral camera can only image on a specified frequency point is overcome, and the method is suitable for detecting novel diseases on the expansion frequency point.
In conclusion, the invention greatly reduces the cost of crop disease detection by using an image technology, makes up the defect that a hyperspectral camera cannot be used for collecting specific frequency spectrum images, and can be used for detecting novel diseases on an extended frequency point. Therefore, the technical method can be widely applied to disease detection in intelligent agriculture.
Detailed Description
Firstly, carrying out image acquisition under the same environment by using a common camera and a hyperspectral camera, and carrying out coordinate system conversion by using a Morojinsky seven-parameter model by using mark points to realize the coordinate system conversion of two types of images; then, the data cleaning method is utilized to realize the consistency of the sizes of the two types of images and the one-to-one correspondence of pixel points; then, establishing a matrix equation by using data acquired by the two types of images, and calculating and acquiring a conversion matrix from an RGB image of the common camera to a hyperspectral image by a least square method; the conversion from the RGB image to the hyperspectral image can be realized by utilizing the conversion matrix; and further adopting a Hermite piecewise interpolation algorithm to carry out polynomial interpolation, realizing spectrum curve expansion fitting, and obtaining a spectrum image on the hyperspectral expansion frequency point.
The present invention will be described in further detail below.
A method for reconstructing a hyperspectral image from an RGB image of a crop specifically comprises the following steps:
crop image data acquisition:
under the same environment, respectively collecting images by using a common camera and a hyperspectral camera, and arranging 4 marks around the collected crops for subsequent positioning of the images and conversion of a camera coordinate system; the collected images are RGB images of a common camera and a plurality of images of a hyperspectral camera on different frequency bands.
Image pixel coordinate system conversion:
because the image resolution, the image length-width ratio, the focal length and other camera parameters of the common camera and the hyperspectral camera are different, so that the images have inconsistent displacement, rotation and scaling.
Data cleaning:
aiming at the difference of image resolutions of a common camera and a hyperspectral camera, the data cleaning method is utilized to realize the one-to-one correspondence of pixel points and the consistent sizes of two types of pictures so as to extract data and convert data of the same position points of the pictures in the following process.
Conversion of RGB image to hyperspectral image:
and establishing an RGB (red, green and blue) acquisition matrix of a common camera and an intensity matrix of the hyperspectral camera, and calculating a conversion matrix under the condition of obtaining the minimum mean square error by using a minimum quadratic method to realize the conversion from the RGB image to the hyperspectral image.
Fitting by high spectral curve extension:
aiming at a frequency point f which is not possessed by the hyperspectral camera, performing polynomial interpolation by using a spectrum intensity sampling value of the existing frequency point and using a two-point three-time Hermite segmented interpolation algorithm, realizing hyperspectral curve extension fitting, and acquiring a spectrum image of the hyperspectral camera extension frequency point f.
The crop image data acquisition part:
under the same environment, a common camera and a hyperspectral camera are used for respectively acquiring images. Around the harvested crop, 4 markers (marked with Q1, Q2, Q3 and Q4) are deployed for subsequent positioning of the images and conversion of the camera coordinate system.
The collected images are RGB images of a common camera and a plurality of images of a hyperspectral camera on different frequency bands. The image collected by the ordinary camera is A, the image collected by the hyperspectral camera is B, the frequency range number of the hyperspectral camera is n, and then the image collected by the n frequency ranges of the hyperspectral camera is Bi(i＝1,2...n)。
The image pixel coordinate system conversion section:
recording on an image A acquired by a common cameraThe pixel coordinates of the recording mark points Q1, Q2, Q3 and Q4 are XA＝(xA,i,yA,i) Wherein i is 1,2,3, 4 is the serial number of 4 marking points; image B acquired by n frequency bands of hyperspectral camerai(i 1,2.. n) are all images obtained by classifying the same acquired data according to frequency bands, and the image pixel positions of the images are all completely the same, so that the image B acquired by the hyperspectral camerai(i 1,2.. n), one image is arbitrarily selected, and the pixel coordinates of the mark points Q1, Q2, Q3 and Q4 are recorded as XB＝(xB,i,yB,i) Wherein, i is 1,2,3, 4 are the serial numbers of 4 marked points. Because camera parameters such as resolution, image length-width ratio, focal length and the like of the two cameras are inconsistent, the two images have inconsistent displacement, rotation and scaling, and the method adopts the Morojinskyqi parameter model to convert the coordinate systems of the two images.
