CN116630750A - Polarized light image processing method, polarized light image processing system, computer device, and storage medium - Google Patents

Polarized light image processing method, polarized light image processing system, computer device, and storage medium Download PDF

Info

Publication number
CN116630750A
CN116630750A CN202310896103.XA CN202310896103A CN116630750A CN 116630750 A CN116630750 A CN 116630750A CN 202310896103 A CN202310896103 A CN 202310896103A CN 116630750 A CN116630750 A CN 116630750A
Authority
CN
China
Prior art keywords
polarization
image
polarized
stokes
polarized light
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
CN202310896103.XA
Other languages
Chinese (zh)
Other versions
CN116630750B (en
Inventor
陆翔
吕新政
孙红雨
郭银景
张荣良
孔芳
温安昊
刘增浩
马宁
张帆
王昊
王渌平
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.)
Shandong University of Science and Technology
Original Assignee
Shandong University of Science and Technology
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 Shandong University of Science and Technology filed Critical Shandong University of Science and Technology
Priority to CN202310896103.XA priority Critical patent/CN116630750B/en
Publication of CN116630750A publication Critical patent/CN116630750A/en
Application granted granted Critical
Publication of CN116630750B publication Critical patent/CN116630750B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/096—Transfer learning
    • 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
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • 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
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • 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
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Multimedia (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Image Processing (AREA)

Abstract

The invention belongs to the technical field of polarized light image processing, and particularly discloses a polarized light image processing method, a polarized light image processing system, computer equipment and a storage medium. According to the invention, the nonlinear operation in the existing polarized image fusion is replaced by the linear operation, and the nonlinear operation is fused into the neural network through transfer learning, so that the operation amount is successfully reduced, the polarized image target recognition algorithm is more efficient, and the instantaneity of the polarized image target recognition technology is improved. By simplifying the calculation content, the method is easier to realize in the field of embedded systems such as FPGA and the like, and the polarized light target recognition technology can be better applied to the aspects of low-power-consumption equipment, wearable equipment and the like. The method and the device remarkably improve the real-time performance of operation, and have positive effects on the application of real-time polarized light hidden target identification in the fields of explosion prevention, anti-terrorism and the like, and the combined navigation of polarized light navigation and the like.

