Infrared image multi-point correction method, device, equipment and medium

Abstract

The invention provides an infrared image multi-point correction method, device, equipment and medium, and the method comprises the steps: collecting more than or equal to 10 frames of infrared images in four temperature scenes of dark field, low illumination, medium illumination and high illumination in real time, carrying out the multi-frame average processing of each frame of image to suppress the time domain noise, and generating a to-be-analyzed image; pixel feature parameter extraction is carried out on the to-be-analyzed image, and blind pixel detection parameters based on the response rate, the noise voltage, the noise equivalent temperature difference and the linearity feature of each pixel are generated; marking blind pixel pixels and normal pixel pixels of the to-be-analyzed image according to a constructed blind pixel table based on blind pixel detection parameters; and for blind pixels, carrying out average interpolation filling by adopting adaptive gain and offset parameters of normal pixels in a 3 * 3 neighborhood to realize correction. According to the method, the multi-frame data, multi-criterion blind pixel detection and multi-point temperature correction methods are combined, wide-temperature-range and high-precision infrared image correction is achieved, and noise suppression and pixel nonlinear compensation are considered at the same time.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method, apparatus, device, and medium for multi-point correction of infrared images. Background Technology

[0002] Due to manufacturing limitations, infrared detectors suffer from problems such as non-uniform pixel response, dead pixels (dead pixels, overheated pixels), and noise interference, which lead to a decline in image quality.

[0003] Existing technologies employ methods such as two-point correction, single-point correction, blind pixel detection, and multi-frame averaging for multi-point image correction. However, the following technical problems exist: (1) Poor adaptability: Two-point correction only covers a limited number of temperature points and cannot cope with response drift in wide temperature range scenarios; (2) Low accuracy of blind pixel detection: A single criterion (such as response rate alone) is insufficient to distinguish between real blind pixels and noise interference, resulting in defects in the corrected image; (3) Insufficient noise suppression: Simple multi-frame averaging has limited effect on suppressing temporal noise, especially in low-light scenes; (4) Pixel nonlinearity is not considered: Traditional methods assume that the pixel response is linear, but the actual nonlinear response leads to correction deviation and affects the target detection accuracy. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a method, apparatus, device, and medium for multi-point correction of infrared images. This invention combines multi-frame data, multi-criteria blind pixel detection, and multi-point temperature correction to achieve wide-temperature-range, high-precision infrared image correction, while also taking into account noise suppression and pixel nonlinearity compensation.

[0005] In a first aspect, the present invention proposes a method for multi-point correction of infrared images, specifically including the following steps: Step S1. Real-time acquisition of infrared images of 10 or more frames under four temperature scenarios: dark field, low illumination, medium illumination, and high illumination. Multi-frame averaging is performed on each frame to suppress temporal noise and generate the image to be analyzed. Step S2. Extract pixel feature parameters from the image to be analyzed to generate blind pixel detection parameters based on the response rate, noise voltage, noise equivalent temperature difference, and linearity features of each pixel; Step S3. Mark the blind pixels and normal pixels of the image to be analyzed according to the constructed blind pixel table based on blind pixel detection parameters; Step S4. For blind pixels, average interpolation is performed using the adaptive gain and offset parameters of normal pixels in the 3×3 neighborhood to achieve correction.

[0006] Furthermore, the process of averaging multiple frames for each image is implemented using the following formula: Where F is the number of frames. For the f-th frame pixel The signal voltage.

[0007] Furthermore, the calculation formulas for the response rate, noise voltage, noise equivalent temperature difference, and linearity characteristics generated in step S2 are respectively expressed as follows: In the above formula, Representing pixels at different temperatures Response rate , Pixels at temperature T in a high-illuminance scene and at temperature T0 in a low-illuminance scene, respectively. Signal voltage; Represents pixels in dark fields noise voltage, Represents pixels in dark fields The response voltage, This represents the pixel in the f-th frame at temperature T0 in a low-light scene. The signal voltage; Indicates the noise equivalent temperature difference; Nonlinear deviation representing linearity characteristics Represents a cell The original signal voltage at temperature T under high illumination conditions. , Represents a cell The coefficients for fitting the response curve.

[0008] Furthermore, the step of constructing a blind pixel table based on blind pixel detection parameters specifically includes fusing spatial response criteria, response rate criteria, temporal noise criteria, linearity criteria, and G-parameter criteria, and generating a blind pixel table through a logical OR operation. When blind pixel labeling of the image to be analyzed is performed using the blind pixel table, if any performance parameter of the spatial response criteria, response rate criteria, temporal noise criteria, linearity criteria, or G-parameter criteria is abnormal, the detection result of the blind pixel is output.