The coordinate transformation formula of the Morojinskyqi parameter model is XB＝Xp+(1+α)R(XA-Xp) + dX, wherein XpIn actual calculation, the geometric gravity center is obtained through Q1, Q2, Q3 and Q4 points, and the coordinates are regarded as known values in an equation; xA、XBFor coordinate vectors of two coordinate systems collected at points Q1, Q2, Q3 and Q4 on the image, dX is the translation amount of three coordinates, alpha is a scaling amount scale parameter, and R is a rotation matrix generated by the rotation angles of three coordinate axes. Three rotation angles ω of translation amount dX, scaling amount α, and rotation selection amount Rx、ωy、ωzI.e. seven parameters to be solved using the morokins model, the 7 parameter vectors to be solved can be represented as Y ═ { dx, dy, dz, ωx,ωy,ωzAnd alpha. Wherein the rotation matrix can be represented as the product of three sub-matrices:
Figure RE-GDA0002944146630000071
it follows that the matrix equation XB＝X0+(1+α)R(XA-X0) + dX is a non-linear equation, usually solved by an assumed rotationAngle of rotation omegax、ωy、ωzSmaller, the equation is linearized and then solved using the least squares method. However, considering that the rotation angle may be large when two cameras perform photographing, it is impossible to solve using the linearization method. In the technical scheme, a Gaussian Newton method is adopted, an iterative algorithm is utilized to solve a nonlinear equation, and finally 7 parameter vectors Y ═ { dx, dy, dz and omega are solvedx,ωy,ωzα } to realize high precision coordinate transformation, the specific calculation steps are as follows:
1. giving an initial value of Y0；
2. From the functional form f (X) X- [ X ] of the Morogkins matrix equationp+(1+α)R(XA-Xp)+dX]Obtaining a Jacobian matrix expression J through differential calculation of f (X); substitution into Y0Calculating J (Y)0)；
3. Calculate H (Y)0)＝JT(Y0)*J(Y0)，B(Y0)＝-JT(Y0)*f(Y0) Wherein f (X) is ═ X- [ Xp+(1+α)R(XA-Xp)+dX]；
4. Solving an equation H, delta Y and B to obtain delta Y;
5. if the delta Y is smaller than the set threshold, stopping iteration, wherein the Y value at the moment is seven parameter values to be solved; otherwise, setting Y as Y0And (5) repeating the steps 2,3 and 4, and repeating the iterative calculation.
The above 7 parameters Y ═ dx, dy, dz, ω, which are obtained by calculation using the gauss-newton method, for coordinate transformationx,ωy,ωzAnd alpha. In the actual coordinate conversion calculation, a matrix equation X is utilizedB＝X0+(1+α)R(XA-X0) And + dX, high-precision conversion of coordinates between the two images can be carried out.
The data cleaning part:
in the images A and B, the sizes of the two types of images are consistent and the pixels correspond to each other one by one through a data cleaning method, and the calculation steps are as follows:
1. selecting a basic image: the A image is a visual RGB image, so that the position of the analysis target crop is easy to judge, and therefore the A image is selected as a basic image.
2. Image cutting: on the A image, according to the range of the target crop, selecting the pixel coordinate range to be analyzed as X on the imageA，min、XA，maxForming a rectangular frame, and performing the following pixel coordinate conversion and pixel convergence in the rectangular frame; mixing XA，min、XA，maxRespectively substituted into coordinate conversion formula XB＝X0+(1+α)R(XA-X0) + dX, calculating to obtain XB，min、XB，maxThereby forming a rectangular frame on picture B. A, B pixel data within a rectangular frame on two images are acquired separately, forming A, B image dataset PA＝{R，G，B，XA},PB＝{I，XBWhere R, G, B is the pixel three channel value, I is the spectral intensity value, XA、XBAre coordinate vectors in respective coordinate systems.
3. And (3) pixel coordinate conversion: for PB＝{I，XBFirstly, the matrix is converted into the formula XB＝X0+(1+α)R(XA-X0) + dX conversion to XA＝X0+R-1(XB-dX-X0) V (1+ α), using this formula, P isBCollective coordinate XBConverting into coordinates on the A image coordinate system to form a B image sampling data set P after coordinate conversionB->A＝{I，XB->A}。
4. Image pixel convergence: will PB->A＝{I，XB->AX coordinates of } XB->AAnd rounding, namely taking the average value of the data of the same integer coordinate value as the data value of the integer pixel point.
Through the steps, A, B one-to-one correspondence of two types of image acquisition data according to pixel coordinates is realized, and the two types of images correspond to the same physical position on the same pixel coordinate point. B, converting the pixel sampling values of the image set to form an image set D after coordinate conversioni(i＝1,2...n)。
The conversion part from the RGB image to the hyperspectral image comprises the following steps:
for coordinate transformed and data scrubbed image C, DiAnd (i ═ 1,2.. n), wherein the pixel positions and the coordinates correspond to each other one by one, and the sampling data on the same position point is identified. Traversing all pixels according to a certain mode aiming at the image C, and establishing an RGB acquisition matrix
Figure RE-GDA0002944146630000081
Wherein R, G, B is the sampling value on the pixel point of 3 channels of RGB on the pixel, and m is the total number of pixels. For a set of images Di(i 1,2.. n), extracting the intensity value on each image by using the same pixel traversal mode of the image C to form a hyperspectral camera intensity sampling value matrix
Figure RE-GDA0002944146630000082
Wherein I is a sampling value of the strength on the pixel point, m is the total number of the pixels, n is the frequency band number of the hyperspectral camera, and Ii,jWhere i is the pixel number and j is the band number. The transformation matrix equation of the S, I matrix is further established as I ═ S × T, where T is the 3 × n transformation matrix to be solved. The equation is subjected to matrix transformation to form a new matrix equation in the form of (S)TS)-1ST*I*IT＝T*ITThe superscript T is the matrix transpose, and the superscript-1 is the inverse of the matrix. Let R ═ STS)-1ST*I*IT，P＝ITThe matrix equation can be converted into a standard matrix equation R ═ T × P, where R, P can be obtained by calculation using a sampling value matrix S, I, and T is a conversion matrix for solving the tape; for the matrix equation R ═ T ═ P, the transformation matrix T ═ P (P) under the condition of minimum mean square error can be calculated by using least square methodTP)-1PTR。
Through calculating the obtained transformation matrix T, the RGB channel numerical values of the pixel points on the RGB image can be utilized, and the intensity values on different frequency bands are calculated by utilizing I ═ S × T, so that the fitting calculation from the RGB image to the high spectrum frequency band intensity is realized.
Traversing each pixel point on the RGB image, calculating I (S) T, and obtaining the intensity value of n hyperspectral frequency bands on each pixel point; and (4) arranging the pixel points according to the hyperspectral frequency bands to reconstruct a spectral image aiming at each frequency band, thereby realizing spectral reconstruction images of n frequency bands.
The hyperspectral curve extension fitting method comprises the following steps:
according to the method, under the same environment, the conversion matrix T is calculated by using the data of n frequency bands generated by the hyperspectral camera and the RGB images shot under the same environment, and the hyperspectral images of the n frequency bands are reconstructed from the RGB images. In the method, for the non-preset frequency point f of the hyperspectral camera, a corresponding spectrogram image cannot be given. The invention further discloses a method for fitting the spectrum image of the non-preset frequency point f of the hyperspectral camera by adopting a polynomial fitting mode.
For an RGB channel sampling value (R, G, B) of a certain pixel on an RGB image, the spectral intensity (I) over n frequency bands can be calculated by using I ═ S × T1,I2...,In) The center frequency of n frequency bands is (f)1,f2...,fn) Therefore, a discrete curve I (f) of I-f can be constructed, a smooth curve I (f) is generated by interpolation by using a Hermite interpolation piecewise curve fitting method, and the spectral intensity value on the frequency can be calculated by substituting the corresponding frequency value f in the curve, so that the fitting calculation of the spectral intensity from the RGB pixel point to the frequency f is realized. The calculation steps are as follows:
for the piecewise fitting of the spectral intensity I, use is made of (I)1,I2...,In) For two adjacent data points Ii,Ii+1Performing two-point three-time Hermite segmented interpolation, wherein the interpolation function is a 3-time function Hi(f)＝a0+a1f+a2f2+a3f3Wherein f is a frequency variable. During Hermite piecewise interpolation, Hi(f) The function satisfies:
1、Ii、Ii+1in function Hi(f) The above.
2. To satisfy the smoothness of the function, at the end point IiAbove, satisfy Hi(f) Derivative and Hi-1(f) Is equal at the endpoint Ii+1Above, satisfy Hi(f) Derivative and Hi+1(f) Are equal to satisfy the smoothness of the fitting function.
Substituting the obtained 4 known conditions into the 3-degree function Hi(f)＝a0+a1f+a2f2+a3f3And solving in sections to fit undetermined coefficient ai(I-0, 1,2,3), i.e. (I) can be fitted1,I2)......(In-1,In) 3 rd order polynomial fitting curve H between pointsi(f)(i＝1,2,....n-1)。
For the frequency f needing fitting calculation, in the frequency set (f)1,f2...,fn) In, judgment of fk＜f≤fk+1I.e. using piecewise fitting of Hk(f) And calculating the intensity value at the frequency f by using the function, thereby realizing the fitting calculation of the intensity.
By using the method, spectral intensity fitting calculation of the frequency f is carried out on each pixel point in the RGB image, so that conversion from the RGB image to the spectral image of the frequency f can be realized, and the spectral image at the frequency f is reconstructed.