Description

Polarized light image processing method, polarized light image processing system, computer device, and storage medium
Technical Field
The invention belongs to the technical field of polarized light image processing, and particularly relates to a polarized light image processing method, a polarized light image processing system, computer equipment and a storage medium.
Background
Traditional Chinese medicineStokesThe two-dimensional polarized image is calculated by a parametric method, which is shown in fig. 1, and the specific process is as follows: firstly, utilizeStokesVector [I,Q,U,V]And the incident light and emergent light relational expression formed by Mueller matrix, and through a plurality of light intensity images with different polarization directions, the incident light corresponding to each pixel point is reversely solved after linear transformationStokesVector quantityS in . wherein ,Vrepresenting the difference between the right circularly polarized light intensity and the left circularly polarized light intensity, in engineering practice,Vcalculated generally as 0, to solve for incident lightStokesAfter vector, according to polarized lightStokesThe mathematical relationship between the vector and the polarization angle and the polarization degree solves the polarization angle and the polarization degree of the incident light, and the polarization angle and the polarization degree are contained in the polarization angle and the polarization degreearctanTransport and deliverAnd 3, nonlinearity is introduced into calculation, and the calculation complexity is increased. If the amount of linear operation in the operation corresponding to the single pixel point isoThe nonlinear operation takes time asa*o(since the amount of non-linear operation is generally larger than that of linear operation)a>1) Since the polarization processing of the image requires traversing each pixel, for one sheetm*nThe above operation is performed on the image of (2) and the time is additionally increasedm*n*a*oThe larger the image is, the larger the operation amount required by nonlinear operation is, so that the polarized image target identification efficiency is lower, and the real-time performance of the polarized image target identification cannot be ensured.
Disclosure of Invention
The invention aims to provide a processing method of a polarized light image, which aims to solve the technical problem of increased operation amount in the polarized light image processing process caused by nonlinear operation in the existing polarized image fusion and neural network identification, thereby improving the efficiency of the polarized image target identification method and further improving the real-time performance of the polarized image target identification.
In order to achieve the above purpose, the invention adopts the following technical scheme:
a method of processing a polarized light image, comprising the steps of:
step 1, firstly, carrying out polarized light intensity image training on each group of polarized light intensity images in a polarized light intensity image training setStokesVector solution to obtainStokesVector diagram, willStokesVector graphics in combination with corresponding tag data compositionStokesA vector dataset;
step 2, forStokesEach of the vector data setsStokesThe vector diagram is used for calculating a polarized image data set in a traditional polarized light fusion calculation mode, and each group of polarized images in the polarized image data set comprise a polarized angle image and a polarized degree image;
step 3. According to step 1StokesCalculating quantization parameters by using the vector diagram and the polarized image obtained in the step 2, wherein the quantization parameters comprise slope parametersa、cBias parameterb、d；
Step 4. The step 1 is carried outStokesVector data setCombining the quantization parameters in step 3a、b、c、dAccording to the linear mapping polarization resolving algorithm, resolving to obtain a linear mapping polarization image dataset; each group of linear mapping polarization images in the linear mapping polarization image data set comprises a linear mapping polarization angle image and a linear mapping polarization degree image;
step 5, performing polarization recognition network model training by using the polarization image data set obtained in the step 2 to obtain a traditional polarization image recognition model;
step 6, combining the linear mapping polarized image data set obtained in the step 4 with the traditional polarized image recognition model in the step 5 to perform migration learning so as to obtain a linear mapping polarized image recognition model;
and 7, carrying out polarized image recognition on the image to be recognized acquired by the polarized light camera by utilizing the linear mapping polarized image recognition model obtained in the step 6, and obtaining a polarized image recognition result.
In addition, on the basis of the polarized light image processing method, the invention also provides a polarized light image processing system which is suitable for the polarized light image processing method, and the polarized light image processing system adopts the following technical scheme:
a polarized image processing system comprising:
Stokesa vector resolving module for performing a processing on each group of polarized light intensity images in the polarized light intensity image training setStokesVector solution to obtain correspondingStokesA vector diagram;
will beStokesVector graphics in combination with corresponding tag data compositionStokesA vector dataset;
a traditional polarized light fusion calculation module for aiming atStokesEach of the vector data setsStokesThe vector diagram is used for calculating a polarized image dataset through a traditional polarized light fusion calculation mode;
