Method and system for rotating a gaussian distribution of infrared point targets

Abstract

The application provides a rotating system and method for Gaussian distribution of an infrared point target, comprising: a target rotating and neighborhood pixel position generating module, which rotates each pixel of the target according to a preset shooting angle to determine the position relationship of four neighborhood pixels of the rotated pixel; a pixel radiation energy Gaussian distribution module, which determines a Gaussian distribution model based on distance change; a pixel distribution module based on energy and barycenter constraint, which determines an energy and barycenter constraint model; and an iterative calculation pixel distribution module, which uses Newton iteration to solve the radiation energy of the four neighborhood pixels based on the Gaussian distribution model and the energy and barycenter constraint model, thereby obtaining the position and radiation energy distribution of each pixel of the rotated infrared target. The application uses the four neighborhood pixels of the rotated pixel as the benchmark on the basis of the conventional rotation transformation, and uses the Gaussian distribution model based on distance change for the pixel radiation energy distribution, thereby ensuring that the radiation energy of the infrared target does not change before and after rotation and that the relative barycenter of each pixel does not change.

Description

Technical Field

[0001] This invention relates to the technical field of target image simulation and transformation, specifically to a method and system for rotating the Gaussian distribution of infrared point targets, and more specifically to a method and system for rotating the Gaussian distribution of infrared point targets based on unchanged energy and relative center of gravity. Background Technology

[0002] Target image rotation simulation is a technique that simulates the pixel distribution of the same target under different shooting directions to verify target detection and tracking performance. In particular, simulations based on constant energy and relative centroid can more realistically reflect the pixel distribution under different shooting directions, thus further validating target detection and tracking performance.

[0003] Patent document CN116843560A (application number: 202310545277.1) discloses an infrared image rotational motion blur recovery method based on a thermal diffusion model. This method uses a thermal balance equation to construct a thermodynamic system consisting of the detector pixel and its surrounding environment to characterize the generation process of the blurred image, thereby achieving infrared image rotational motion blur recovery. This method is an image rotational blur recovery method based on a thermal diffusion model.

[0004] Patent document CN109034153A (application number: 201810803980.7) discloses a method and apparatus for image fidelity preservation applied to image rotation. The method determines a fidelity sample image by using the minimum bounding rectangle of the key information region after image rotation, thereby expanding the number of training samples. This method is a way to obtain fidelity images of key information regions.

[0005] Patent document CN108492243A (application number: 201810333559.4) discloses an image rotation device, system and method based on block processing. By performing coordinate rotation mapping on the image according to the image block division rules, and outputting the blocks to a predetermined cache in a preset order, the efficiency of DDR memory access during image rotation processing is achieved. This method is a way to improve the processing speed of image rotation.

[0006] Patent document CN109448038A (application number: 201811310207.3) discloses a feature extraction method for seabed sediment sonar images based on DRLBP and random forest. This method improves upon traditional LBP by using DRLBP to extract features from images at different scales, addressing issues such as image rotation variations, high computational cost, and feature redundancy. This method is an image rotation variation-based approach using image feature extraction.

[0007] Patent document CN113421248A (application number: 202110742772.2) discloses a numerical processing method for rotating images of substation equipment. This method calculates the deflection arc value of the target real-time image by calculating and comparing the extreme point vector moments of a reference image and a real-time image. The method obtains the deflection of the rotating image by comparing the real-time image with the reference image.

[0008] Traditional image rotation methods can cause changes in the pixel distribution and relative centroid of the pixels before and after rotation, resulting in image distortion. Therefore, ensuring that the pixel distribution of the target image remains unchanged before and after rotation and that the image is not distorted has become a problem that needs to be solved. Summary of the Invention

[0009] In view of the deficiencies in the prior art, the purpose of this invention is to provide a rotation system and method for Gaussian distribution of infrared point targets.