Claims (10)

1. A method for reconstructing a hyperspectral image from an RGB image of a crop is characterized by comprising the following steps:
a. crop image data acquisition:
respectively collecting images of crops by using a common camera and a hyperspectral camera under the same environment;
b. image pixel coordinate system conversion:
calculating a coordinate system conversion parameter by adopting a Morojinsky parameter model to convert the coordinate systems of the two pictures;
c. data cleaning:
through data cleaning, the sizes of the two types of pictures are consistent, and pixel points are in one-to-one correspondence;
d. conversion of RGB image to hyperspectral image:
establishing an RGB (red, green and blue) acquisition matrix of a common camera and an intensity matrix of a hyperspectral camera, and calculating a conversion matrix under the condition of obtaining the minimum mean square error by using a minimum quadratic method to convert an RGB image into a hyperspectral image;
e. fitting by high spectral curve extension:
and performing polynomial interpolation by adopting a Hermite piecewise interpolation algorithm, so that the spectrum curve is subjected to expansion fitting to obtain a spectrum image on the hyperspectral expansion frequency point.
2. The method for reconstructing the hyperspectral image of the crop RGB image as claimed in claim 1, wherein: in the step a, 4 marking points Q1, Q2, Q3 and Q4 are deployed around the collected crops and are used for positioning images and converting a camera coordinate system; the collected images are RGB images of a common camera and a plurality of images of a hyperspectral camera on different frequency bands; the image collected by the ordinary camera is A, the image collected by the hyperspectral camera is B, the frequency range number of the hyperspectral camera is n, and then the image collected by the n frequency ranges of the hyperspectral camera is Bi(i＝1,2...n)。
3. The method for reconstructing the hyperspectral image of the crop RGB image as claimed in claim 2, wherein: in the step b, the pixel coordinates of the marking points Q1, Q2, Q3 and Q4 are recorded as X in the image A acquired by the common cameraA＝(xA,i,yA,i) Wherein i is 1,2,3, 4 is the serial number of 4 marking points; image B acquired at hyperspectral camerai(i 1,2.. n), one image is arbitrarily selected, and the pixel coordinates of the mark points Q1, Q2, Q3 and Q4 are recorded as XB＝(xB,i,yB,i) Wherein, i is 1,2,3, 4 are the serial numbers of 4 marked points.
4. The method for reconstructing the hyperspectral image of the crop RGB image as claimed in claim 3, wherein: the coordinate transformation formula of the Morojinskyseven parameter model is XB＝Xp+(1+α)R(XA-Xp) + dX, wherein XpFor transition point coordinates, by Q1. The points Q2, Q3 and Q4 are used for obtaining the geometric gravity center, and can be regarded as known values in the equation; xA、XBThe coordinate vectors of two coordinate systems are acquired by points Q1, Q2, Q3 and Q4 on the image, dX is the translation amount of three coordinates, alpha is a scaling parameter, and R is a rotation matrix generated by the rotation angle of three coordinate axes; three rotation angles ω of translation amount dX, scaling amount α, and rotation selection amount Rx、ωy、ωzI.e. seven parameters to be solved using the morokins model, the 7 parameter vectors to be solved can be represented as Y ═ { dx, dy, dz, ωx,ωy,ωzα }; wherein the rotation matrix can be represented as the product of three sub-matrices:
Figure RE-FDA0002944146620000021
it follows that the matrix equation XB＝X0+(1+α)R(XA-X0) + dX is a nonlinear equation, and the nonlinear equation is solved by using a gauss-newton method and an iterative algorithm, and finally 7 parameter vectors Y ═ { dX, dy, dz, ω are solvedx,ωy,ωzAnd alpha, completing the coordinate transformation.
5. The crop RGB image hyperspectral image reconstruction method according to claim 4, wherein the hyperspectral image reconstruction method comprises the following steps: the coordinate transformation specifically comprises the following steps:
1. giving an initial value of Y0；