[0009] Furthermore, the spatial response criterion specifically generates a first threshold interval based on the global mean and global standard deviation of the pixel response values ​​under different temperature scenarios such as dark field, medium illumination, and high illumination. If the pixel response value is greater than the maximum value of the first threshold interval or less than the minimum value of the first threshold interval, the detection result of the blind pixel is output. The response rate criterion is specifically to generate a second threshold interval based on the mean and standard deviation of the pixel response value difference under different temperature scenarios of high illumination and dark field. If the pixel response value difference is greater than the maximum value of the second threshold interval or less than the minimum value of the second threshold interval, the detection result of the blind cell is output. The temporal noise criterion is specifically generated by generating a third threshold interval based on the mean and standard deviation of the noise standard deviation under different temperature scenarios such as dark field, medium illumination and high illumination. If the temporal noise value is greater than the maximum value of the third threshold interval or less than the minimum value of the third threshold interval, the detection result of the blind cell is output. The linearity criterion is specifically to generate a nonlinear deviation value based on the slope and intercept obtained by fitting the pixel response values ​​under different temperature scenarios of high illumination and dark field. If the nonlinear deviation value is greater than the fourth threshold, the detection result of the blind cell is output. The G-parameter criterion is specifically that if the correction gain parameter is greater than the maximum value of the fifth threshold interval or less than the minimum value of the fifth threshold interval, then the detection result of the blind cell is output.

[0010] Furthermore, for blind pixels, the process of average interpolation filling using adaptive gain and offset parameters of normal pixels in a 3×3 neighborhood to achieve correction includes the following formula: In the above formula, Represents a blind pixel At temperature The corrected signal voltage; N represents the number of normal pixels in the 3×3 neighborhood. Represents a blind pixel The coordinates of normal pixels within a 3×3 neighborhood. Represents the normal pixels in the neighborhood The corrected signal voltage at temperature T.

[0011] Furthermore, the method also includes correcting normal pixels using a correction response curve based on nonlinear deviation pixel fitting, expressed as: in, Represents a cell Adaptive gain parameters Represents a cell The adaptive offset parameters, Represents a cell At temperature The corrected signal voltage; Represents a cell At temperature The original signal voltage.

[0012] In a second aspect, the present invention also provides an apparatus for infrared image multi-point correction to perform the method as described in the first aspect, the apparatus comprising: The image acquisition module is used to acquire infrared images of 10 or more frames in real time under four temperature scenarios: dark field, low illumination, medium illumination, and high illumination. Each frame is averaged to suppress temporal noise and generate an image to be analyzed. The feature extraction module is used to extract pixel feature parameters from the image to be analyzed and generate blind pixel detection parameters based on the response rate, noise voltage, noise equivalent temperature difference and linearity features of each pixel. The blind pixel labeling module is used to label the blind pixels and normal pixels of the image to be analyzed according to the constructed blind pixel table based on blind pixel detection parameters. The blind pixel correction module is used to fill blind pixels with average interpolation parameters of normal pixels in a 3×3 neighborhood to achieve correction.

[0013] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in the first aspect.

[0014] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect.

[0015] The beneficial effects of this invention are: (1) Wide temperature range adaptability: Covering multiple temperatures from dark field to high illumination, the non-uniformity of the corrected image in the range of -10℃ to 50℃ is reduced to less than 5%; (2) Accurate blind element detection: This invention improves the accuracy of blind element recognition to over 98% and reduces the false positive rate by 20% by fusing five criteria; (3) Significant noise suppression: Multi-frame averaging combined with pixel noise characteristics reduces temporal noise by 30%~50%, especially in low-light scenes; (4) Nonlinear compensation: Multi-point response curve fitting reduces pixel nonlinear deviation to below 3%, improving the accuracy of target temperature measurement; (5) High versatility: compatible with linear / area array and short-wave / mid-wave infrared cameras, requiring no hardware modification and easy to implement in engineering. Attached Figure Description

[0016] Figure 1 This is a schematic flowchart of the infrared image multi-point correction method proposed in this invention; Figure 2 This is a schematic diagram showing the image correction comparison in the infrared image multi-point correction method proposed in this invention; Figure 3 This is a schematic diagram of the frame of the infrared image multi-point correction device proposed in this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] like Figure 1 As shown in the figure, this embodiment proposes a method for multi-point correction of infrared images, which specifically includes the following steps: Step S1. Real-time acquisition of infrared images of 10 or more frames under four temperature scenarios: dark field, low illumination, medium illumination, and high illumination. Multi-frame averaging is performed on each frame to suppress temporal noise and generate the image to be analyzed.