[0010] A rotating system for an infrared point target with a Gaussian distribution, provided by the present invention, comprises:

[0011] Target rotation and neighboring cell position generation module: Rotates each cell of the target according to a preset azimuth angle and determines the positional relationship of the four neighboring cells of the rotated cell;

[0012] Pixel Radiant Energy Gaussian Distribution Module: Determines a Gaussian distribution model based on distance variation;

[0013] Pixel distribution module based on energy and centroid constraints: Determine the energy and centroid constraint model;

[0014] Iterative pixel distribution calculation module: Based on the Gaussian distribution model, energy and centroid constraint model, Newton iteration is used to solve the radiation energy of four neighboring pixels, thereby obtaining the position of each pixel of the infrared target and the distribution of radiation energy after rotation.

[0015] Preferably, in the target rotation and neighboring cell position generation module,

[0016] Assume the target's trajectory angle before rotation is θ1, the target's trajectory angle after rotation is θ2, and the target's centroid coordinates before rotation are [X]. c ,Y c ]; taking the coordinates of a certain position before rotation as [r k_pre ,c k_pre The radiant energy is I. k Taking a pixel as an example, after the rotation matrix transformation, the radiant energy of the rotated pixel is still I. k After rotation, the row and column coordinates of the pixel [r] k_aft ,c k_aft The following is represented:

[0017] r k _aft =c k _ pre -Y c )*sin(θ2-θ1)+r k _ pre -X c )*cos(θ2-θ1)

[0018] c k _ aft =c k _-Y c )*cos(θ2-θ1)-r k _ pre -X c sin(θ2-θ1)

[0019] Then the four neighboring pixels of the rotated pixel are N k1 N k2 N k3 N k4 This indicates that its corresponding row and column coordinates are [r k_adj ,c k_ ], r k_adj ,c k_adj +1],[r k_adj +1,c k_ ], [r k_adj +1,c k_adj +1], the corresponding radiation energy is I k1 I k2 I k3 I k4 ;

[0020] in:

[0021] r k _ adj =fix(r k _ aft ),c k _ adj =fix(c k _)

[0022] Where: fix means rounding down.

[0023] Preferably, in the Gaussian distribution module for pixel radiation energy, the radiation energy of the four neighboring pixels is allocated using the distance between the four neighboring pixels and the centroid coordinates as weighting coefficients. At the same time, the Gaussian distribution model of the energy of the four neighboring pixels is required to satisfy a normal distribution with the centroid coordinates as the origin and the distance as the variable.

[0024] Preferably, in the Gaussian distribution module of pixel radiant energy, the distribution of pixel radiant energy follows a mathematical expectation μ and a variance σ. 2Normal distribution:

[0025]

[0026] Where: p is the radiant energy coefficient of the neighboring pixel, I k Let be the radiant energy of the rotated pixel, e be the exponential function, σ be the variance, and α be the weighting coefficients corresponding to the distances from the rotated pixel to its four neighboring pixels. The calculation is as follows:

[0027]

[0028] Where: r i Let i be the distance from the centroid coordinates of the rotated pixel to the coordinates of the neighboring pixel, i = 1:4.

[0029] Preferably, in the pixel distribution module based on energy and centroid constraints,

[0030] The sum of the radiated energy of the four neighboring regions is equivalent to the radiated energy of the rotating pixel;

[0031] The centroids of the four neighboring pixels are the coordinates of the rotated pixels. The centroids of the four pixels are calculated by using the radiation energy of each pixel as a weighting coefficient, thereby constraining the coordinates and radiation energy distribution of each pixel after rotation.