2. From the functional form f (X) X- [ X ] of the Morogkins matrix equationp+(1+α)R(XA-Xp)+dX]Obtaining a Jacobian matrix expression J through differential calculation of f (X); substitution into Y0Calculating J (Y)0)；
3. Calculate H (Y)0)＝JT(Y0)*J(Y0)，B(Y0)＝-JT(Y0)*f(Y0) Wherein f (X) is ═ X- [ Xp+(1+α)R(XA-Xp)+dX]；
4. Solving an equation H, delta Y and B to obtain delta Y;
5. if the delta Y is smaller than the set threshold, stopping iteration, wherein the Y value at the moment is seven parameter values to be solved; otherwise, setting Y as Y0Step 2, step 3, step 4 are repeated, and iterative calculation is repeated;
7 parameters Y ═ { dx, dy, dz, ω) obtained by calculation to coordinate conversionx,ωy,ωzAlpha, when the actual coordinate is converted and calculated, the matrix equation X is usedB＝X0+(1+α)R(XA-X0) And + dX, high-precision conversion of coordinates between the two images can be carried out.
6. The method for reconstructing the hyperspectral image of the crop RGB image as claimed in claim 5, wherein: the step c comprises the following specific steps:
1. selecting a basic image: selecting an image A as a basic image;
2. image cutting: on the A image, according to the range of the target crop, selecting the pixel coordinate range to be analyzed as X on the imageA，min、XA，maxForming a rectangular frame, and performing the following pixel coordinate conversion and pixel convergence in the rectangular frame; mixing XA，min、XA，maxRespectively substituted into coordinate conversion formula XB＝X0+(1+α)R(XA-X0) + dX, calculating to obtain XB，min、XB，maxThereby forming a rectangular frame on picture B; a, B pixel data within a rectangular frame on two images are acquired separately, forming A, B image dataset PA＝{R，G，B，XA},PB＝{I，XBWhere R, G, B is the pixel three channel value, I is the spectral intensity value, XA、XBCoordinate vectors under respective coordinate systems;
3. and (3) pixel coordinate conversion: for PB＝{I，XBFirstly, the matrix is converted into the formula XB＝X0+(1+α)R(XA-X0) + dX conversion to XA＝X0+R-1(XB-dX-X0) V (1+ α), using this formula, P isBCoordinates of the setsXBConverting into coordinates on the A image coordinate system to form a B image sampling data set P after coordinate conversionB->A＝{I，XB->A}；
4. Image pixel convergence: will PB->A＝{I，XB->AX coordinates of } XB->ARounding, namely, taking the average value of the data of the same integer coordinate value as the data value of the integer pixel point;
through the steps, A, B two types of image acquisition data correspond to each other according to the pixel coordinates, and the two types of images correspond to the same physical position on the same pixel coordinate point; b, converting the pixel sampling values of the image set to form an image set D after coordinate conversioni(i＝1,2...n)。
7. The method for reconstructing the hyperspectral image of the crop RGB image according to claim 6, wherein: the step d specifically comprises:
1. for coordinate transformed and data scrubbed image C, Di(i 1,2.. n), the pixel positions and the coordinates of which correspond one to one, and the sampling data on the same position point is marked; traversing all pixels and establishing an RGB acquisition matrix aiming at the image C
Figure RE-FDA0002944146620000031
R, G, B is a sampling value on a pixel point of 3 channels of RGB on the pixel, and m is the total number of the pixel; for a set of images Di(i 1,2.. n), extracting the intensity value on each image by using the same pixel traversal mode of the image C to form a hyperspectral camera intensity sampling value matrix
Figure RE-FDA0002944146620000032
Wherein I is a sampling value of the strength on the pixel point, m is the total number of the pixels, n is the frequency band number of the hyperspectral camera, and Ii,jIn the formula, i is a pixel serial number, and j is a frequency band serial number; further establishing a conversion matrix equation of the S, I matrix as I ═ S × T, wherein T is a 3 × n conversion matrix to be solved; subjecting the equation to matrix transformationForming a new matrix equation of the form (S)TS)-1ST*I*IT＝T*ITWherein, superscript T is the matrix transposition, and superscript-1 is the inverse of the matrix; let R ═ STS)-1ST*I*IT，P＝ITThe matrix equation is converted into a standard matrix equation R ═ T × P, wherein R, P is obtained by calculating a sampling value matrix S, I, and T is a conversion matrix for solving the tape; aiming at the matrix equation R ═ T ═ P, a least square method is utilized to calculate the transformation matrix T ═ P (P) under the condition of minimum mean square errorTP)-1PTR；