wherein each set of polarized images in the polarized image dataset comprises a polarization angle image and a polarization degree image;
a quantization parameter calculation module for calculating a quantization parameter based on the obtained quantization parameterStokesCalculating quantization parameters from the vector image and the resulting polarized image, wherein the quantization parameters include slope parametersa、cBias parameterb、d；
A linear mapping polarization resolving module for resolving the obtainedStokesVector data set incorporating quantization parametersa、b、c、dAccording to the linear mapping polarization resolving algorithm, resolving to obtain a linear mapping polarization image dataset;
wherein each set of linear mapped polarization images in the linear mapped polarization image dataset comprises a linear mapped polarization angle image and a linear mapped polarization degree image;
the polarization recognition network module is used for carrying out polarization recognition network model training according to the obtained polarization image dataset to obtain a traditional polarization image recognition model;
the transfer learning module is used for carrying out transfer learning by combining the traditional polarized image recognition model according to the obtained linear mapping polarized image data set to obtain a linear mapping polarized image recognition model;
and the polarized image recognition module is used for carrying out polarized image recognition on the image to be recognized acquired by the polarized light camera by using the obtained linear mapping polarized image recognition model, so as to obtain a polarized image recognition result.
In addition, on the basis of the polarized light image processing method, the invention also provides computer equipment which comprises a memory and one or more processors. The memory stores executable codes, and the processor is used for realizing the steps of the polarized light image processing method when executing the executable codes.
Furthermore, on the basis of the polarized light image processing method, the invention also provides a computer readable storage medium on which a program is stored. The program, when executed by a processor, is adapted to carry out the steps of the above-mentioned polarized light image processing method.
The invention has the following advantages:
as described above, the invention provides a processing method of polarized light images, which replaces the nonlinear operation in the existing polarized image fusion with linear operation, and fuses the nonlinear operation into a neural network through migration learning, so that the operation amount is successfully reduced, the polarized image target recognition algorithm is more efficient, and the real-time performance of the polarized image target recognition technology is improved. By simplifying the calculation content, the method is easier to realize in the field of embedded systems such as FPGA and the like, and the polarized light target recognition technology can be better applied to the aspects of low-power-consumption equipment, wearable equipment and the like. The method and the device remarkably improve the real-time performance of operation, and have positive effects on the application of real-time polarized light hidden target identification in the fields of explosion prevention, anti-terrorism and the like, and the combined navigation of polarized light navigation and the like.
Drawings
FIG. 1 is a conventional viewStokesAnd solving a schematic diagram of the two-dimensional polarized image by a parametric method.
Fig. 2 is a flowchart of a processing method of a polarized light image according to an embodiment of the present invention.
Detailed Description
The invention is described in further detail below with reference to the attached drawings and detailed description:
example 1
As shown in fig. 2, embodiment 1 describes a processing method of polarized light image based onStokesThe hidden target identification combining the parametric method polarization solution image and the neural network specifically comprises the following steps:
step 1, firstly, carrying out polarized light intensity image processing on each group of polarized light intensity images in a polarized light intensity image training setStokesVector solution to obtain correspondingStokesVector diagram, willStokesVector graphics in combination with corresponding tag data compositionStokesVector data sets.
The user provides polarized light intensity images of different polarized angles shot by a plurality of polarized light cameras, and combines the label information of the objects in each group of polarized light intensity images to form a polarized light intensity image training set.
According toStokesVector [I,Q,U,V]And the incident light and emergent light relation (1) formed by Mueller matrix, traversing polarized light intensity images of different angles of each group of the same object in the polarized light intensity image training set, and calculating the incident light Stokes of the polarized light intensity images according to the polarized anglesFour parameters of the vectorI、Q、U、VAnd four parameters of Stokes vectorI´、Q´、U´、VAnd (2) and jointly solving the set of images to obtain the formula (2).
（1）
wherein ,S in representing incident lightStokesThe vector of the vector,S out representing the emitted lightStokesAnd (5) vector.
M θ Representing a specific angleθA Muller matrix corresponding to the polarizer; mueller matrix describes the passage of incident polarized light through a particular angleθChanges in polarization state after the polarizer.
IIndicating the total intensity of the incident light,Qfor polarization differences between the horizontal and vertical directions of the incident light,Ufor polarization differences between +45° and-45 ° directions of incident light,Vrepresenting the difference between the left-hand and right-hand circular polarization of the incident light.