[0032] Preferably, in the pixel distribution module based on energy and centroid constraints,

[0033] I k =I k1 +I k2 +I k3 +I k4

[0034]

[0035] Preferably, in the iterative method module for solving the pixel distribution,

[0036]

[0037] A method for rotating an infrared point target with a Gaussian distribution according to the present invention includes:

[0038] Step S1: Rotate each pixel of the target according to the preset azimuth angle to determine the positional relationship of the four neighboring pixels of the rotated pixel;

[0039] Step S2: Determine the Gaussian distribution model based on distance variation;

[0040] Step S3: Determine the energy and center of gravity constraint model;

[0041] Step S4: Based on the Gaussian distribution model, energy and centroid constraint model, Newton iteration is used to solve the radiation energy of the four neighboring pixels, so as to obtain the position of each pixel of the infrared target and the distribution of radiation energy after rotation.

[0042] Preferably, step S1 employs:

[0043] Assume the target's trajectory angle before rotation is θ1, the target's trajectory angle after rotation is θ2, and the target's centroid coordinates before rotation are [X]. c ,Y c ]; taking the coordinates of a certain position before rotation as [r k_pre ,c k_pre The radiant energy is I. k Taking a pixel as an example, after the rotation matrix transformation, the radiant energy of the rotated pixel is still I. k After rotation, the row and column coordinates of the pixel [r] k_aft ,c k_aft The following is represented:

[0044] r k _ aft =c k _ pre -Y c )*sin(θ2-θ1)+r k _ pre -X c )*cos(θ2-θ1)

[0045] c k _ aft =c k _-Y c )*cos(θ2-θ1)-r k _ pre -X c sin(θ2-θ1)

[0046] Then the four neighboring pixels of the rotated pixel are N k1 N k2 N k3 N k4 This indicates that its corresponding row and column coordinates are [r k_adj ,c k_ ], r k_adj ,c k_adj +1],[r k_adj +1,c k_ ], [r k_adj +1,c k_adj +1], the corresponding radiation energy is I k1 I k2 I k3 I k4 ;

[0047] in:

[0048] r k _ adj =fix(r k _ aft ),c k _ adj =fix(c k _)

[0049] Where: fix means rounding down.

[0050] Preferably, step S2 involves: allocating the radiation energy of the four neighboring pixels using the distance between the four neighboring pixels and the centroid coordinates as weighting coefficients, while requiring the Gaussian distribution model of the energy of the four neighboring pixels to satisfy a normal distribution with the centroid coordinates as the origin and the distance as the variable;

[0051] Step S2 includes: the pixel radiant energy distribution follows a mathematical expectation μ and a variance σ. 2 Normal distribution:

[0052]

[0053] Where: p is the radiant energy coefficient of the neighboring pixel, I k Let be the radiant energy of the rotated pixel, e be the exponential function, σ be the variance, and α be the weighting coefficients corresponding to the distances from the rotated pixel to its four neighboring pixels. The calculation is as follows:

[0054]

[0055] Where: r i The distance from the centroid coordinates of the rotated pixel to the coordinates of the neighboring pixel is 1:4;

[0056] Step S3 includes:

[0057] The sum of the radiated energy of the four neighboring regions is equivalent to the radiated energy of the rotating pixel;

[0058] The centroid of the four neighboring pixels is the coordinate of the rotated pixel. The centroid of the four pixels is calculated by using the radiation energy of each pixel as a weighting coefficient, thereby constraining the coordinates and radiation energy distribution of each pixel after rotation.

[0059] Step S3 includes:

[0060] I k =I k1 +I k2 +I k3 +I k4

[0061]

[0062] Step S4 includes:

[0063]

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] 1. The present invention discloses a rotation method for infrared point targets based on Gaussian distribution with unchanged energy and relative center of gravity, which can be used to improve the pixel simulation distortion caused by conventional image rotation, so as to verify the target detection performance;

[0066] 2. The simulation method of the present invention, by using the Gaussian distribution technology feature based on distance variation, ensures that the target centroid and energy remain unchanged before and after image rotation, avoids pixel blurring after rotation, reduces false detection caused by image simulation distortion, and avoids affecting the correctness of verification detection. Attached Figure Description

[0067] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0068] Figure 1 This is a schematic diagram of the image rotation process based on constant energy and center of gravity.