2. Calculating intensity values on different frequency bands by calculating the obtained conversion matrix T, using RGB channel numerical values of pixel points on the RGB image and I ═ S × T, and completing fitting calculation from the RGB image to the high spectral frequency band intensity;
3. traversing each pixel point on the RGB image, calculating I (S) T, and obtaining the intensity value of n hyperspectral frequency bands on each pixel point; and (4) arranging pixel points according to the hyperspectral frequency bands, reconstructing a spectral image aiming at each frequency band, and finishing spectral reconstruction images of n frequency bands.
8. The method for reconstructing the hyperspectral image of the crop RGB image according to claim 7, wherein: in the step e, a conversion matrix T is calculated by using the data of the n frequency bands generated by the hyperspectral camera and the RGB images shot in the same environment, and the hyperspectral images of the n frequency bands are reconstructed from the RGB images.
9. The method for reconstructing the hyperspectral image of the crop RGB image according to claim 8, wherein: in the step e, for an RGB channel sampling value (R, G, B) of a certain pixel on the RGB image, the spectral intensity (I) at n frequency bands is calculated by using I ═ S × T1,I2...,In) The center frequency of n frequency bands is (f)1,f2...,fn) Constructing a discrete curve I (f) of I-f, further utilizing a Hermite interpolation piecewise curve fitting method to interpolate to generate a smooth curve I (f),and substituting the corresponding frequency value f into the curve, calculating the spectral intensity value at the frequency, and finishing the fitting calculation of the spectral intensity from the RGB pixel point to the frequency f.
10. The method for reconstructing a hyperspectral image from an RGB image of a crop as claimed in claim 9, wherein: the fitting calculation steps of the spectrum intensity from the RGB pixel point to the frequency f are as follows:
for the piecewise fitting of the spectral intensity I, use is made of (I)1,I2...,In) For two adjacent data points Ii,Ii+1Performing two-point three-time Hermite segmented interpolation, wherein the interpolation function is a 3-time function Hi(f)＝a0+a1f+a2f2+a3f3Wherein f is a frequency variable; during Hermite piecewise interpolation, Hi(f) The function satisfies:
1、Ii、Ii+1in function Hi(f) The above step (1);
2. to satisfy the smoothness of the function, at the end point IiAbove, satisfy Hi(f) Derivative and Hi-1(f) Is equal at the endpoint Ii+1Above, satisfy Hi(f) Derivative and Hi+1(f) Are equal, so that the smoothness of the fitting function is satisfied;
substituting the obtained 4 known conditions into the 3-degree function Hi(f)＝a0+a1f+a2f2+a3f3And solving in sections to fit undetermined coefficient ai(I-0, 1,2,3) to obtain (I)1,I2)......(In-1,In) 3 rd order polynomial fitting curve H between pointsi(f)(i＝1,2,....n-1)；
For the frequency f needing fitting calculation, in the frequency set (f)1,f2...,fn) In, judgment of fk＜f≤fk+1H by piecewise fittingk(f) A function, calculating the intensity value at the frequency f to complete the fitting calculation of the intensity;
by using the method, spectral intensity fitting calculation of the frequency f is carried out on each pixel point in the RGB image, the conversion from the RGB image to the spectral image of the frequency f is completed, and the spectral image at the frequency f is reconstructed.
CN202011494670.5A 2020-12-17 2020-12-17 Method for reconstructing hyperspectral image by crop RGB image Active CN112561883B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011494670.5A CN112561883B (en) 2020-12-17 2020-12-17 Method for reconstructing hyperspectral image by crop RGB image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011494670.5A CN112561883B (en) 2020-12-17 2020-12-17 Method for reconstructing hyperspectral image by crop RGB image