IAnd represents the total intensity of the outgoing light,Qis the polarization difference between the horizontal and vertical directions of the outgoing light,Uis the polarization difference between the +45° and-45 ° directions of the outgoing light,Vand (f) represents the difference between the left-hand and right-hand circular polarization of the exiting light.
Different polarization angles according to the polarized light intensity images of different angles of the same objectθBut the light intensity value corresponding to the pixel point at the same positionI n And (2) reversely calculating the incident light corresponding to the pixel pointStokesVector quantityS in 。
In which incident lightStokesVector quantityS in The inverse solution expression of (2) is represented by the following formula.
The formula (2) table is derived from the formula (1), and the incident light parameter V is calculated according to 0 in engineering practice.
（2）
wherein ,I 1 ´、I 2 ´、I 3 respectively represents that the pixel point at the position is at the first polarization angleθ 1 Light intensity measured in the image at a second polarization angleθ 2 Light intensity measured in image, at third polarization angleθ 3 The measured light intensity in the image.
For each group of polarized light intensity images, calculating polarized light intensity images of different angles of the same object according to a formula (1) and a formula (2)StokesVector images, combined with label data of the set of polarized light intensity images, collectively formStokesA data set.
Step 2, forStokesEach of the vector data setsStokesAnd the vector diagram is used for calculating a polarized image data set through a traditional polarized light fusion calculation mode, wherein each group of polarized images comprise a polarized angle image and a polarized degree image.
For the purpose ofStokesEach image in the vector dataset is traversed through each pixel according toStokesVector and polarization angleAOPAnd degree of polarizationDOPThe corresponding pixel is calculated by the calculation relation (3) and the calculation relation (4)AOPAndDOPvalues.
（3）
（4）
Obtained according to formulas (3) and (4)AOPAndDOPthe values generate new polarization angle images and polarization degree images, and then the polarization angle images and the polarization degree images and labels corresponding to each group of polarization images are combined to form a polarization image data set.
Step 3. According to step 1StokesCalculating quantization parameters by using the vector diagram and the polarized image obtained in the step 2, wherein the quantization parameters comprise slope parametersNumber of digitsa、cBias parameterb、d。
Wherein the slope parametera、cTo express the slope of the linear mapping function.
Bias parameterb、dTo represent the bias of the polarization mapping function.
Step 3.1. Traversing each group of polarization angle images and polarization degree images in the polarized image data set, and solving the numerical maximum value of all pixels in the whole polarized image data set for each pixelAOP max 、DOP max Minimum value of numerical valueAOP min 、DOP min 。
wherein ,AOP max represents the maximum value of the polarization angle,DOP max representing the maximum value of the degree of polarization.
AOP min Representing the minimum value of the polarization angle,DOP min representing the minimum value of the degree of polarization.
Step 3.2 findAOP max 、DOP max AndAOP min 、DOP min the pixel points corresponding to the four values are atStokesCorresponding pixel points in the vector image data set, and solving the calculation results of the four pixel points, which are respectively recorded as:
、/>、/>、/>。
step 3.3. Solve quantization parameters using the data in steps 3.1 and 3.2a、b、c、dQuantization parametera、b、c、dThe solving process of (2) is shown in the formulas (5) to (8), respectively.
（5）
（6）
（7）
（8）。
Step 4. The step 1 is carried outStokesAnd (3) respectively calculating a linear mapping polarization angle image and a linear mapping polarization degree image data set according to the linear mapping polarization resolving algorithm by combining the vector data set and the quantization parameter in the step (3).
Wherein each set of linear mapped polarization images includes a linear mapped polarization angle image and a linear mapped polarization degree image.
By aiming atStokesTraversing each pixel point of each image in the vector data set, and calculating the linear mapping polarization angle of the corresponding pixel according to the formulas (9) to (10)lqAOPValue and linear mapping polarization degreelqDOPValues.
（9）
（10）
wherein ,lqAOPrepresenting the intermediate solution of the polarization angle after linear mapping operation,lqAOPobtaining a polarization angle through nonlinear transformation;lqDOPan intermediate solution representing the degree of polarization after linear mapping operation,lqDOPthe degree of polarization is solved by nonlinear transformation.
And 5, training the polarized image recognition network model by using the polarized image data set to obtain a traditional polarized image recognition model.
In this embodiment, the polarization recognition network model adopts the Yolo v5 model.
And 6, combining the linear mapping polarized image data set obtained in the step 4 with the traditional polarized image recognition model in the step 5 to perform migration learning, and finally generating the linear mapping polarized image recognition model.
Specifically, the transfer learning is to input the linear mapping polarized image data set to the traditional polarized image recognition model obtained in the step 5 for training again, so that the traditional polarized image recognition model further learns to obtain the linear mapping polarized image recognition capability.