[0069] Figure 2 This is a schematic diagram of the coordinates and energy distribution of neighboring pixels.

[0070] Figure 3 This is a schematic diagram of the target pixel and energy distribution calculation after rotation. Detailed Implementation

[0071] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0072] Example 1

[0073] This invention provides a rotating system based on a Gaussian distribution of an infrared point target with invariant energy and relative center of gravity, such as... Figures 1 to 3 As shown, it includes: a target rotation and neighborhood cell position generation module, a cell radiation energy Gaussian distribution module, a cell distribution module based on energy and centroid constraints, and an iterative solution module for cell distribution.

[0074] Based on traditional rotation transformation, this invention uses the four neighboring pixels of the rotated pixel as a reference. The pixel radiation energy distribution adopts a Gaussian distribution model based on distance variation, which ensures that the radiation energy of the infrared target remains unchanged before and after rotation and that the relative center of gravity of each pixel remains unchanged. This improves the defect of uneven pixel distribution caused by the rotation transformation of multiple pixel points superimposed on the same point.

[0075] The target rotation and neighborhood cell position generation module includes two sub-modules: a target rotation transformation sub-module and a neighborhood cell coordinate interpolation sub-module;

[0076] The target rotation transformation submodule rotates each pixel of the target according to a specific azimuth angle based on a traditional rotation transformation matrix.

[0077] The neighborhood cell coordinate interpolation submodule is obtained by interpolating the coordinates of the four neighboring cells based on the rotated cell coordinates.

[0078] The Gaussian distribution module for pixel radiant energy employs a distance-based Gaussian distribution model. The radiant energy of the four neighboring pixels is allocated using the distance between the four neighboring pixels and the centroid coordinates as weighting coefficients. Simultaneously, the Gaussian distribution model of the energy of the four neighboring pixels is required to satisfy a normal distribution with the centroid coordinates as the origin and distance as the variable.

[0079] The pixel radiant energy distribution follows the mathematical expectation μ and variance σ. 2 The normal distribution is N(μ,σ). 2 This invention considers that it follows a standard normal distribution of N(0,1).

[0080] The pixel distribution module based on energy and centroid constraints requires that the sum of the radiation energy of the four neighboring pixels be equivalent to the radiation energy of the rotated pixel, and the centroid of the four neighboring pixels be the coordinates of the rotated pixel. The centroid of the four pixels is calculated by using the radiation energy of each pixel as a weighting coefficient to calculate the pixel centroid, thereby constraining the coordinates and radiation energy distribution of each pixel after rotation.

[0081] The iterative method for calculating pixel distribution uses Newton's iteration to solve for the radiation energy of four neighboring pixels, thereby obtaining the pixel distribution after rotation.

[0082] In this embodiment, as Figures 1-3 As shown, the specific implementation steps are as follows:

[0083] The target rotation and neighboring cell position generation module: determines the positional relationship of the four neighboring cells of the rotated cell;

[0084] Assume the target's trajectory angle before rotation is θ1, the target's trajectory angle after rotation is θ2, and the target's centroid coordinates before rotation are [X]. c ,Y c]; taking the coordinates of a certain position before rotation as [r k_pre ,c k_pre The radiant energy is I. k Taking a pixel as an example, after the traditional rotation matrix transformation, the radiant energy of the rotated pixel is still I. k After rotation, the row and column coordinates of the pixel [r] k_aft ,c k_aft The following is represented:

[0085] r k _ aft =c k _ pre -Y c )*sin(θ2-θ1)+r k _ pre -X c )*cos(θ2-θ1)

[0086] c k _ aft =c k _-Y c )*cos(θ2-θ1)-r k _ pre -X c sin(θ2-θ1)

[0087] Then the four neighboring pixels of the rotated pixel are N k1 N k2 N k3 N k4 This indicates that its corresponding row and column coordinates are [r k_adj ,c k_ ], r k_adj ,c k_adj +1],[r k_adj +1,c k_ ], [r k_adj +1,c k_adj +1], the corresponding radiation energy is I k1 I k2 I k3 I k4 ;

[0088] in:

[0089] r k _ adj =fix(r k _ aft ),c k _ adj =fix(c k _)

[0090] Where: fix means rounding down.