Publications (2)

Publication Number Publication Date
CN112561883A true CN112561883A (en) 2021-03-26
CN112561883B CN112561883B (en) 2024-06-21

Family

ID=75064451

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011494670.5A Active CN112561883B (en) 2020-12-17 2020-12-17 Method for reconstructing hyperspectral image by crop RGB image

Country Status (1)

Country Link
CN (1) CN112561883B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114663785A (en) * 2022-03-18 2022-06-24 华南农业大学 Litchi disease detection method and system based on hyperspectrum of unmanned aerial vehicle
CN115753691A (en) * 2022-08-23 2023-03-07 合肥工业大学 Water quality parameter detection method based on RGB reconstruction hyperspectrum
CN116188465A (en) * 2023-04-26 2023-05-30 济宁市保田农机技术推广专业合作社 Crop growth state detection method based on image processing technology
CN118505573A (en) * 2024-07-18 2024-08-16 奥谱天成(湖南)信息科技有限公司 Spectral data recovery method, device and storage medium

Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2959092A1 (en) * 2010-04-20 2011-10-21 Centre Nat Rech Scient DIGITAL COMPENSATION PROCESSING OF SIGNALS FROM PHOTOSITES OF A COLOR SENSOR
CN103808286A (en) * 2012-11-08 2014-05-21 谢荣 Total station-based steel structure three dimensional precision detection analysis method and application thereof
CN107133976A (en) * 2017-04-24 2017-09-05 浙江大学 A kind of method and apparatus for obtaining three-dimensional hyperspectral information
US20170354358A1 (en) * 2016-06-13 2017-12-14 Umm Al-Qura University Hyperspectral image visualization in patients with medical conditions
CN107783937A (en) * 2017-10-19 2018-03-09 西安科技大学 A kind of method for solving any anglec of rotation three-dimensional coordinate conversion parameter
CN108520495A (en) * 2018-03-15 2018-09-11 西北工业大学 Hyperspectral image super-resolution reconstruction method based on cluster manifold priori
CN108759665A (en) * 2018-05-25 2018-11-06 哈尔滨工业大学 A kind of extraterrestrial target reconstruction accuracy analysis method based on coordinate conversion
CN109146787A (en) * 2018-08-15 2019-01-04 北京理工大学 A kind of real-time reconstruction method of the double camera spectrum imaging system based on interpolation
CN109978931A (en) * 2019-04-04 2019-07-05 北京悉见科技有限公司 Method for reconstructing three-dimensional scene and equipment, storage medium
CN110520768A (en) * 2017-06-05 2019-11-29 上海科技大学 EO-1 hyperion optical field imaging method and system
WO2020199205A1 (en) * 2019-04-04 2020-10-08 合刃科技（深圳）有限公司 Hybrid hyperspectral image reconstruction method and system