Through transfer learning, the traditional polarized image recognition model can quickly obtain the linear mapping polarized image recognition capability.
And 7, carrying out polarized image recognition on the image to be recognized acquired by the polarized light camera by utilizing the linear mapping polarized image recognition model obtained in the step 6.
Step 7.1. Firstly, the image to be identified is processed by the step 1StokesVector solution, then using step 4 pairStokesThe result after vector calculation is further subjected to linear mapping polarization solution;
and 7.2, carrying out corresponding image recognition on the linear mapping polarization image obtained by the linear mapping polarization solution in the step 4 by using the linear mapping polarization image recognition model obtained in the step 6, and obtaining an image recognition result.
According to the invention, through optimizing the network model training process, the nonlinear processing of the polarized image part is integrated into the neural network on the basis of not changing the network structure, so that the operation amount of the neural network identification is simplified, and the polarized light target identification efficiency is improved.
Example 2
This embodiment 2 describes a polarized light image processing system based on the same inventive concept as the polarized light image processing method described in embodiment 1 above. Specifically, the polarized light image processing system includes:
Stokesa vector resolving module for performing a processing on each group of polarized light intensity images in the polarized light intensity image training setStokesVector solution to obtain correspondingStokesA vector diagram;
will beStokesVector graphics in combination with corresponding tag data compositionStokesA vector dataset;
a traditional polarized light fusion calculation module for aiming atStokesEach of the vector data setsStokesThe vector diagram is used for calculating a polarized image dataset through a traditional polarized light fusion calculation mode;
wherein each set of polarized images in the polarized image dataset comprises a polarization angle image and a polarization degree image;
a quantization parameter calculation module for calculating a quantization parameter based on the obtained quantization parameterStokesCalculating quantization parameters from the vector image and the resulting polarized image, wherein the quantization parameters include slope parametersa、cBias parameterb、d；
A linear mapping polarization resolving module for resolving the obtainedStokesVector data set incorporating quantization parametersa、b、c、dAccording to the linear mapping polarization resolving algorithm, resolving to obtain a linear mapping polarization image dataset;
wherein each set of linear mapped polarization images in the linear mapped polarization image dataset comprises a linear mapped polarization angle image and a linear mapped polarization degree image;
the polarization recognition network module is used for carrying out polarization recognition network model training according to the obtained polarization image dataset to obtain a traditional polarization image recognition model;
the transfer learning module is used for carrying out transfer learning by combining the traditional polarized image recognition model according to the obtained linear mapping polarized image data set to obtain a linear mapping polarized image recognition model;
and the polarized image recognition module is used for carrying out polarized image recognition on the image to be recognized acquired by the polarized light camera by using the obtained linear mapping polarized image recognition model, so as to obtain a polarized image recognition result.
It should be noted that, in the polarized light image processing system, the implementation process of the functions and roles of each functional module is specifically detailed in the implementation process of the corresponding steps in the method in the above embodiment 1, which is not described herein again.
Example 3
Embodiment 3 describes a computer apparatus for implementing the polarized light image processing method described in embodiment 1 above.
In particular, the computer device includes a memory and one or more processors. Executable code is stored in the memory for implementing the steps of the polarized light image processing method described above when the executable code is executed by the processor.
In this embodiment, the computer device is any device or apparatus having data processing capability, which is not described herein.
Example 4
Embodiment 4 describes a computer-readable storage medium for implementing the polarized light image processing method described in embodiment 1 above.
Specifically, the computer-readable storage medium in this embodiment 4 has stored thereon a program for implementing the steps of the above-described polarized light image processing method when executed by a processor.
The computer readable storage medium may be an internal storage unit of any device or apparatus having data processing capability, such as a hard disk or a memory, or may be an external storage device of any device having data processing capability, such as a plug-in hard disk, a Smart Media Card (SMC), an SD Card, a Flash memory Card (Flash Card), or the like, which are provided on the device.
The foregoing description is, of course, merely illustrative of preferred embodiments of the present invention, and it should be understood that the present invention is not limited to the above-described embodiments, but is intended to cover all modifications, equivalents and alternatives falling within the spirit and scope of the present invention as defined by the appended claims.