[0091] The Gaussian distribution module for pixel radiant energy determines the pixel radiant energy distribution model. The pixel radiant energy distribution follows a mathematical expectation μ and a variance σ. 2 Normal distribution:

[0092]

[0093] Where: p is the radiant energy coefficient of the neighboring pixel, I k Let be the radiant energy of the rotated pixel, e be an exponential function, σ be the variance (here we take σ = 1), and α be the weighting coefficients corresponding to the distances from the rotated pixel to its four neighboring pixels, calculated as follows:

[0094]

[0095] Where: r i Let i be the distance from the centroid coordinates of the rotated pixel to the coordinates of the neighboring pixel, i = 1:4.

[0096] The cell distribution module based on energy and centroid constraints determines the energy and centroid constraint model. It requires the radiant energy of the four neighboring pixels and the radiant energy equivalent to the rotated pixel, where the centroid of the four neighboring pixels is the coordinate of the rotated pixel.

[0097] I k =I k1 +I k2 +I k3 +I k4

[0098]

[0099] The iterative pixel distribution calculation module iteratively calculates the energy and centroid constraint model parameters. Based on the Gaussian distribution model of pixel radiation energy and the constraints of constant energy and constant relative centroid, it iteratively solves for the radiation energy of four neighboring pixels, thereby obtaining the position of each pixel of the rotated infrared target and the radiation energy distribution.

[0100]

[0101] The present invention also provides a rotation system based on an infrared point target Gaussian distribution with unchanged energy and relative center of gravity. The rotation system based on an infrared point target Gaussian distribution with unchanged energy and relative center of gravity can be implemented by executing the process steps of the rotation method based on an infrared point target Gaussian distribution with unchanged energy and relative center of gravity. That is, those skilled in the art can understand the rotation method based on an infrared point target Gaussian distribution with unchanged energy and relative center of gravity as a preferred embodiment of the rotation system based on an infrared point target Gaussian distribution with unchanged energy and relative center of gravity.

[0102] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0103] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A rotating system for a Gaussian distribution of infrared point targets, characterized in that, include: Target rotation and neighboring cell position generation module: Rotates each image cell of the target according to a preset azimuth angle, and determines the positional relationship of the four neighboring image cells of the rotated image cell; Pixel Radiant Energy Gaussian Distribution Module: Determines a Gaussian distribution model based on distance variation; Pixel distribution module based on energy and centroid constraints: Determine the energy and centroid constraint model; Iterative pixel distribution calculation module: Based on the Gaussian distribution model, energy and centroid constraint model, Newton iteration is used to solve the radiation energy of four neighboring image pixels, so as to obtain the position of each image pixel and the radiation energy distribution of the infrared target after rotation; In the Gaussian distribution module for pixel radiation energy, the distance between four neighboring image pixels and the centroid coordinates is used as a weighting coefficient to allocate the radiation energy of four neighboring image pixels. At the same time, the Gaussian distribution model of the energy of the four neighboring image pixels is required to satisfy a normal distribution with the centroid coordinates as the origin and the distance as the variable. In the Gaussian distribution module for pixel radiant energy, the distribution of image pixel radiant energy follows a mathematical expectation. ,variance Normal distribution: in: The radiant energy coefficient of a neighboring image pixel. The radiant energy of the rotated image pixels. It is an exponential function. For variance, The weighting coefficients corresponding to the distances from the rotated image pixel to its four neighboring image pixels are calculated as follows: in: Let i be the distance from the centroid coordinates of the rotated image pixel to the position coordinates of the neighboring image pixel, i = 1:

4.