Patent Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2959092A1 (en) * 2010-04-20 2011-10-21 Centre Nat Rech Scient DIGITAL COMPENSATION PROCESSING OF SIGNALS FROM PHOTOSITES OF A COLOR SENSOR
CN103808286A (en) * 2012-11-08 2014-05-21 谢荣 Total station-based steel structure three dimensional precision detection analysis method and application thereof
US20170354358A1 (en) * 2016-06-13 2017-12-14 Umm Al-Qura University Hyperspectral image visualization in patients with medical conditions
CN107133976A (en) * 2017-04-24 2017-09-05 浙江大学 A kind of method and apparatus for obtaining three-dimensional hyperspectral information
CN110520768A (en) * 2017-06-05 2019-11-29 上海科技大学 EO-1 hyperion optical field imaging method and system
CN107783937A (en) * 2017-10-19 2018-03-09 西安科技大学 A kind of method for solving any anglec of rotation three-dimensional coordinate conversion parameter
CN108520495A (en) * 2018-03-15 2018-09-11 西北工业大学 Hyperspectral image super-resolution reconstruction method based on cluster manifold priori
CN108759665A (en) * 2018-05-25 2018-11-06 哈尔滨工业大学 A kind of extraterrestrial target reconstruction accuracy analysis method based on coordinate conversion
CN109146787A (en) * 2018-08-15 2019-01-04 北京理工大学 A kind of real-time reconstruction method of the double camera spectrum imaging system based on interpolation
CN109978931A (en) * 2019-04-04 2019-07-05 北京悉见科技有限公司 Method for reconstructing three-dimensional scene and equipment, storage medium
WO2020199205A1 (en) * 2019-04-04 2020-10-08 合刃科技（深圳）有限公司 Hybrid hyperspectral image reconstruction method and system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
库静: ""不同高光谱成像方式的马铃薯内外部品质检测方法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》, 15 February 2017 (2017-02-15) *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114663785A (en) * 2022-03-18 2022-06-24 华南农业大学 Litchi disease detection method and system based on hyperspectrum of unmanned aerial vehicle
CN114663785B (en) * 2022-03-18 2024-06-07 华南农业大学 Unmanned aerial vehicle hyperspectral-based litchi disease detection method and system
CN115753691A (en) * 2022-08-23 2023-03-07 合肥工业大学 Water quality parameter detection method based on RGB reconstruction hyperspectrum
CN116188465A (en) * 2023-04-26 2023-05-30 济宁市保田农机技术推广专业合作社 Crop growth state detection method based on image processing technology
CN118505573A (en) * 2024-07-18 2024-08-16 奥谱天成(湖南)信息科技有限公司 Spectral data recovery method, device and storage medium
CN118505573B (en) * 2024-07-18 2024-10-01 奥谱天成(湖南)信息科技有限公司 Spectral data recovery method, device and storage medium

Also Published As

Publication number Publication date
CN112561883B (en) 2024-06-21

Similar Documents

Publication Publication Date Title
CN112561883A (en) Method for reconstructing hyperspectral image from crop RGB image
CN112418188B (en) Crop growth whole-course digital evaluation method based on unmanned aerial vehicle vision
Hall et al. Characterising and mapping vineyard canopy using high-spatial-resolution aerial multispectral images
CN102829739B (en) Object-oriented remote sensing inversion method of leaf area index of crop
CN110211044B (en) Multispectral imaging method and device based on demosaicing algorithm and principal component analysis
CN109102469A (en) A kind of panchromatic sharpening method of remote sensing images based on convolutional neural networks
CN108288256B (en) Multispectral mosaic image restoration method
CN112614147B (en) Crop seedling stage plant density estimation method and system based on RGB image
CN112200854B (en) Leaf vegetable three-dimensional phenotype measuring method based on video image
CN113177548B (en) Key area identification method for immune fixed electrophoresis
CN115272187A (en) Vehicle-mounted dynamic field frame-to-frame relevance based field rice and wheat lodging global evaluation method
CN104318583A (en) Visible light broadband spectrum image registration method
Ahamed et al. Tower remote-sensing system for monitoring energy crops; image acquisition and geometric corrections
CN105069749A (en) Splicing method for tire mold images
CN112465832A (en) Single-sided tree point cloud skeleton line extraction method and system based on binocular vision
CN114612794A (en) Remote sensing identification method for land covering and planting structure in finely-divided agricultural area
CN110736750B (en) Wheat scab detection method based on multi-angle field high-definition imaging
He et al. A calculation method of phenotypic traits of soybean pods based on image processing technology
CN112836678B (en) Intelligent planning method for intelligent agricultural park
CN107818565B (en) Method for investigating quantity of wheat seedlings in one-to-three-leaf period of wheat
CN111896045B (en) Greenhouse crop three-dimensional directional sensing and fine-grained automatic acquisition device and method
CN111798378A (en) Thermal infrared image super-resolution reconstruction evaluation method
CN109410165B (en) Multispectral remote sensing image fusion method based on classification learning
Schneider et al. Towards predicting vine yield: Conceptualization of 3d grape models and derivation of reliable physical and morphological parameters
CN115841615A (en) Tobacco yield prediction method and device based on multispectral data of unmanned aerial vehicle

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