Claims (9)

1. A method of processing a polarized light image, comprising the steps of:
step 1, firstly, carrying out polarized light intensity image training on each group of polarized light intensity images in a polarized light intensity image training setStokesVector solution to obtainStokesVector diagram, willStokesVector graphics in combination with corresponding tag data compositionStokesA vector dataset;
step 2, forStokesEach of the vector data setsStokesThe vector diagram is used for calculating a polarized image data set in a traditional polarized light fusion calculation mode, and each group of polarized images in the polarized image data set comprise a polarized angle image and a polarized degree image;
step 3. According to step 1StokesCalculating quantization parameters by the vector diagram and the polarized image obtained in the step 2; wherein the quantization parameter comprises a slope parametera、cBias parameterb、d；
Step 4. The step 1 is carried outStokesVector data set, combined with quantization parameters in step 3a、b、c、dAccording to the linear mapping polarization resolving algorithm, resolving to obtain a linear mapping polarization image dataset; each group of linear mapping polarization images in the linear mapping polarization image data set comprises a linear mapping polarization angle image and a linear mapping polarization degree image;
training the polarized image data set to obtain a traditional polarized image recognition model;
step 6, combining the linear mapping polarized image data set obtained in the step 4 with the traditional polarized image recognition model in the step 5 to perform migration learning so as to obtain a linear mapping polarized image recognition model;
and 7, carrying out polarized image recognition on the image to be recognized acquired by the polarized light camera by utilizing the linear mapping polarized image recognition model obtained in the step 6, and obtaining a polarized image recognition result.
2. The method for processing a polarized light image according to claim 1, wherein,
the step 1 specifically comprises the following steps:
polarized light intensity images of different polarized angles shot by a plurality of polarized light cameras are combined with label information of objects in each group of polarized light intensity images to form a polarized light intensity image training set;
according toStokesVector [I,Q,U,V]And the incident light and emergent light relation (1) formed by Mueller matrix, traversing polarized light intensity images of different angles of each group of same object in polarized light intensity image training set, and calculating four parameters of the Stokes vector of the incident light of the polarized light intensity images according to the polarized anglesI、Q、U、VAnd four parameters of the exit light Stokes vectorI´、Q´、U´、VAnd (2) to obtain the formula (2) by joint solution;
（1）
wherein ,S in representing incident lightStokesThe vector of the vector,S out representing the emitted lightStokesA vector;
M θ representing a specific angleθA Muller matrix corresponding to the polarizer;
Iindicating the total intensity of the incident light,Qfor polarization differences between the horizontal and vertical directions of the incident light,Ufor polarization differences between +45° and-45 ° directions of incident light,Vrepresenting the difference between the left-hand and right-hand circular polarization of the incident light;
Iand represents the total intensity of the outgoing light,Qis the polarization difference between the horizontal and vertical directions of the outgoing light,Uis the polarization difference between the +45° and-45 ° directions of the outgoing light,Vrepresents the difference between the left-hand and right-hand circular polarization of the exiting light;
according to the light intensity values corresponding to the pixel points at different polarization angles and the same position of the polarized light intensity images of different angles of the same objectI n And (2) reversely calculating the incident light corresponding to the pixel point at the positionStokesVector quantityS in ；
Incident lightStokesVector quantityS in The inverse solution expression of (2) is expressed by the formula (2) and the incident light parameterVCalculated as 0;
（2）
wherein ,I 1 ´、I 2 ´、I 3 respectively represents that the pixel point at the position is at the first polarization angleθ 1 Light intensity measured in the image at a second polarization angleθ 2 Light intensity measured in image, at third polarization angleθ 3 Light intensity measured in the image;
for each group of polarized light intensity images, calculating polarized light intensity images of different angles of the same object according to a formula (1) and a formula (2)StokesVector images, combined with label data of the set of polarized light intensity images, collectively formStokesA data set.
3. The method for processing a polarized light image according to claim 1, wherein,
the step 2 specifically comprises the following steps:
for the purpose ofStokesEach of the vector data setsStokesTraversing each pixel point according to the vector diagramStokesCalculating the corresponding pixels according to the calculated relational expression (3) and expression (4) of the vector and the polarization angle AOP and the polarization degree DOPAOPAndDOPa value;
（3）
（4）
wherein ,Iindicating the total intensity of the incident light,Qbetween the horizontal and vertical directions of the incident lightIs used for the polarization difference of (a),Ufor polarization differences between +45° and-45 ° directions of incident light,Vrepresenting the difference between the left-hand and right-hand circular polarization of the incident light;
obtained according to formulas (3) and (4)AOPAndDOPthe values generate new polarization angle images and polarization degree images, and then the polarization angle images and the polarization degree images and label data corresponding to each group of polarized light intensity images are combined to form a polarized image data set.
4. The method for processing a polarized light image according to claim 1, wherein,
the step 3 specifically comprises the following steps:
step 3.1. Traversing each group of polarization angle images and polarization degree images in the polarized image data set, and solving the numerical maximum value of all pixels in the whole polarized image data set for each pixelAOP max 、DOP max Minimum value of numerical valueAOP min 、DOP min ；
wherein ,AOP max represents the maximum value of the polarization angle,DOP max representing the maximum value of the polarization degree;
AOP min representing the minimum value of the polarization angle,DOP min representing a minimum value of the degree of polarization;
step 3.2 findAOP max 、DOP max AndAOP min 、DOP min the pixel points corresponding to the four values are atStokesCorresponding pixel points in the vector image data set, and solving the calculation results of the four pixel points, which are respectively recorded as:
、/>、/>、/>；
wherein ,Iindicating the total intensity of the incident light,Qfor polarization differences between the horizontal and vertical directions of the incident light,Ufor polarization differences between +45° and-45 ° directions of incident light,Vrepresenting the difference between the left-hand and right-hand circular polarization of the incident light;
step 3.3. Solve quantization parameters using the data in steps 3.1 and 3.2a、b、c、dQuantization parametera、b、c、dThe solving process of (2) is shown as formula (5) to formula (8) respectively;
（5）
（6）
（7）
（8）。
5. the method for processing a polarized light image according to claim 1, wherein,
the step 4 specifically comprises the following steps:
by aiming atStokesTraversing each pixel point of each image in the vector data set, and calculating the linear mapping polarization angle of the corresponding pixel according to the formulas (9) to (10)lqAOPValue and linear mapping polarization degreelqDOPA value;
（9）
（10）
wherein ,Iindicating the total intensity of the light and,Qfor the polarization difference between the horizontal and vertical directions,Uis the polarization difference between +45° and-45 ° directions,Vrepresenting the difference between left-hand and right-hand circular polarization;
lqAOPrepresenting the intermediate solution of the polarization angle after linear mapping operation,lqAOPobtaining a polarization angle through nonlinear transformation;lqDOPan intermediate solution representing the degree of polarization after linear mapping operation,lqDOPthe degree of polarization is solved by nonlinear transformation.
6. The method for processing a polarized light image according to claim 1, wherein,
the step 7 specifically comprises the following steps:
step 7.1. Firstly, the image to be identified is processed by the step 1StokesVector solution, then using step 4 pairStokesThe result after vector calculation is further subjected to linear mapping polarization solution;
and 7.2, carrying out corresponding image recognition on the linear mapping polarization image obtained by the linear mapping polarization solution in the step 4 by using the linear mapping polarization image recognition model obtained in the step 6, and obtaining an image recognition result.
7. A polarized light image processing system, comprising:
Stokesthe vector resolving module is used for carrying out the processing on each group of polarized light intensity images in the polarized light intensity image training setStokesVector solution to obtain correspondingStokesA vector diagram;
will beStokesVector graphics in combination with corresponding tag data compositionStokesVector dataA collection;
a traditional polarized light fusion calculation module for aiming atStokesEach of the vector data setsStokesThe vector diagram is used for calculating a polarized image dataset through a traditional polarized light fusion calculation mode;
wherein each set of polarized images in the polarized image dataset comprises a polarization angle image and a polarization degree image;
a quantization parameter calculation module for calculating a quantization parameter based on the obtained quantization parameterStokesCalculating quantization parameters from the vector image and the resulting polarized image, wherein the quantization parameters include slope parametersa、cBias parameterb、d；
A linear mapping polarization resolving module for resolving the obtainedStokesVector data set incorporating quantization parametersa、b、c、dAccording to the linear mapping polarization resolving algorithm, resolving to obtain a linear mapping polarization image dataset;
wherein each set of linear mapped polarization images in the linear mapped polarization image dataset comprises a linear mapped polarization angle image and a linear mapped polarization degree image;
the polarization recognition network module is used for carrying out polarization recognition network model training according to the obtained polarization image dataset to obtain a traditional polarization image recognition model;
the transfer learning module is used for carrying out transfer learning by combining the traditional polarized image recognition model according to the obtained linear mapping polarized image data set to obtain a linear mapping polarized image recognition model;
and the polarized image recognition module is used for carrying out polarized image recognition on the image to be recognized acquired by the polarized light camera by using the obtained linear mapping polarized image recognition model, so as to obtain a polarized image recognition result.
8. A computer device comprising a memory and one or more processors, the memory having executable code stored therein, wherein the processor, when executing the executable code,
a step of realizing the polarized light image processing method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a program stored thereon, which when executed by a processor, implements the steps of the polarized light image processing method according to any one of claims 1 to 6.
CN202310896103.XA 2023-07-21 2023-07-21 Polarized light image processing method, polarized light image processing system, computer device, and storage medium Active CN116630750B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310896103.XA CN116630750B (en) 2023-07-21 2023-07-21 Polarized light image processing method, polarized light image processing system, computer device, and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310896103.XA CN116630750B (en) 2023-07-21 2023-07-21 Polarized light image processing method, polarized light image processing system, computer device, and storage medium