2. The rotating system for Gaussian distribution of infrared point targets according to claim 1, characterized in that, In the target rotation and neighboring cell position generation module, Assuming the target's firing angle before rotation After rotation, the target's firing angle The coordinates of the target's center of gravity before rotation [ , ]; Let the coordinates of a certain position before rotation be [ , ], rotated cell row and column coordinates[ , The following is represented: Then the four neighboring pixels of the rotated pixel are... , , , This indicates that its corresponding row and column coordinates are [ , ], , ], [ , ], [ , The corresponding radiant energy is , , , ; in: Where: fix means rounding down.

3. The rotating system for Gaussian distribution of infrared point targets according to claim 1, characterized in that, In the aforementioned pixel distribution module based on energy and centroid constraints, The sum of the radiated energy of the four neighboring regions is equivalent to the radiated energy of the rotating pixel; The centroids of the four neighboring pixels are the coordinates of the rotated pixels. The centroids of the four pixels are calculated by using the radiation energy of each pixel as a weighting coefficient, thereby constraining the coordinates and radiation energy distribution of each pixel after rotation.

4. The rotating system for Gaussian distribution of infrared point targets according to claim 3, characterized in that, In the aforementioned pixel distribution module based on energy and centroid constraints, 。 5. The rotating system for Gaussian distribution of infrared point targets according to claim 1, characterized in that, In the iterative method module for solving cell distribution, 。 6. A method for rotating a Gaussian distribution of an infrared point target, characterized in that, include: Step S1: Rotate each image pixel of the target according to a preset azimuth angle, and determine the positional relationship of the four neighboring image pixels of the rotated image pixel; Step S2: Determine the Gaussian distribution model based on distance variation; Step S3: Determine the energy and center of gravity constraint model; Step S4: Based on the Gaussian distribution model, energy and centroid constraint model, Newton iteration is used to solve the radiation energy of four neighboring image pixels, thereby obtaining the position of each image pixel and the radiation energy distribution of the infrared target after rotation; In step S2, the following steps are adopted: the distance between the four neighboring image pixels and the centroid coordinates is used as the weighting coefficient to allocate the radiation energy of the four neighboring image pixels. At the same time, the Gaussian distribution model of the energy of the four neighboring image pixels is required to satisfy a normal distribution with the centroid coordinates as the origin and the distance as the variable. Step S2 includes: the distribution of radiant energy of image pixels conforms to a mathematical expectation. ,variance Normal distribution: in: The radiant energy coefficient of a neighboring image pixel. The radiant energy of the rotated image pixels. It is an exponential function. For variance, The weighting coefficients corresponding to the distances from the rotated image pixel to its four neighboring image pixels are calculated as follows: in: Let i be the distance from the centroid coordinates of the rotated image pixel to the position coordinates of the neighboring image pixel, i = 1:

4.

7. The rotation method for Gaussian distribution of infrared point targets according to claim 6, characterized in that, Step S1 adopts the following: Assuming the target's firing angle before rotation After rotation, the target's firing angle The coordinates of the target's center of gravity before rotation [ , ]; Let the coordinates of a certain position before rotation be [ , ], rotated cell row and column coordinates[ , The following is represented: Then the four neighboring pixels of the rotated pixel are... , , , This indicates that its corresponding row and column coordinates are [ , ], , ], [ , ], [ , The corresponding radiant energy is , , , ; in: Where: fix means rounding down.

8. The method for rotating the Gaussian distribution of an infrared point target according to claim 6, characterized in that, Step S3 includes: The sum of the radiated energy of the four neighboring regions is equivalent to the radiated energy of the rotating pixel; The centroid of the four neighboring pixels is the coordinate of the rotated pixel. The centroid of the four pixels is calculated by using the radiation energy of each pixel as a weighting coefficient, thereby constraining the coordinates and radiation energy distribution of each pixel after rotation. Step S3 includes: ； Step S4 includes: 。

Images