Publications (2)

Publication Number Publication Date
CN116630750A true CN116630750A (en) 2023-08-22
CN116630750B CN116630750B (en) 2023-09-26

Family

ID=87638514

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310896103.XA Active CN116630750B (en) 2023-07-21 2023-07-21 Polarized light image processing method, polarized light image processing system, computer device, and storage medium

Country Status (1)

Country Link
CN (1) CN116630750B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060215879A1 (en) * 2005-02-02 2006-09-28 Whitaker Sheila G Real-time image detection using polarization data
US20130308132A1 (en) * 2012-05-17 2013-11-21 The University Of Akron System and method for polarimetric wavelet fractal detection and imaging
CN115393233A (en) * 2022-07-25 2022-11-25 西北农林科技大学 Full-linear polarization image fusion method based on self-encoder
CN115468654A (en) * 2022-09-05 2022-12-13 长春理工大学 Method for acquiring polarization difference characteristic image under optimal angle

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060215879A1 (en) * 2005-02-02 2006-09-28 Whitaker Sheila G Real-time image detection using polarization data
US20130308132A1 (en) * 2012-05-17 2013-11-21 The University Of Akron System and method for polarimetric wavelet fractal detection and imaging
CN115393233A (en) * 2022-07-25 2022-11-25 西北农林科技大学 Full-linear polarization image fusion method based on self-encoder
CN115468654A (en) * 2022-09-05 2022-12-13 长春理工大学 Method for acquiring polarization difference characteristic image under optimal angle

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
王新;王学勤;孙金祚;: "基于偏振成像和图像融合的目标识别技术", 激光与红外, no. 07 *

Also Published As

Publication number Publication date
CN116630750B (en) 2023-09-26

Similar Documents

Publication Publication Date Title
Yang et al. A survey of deep learning-based source image forensics
WO2023082784A1 (en) Person re-identification method and apparatus based on local feature attention
Bakas et al. A digital forensic technique for inter–frame video forgery detection based on 3D CNN
Zhang et al. Visual tracking via constrained incremental non-negative matrix factorization
CN110852311A (en) Three-dimensional human hand key point positioning method and device
US11501470B2 (en) Geometric encoding of data
CN112435223B (en) Target detection method, device and storage medium
CN112560753A (en) Face recognition method, device and equipment based on feature fusion and storage medium
Novozámský et al. Detection of copy-move image modification using JPEG compression model
Jia et al. EMBDN: An efficient multiclass barcode detection network for complicated environments
Sarmah et al. Optimization models in steganography using metaheuristics
Xiao et al. Robust license plate detection and recognition with automatic rectification
CN110163095B (en) Loop detection method, loop detection device and terminal equipment
CN113191189A (en) Face living body detection method, terminal device and computer readable storage medium
CN116630750B (en) Polarized light image processing method, polarized light image processing system, computer device, and storage medium
US20110044497A1 (en) System, method and program product for camera-based object analysis
Sheng et al. Detection of content-aware image resizing based on Benford’s law
Otta et al. User identification with face recognition: A systematic analysis
CN116503761A (en) High-voltage line foreign matter detection method, model training method and device
CN115410257A (en) Image protection method and related equipment
Chen et al. 360-degree gaze estimation in the wild using multiple zoom scales
Wu et al. Research of quickly identifying markers on augmented reality
CN115546192B (en) Livestock quantity identification method, device, equipment and storage medium
Zhu et al. Recaptured image detection based on convolutional neural networks with local binary patterns coding
CN111079704A (en) Face recognition method and device based on quantum computation

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