[0019] In this embodiment, different temperature scenarios are set as dark field, low illumination, medium illumination and high illumination. Dark field means no light field, low illumination is set to T0 with a value of 20℃, medium illumination is set to the intermediate temperature Tm with a value of 25-30℃, and high illumination is set to T with a value of 35℃. In this embodiment, the multi-frame infrared image is generally greater than or equal to 10 frames.

[0020] Step S2. Extract pixel feature parameters from the image to be analyzed to generate blind pixel detection parameters based on the response rate, noise voltage, noise equivalent temperature difference, and linearity features of each pixel.

[0021] In this embodiment, the process of averaging multiple frames for each image is implemented using the following formula: Where F is the number of frames. For the f-th frame pixel The signal voltage.

[0022] In this embodiment, temporal noise is suppressed by averaging each frame of the image over multiple frames.

[0023] In this embodiment, the calculation formulas for the generated response rate, noise voltage, noise equivalent temperature difference, and linearity characteristics are expressed as follows: In the above formula, Representing pixels at different temperatures Response rate , Pixels at temperature T in a high-illuminance scene and at temperature T0 in a low-illuminance scene, respectively. Signal voltage; Represents pixels in dark fields Noise voltage, quantized in DN; Represents pixels in dark fields The response voltage is equivalent to the product of the response rate RIJ(x,y) and the temperature difference (T-T0); This represents the pixel in the f-th frame at temperature T0 in a low-light scene. The signal voltage; Indicates the noise-equivalent temperature difference, quantified in °C; Nonlinear deviation representing linearity characteristics Represents a cell The original signal voltage at temperature T under high illumination conditions. , Represents a cell The response curve fitting coefficients are given by std(), which represents the standard deviation. It should be noted that in this embodiment, VNIJ is the basic input parameter of NETDIJ, and NETDIJ is a derived evaluation parameter of VNIJ.

[0024] It should be noted that the responsivity is the voltage responsivity of a pixel at different temperatures, used to reflect the pixel's sensitivity to radiation. Noise voltage calculates the temporal noise of a pixel in dark conditions, used to characterize pixel stability. Noise equivalent temperature difference is used to measure the pixel's ability to resolve temperature changes.

[0025] Step S3. Mark the blind pixels and normal pixels of the image to be analyzed according to the constructed blind pixel table based on blind pixel detection parameters.

[0026] In this embodiment, the step of constructing a blind pixel table based on blind pixel detection parameters specifically includes fusing spatial response criteria, response rate criteria, temporal noise criteria, linearity criteria, and G-parameter criteria, and generating a blind pixel table through a logical OR operation. When marking blind pixels in the image to be analyzed using this blind pixel table, if any one of the performance parameters in the spatial response criteria, response rate criteria, temporal noise criteria, linearity criteria, or G-parameter criteria is abnormal, the detection result of the blind pixel is output. This embodiment integrates multi-dimensional criteria results to achieve a veto-defeat blind pixel judgment logic, ensuring no blind pixels are missed. The core characteristic of a blind pixel is the presence of any performance anomaly, not all performance anomalies. The logical OR operation can cover different types of blind pixels (dead pixels, overheated pixels, nonlinear pixels, etc.). Furthermore, the results of the five independent criteria are summarized into a unified blind pixel table, providing clear blind pixel location information for subsequent correction steps.

[0027] In this embodiment, the spatial response criterion is specifically to generate a first threshold interval based on the global mean and global standard deviation of the pixel response values ​​under different temperature scenarios such as dark field, medium illumination and high illumination. If the pixel response value is greater than the maximum value of the first threshold interval or less than the minimum value of the first threshold interval, the detection result of the blind pixel is output.

[0028] Specifically, the pixel response values ​​for dark fields, medium illumination, and high illumination are calculated separately: Global mean: , where data is the pixel response dataset of the current scene, reflecting the average response level of the pixels in the corresponding scene; Global standard deviation: This reflects the global dispersion of the response value; Upper limit threshold: ; Lower threshold: ; Judgment: If the pixel response value , marked as blind element.

[0029] The proportional threshold for spatial response criteria (fixed at 0.2, i.e. ±20%) controls the range of proportional anomalies. Indicates a statistical threshold (based on normal distribution 3) The principle is extended to cover extreme outliers, avoiding threshold failure caused by global mean shift; This represents the final upper and lower thresholds (unit: DN), which is the maximum / minimum value of the proportional threshold and the statistical threshold, taking into account both proportional anomalies and absolute anomalies. Represents a cell The response value in the current scenario (unit: DN).

[0030] In this embodiment, the response rate criterion is specifically to generate a second threshold interval based on the mean and standard deviation of the pixel response value difference under different temperature scenarios of high illumination and dark field. If the pixel response value difference is greater than the maximum value of the second threshold interval or less than the minimum value of the second threshold interval, the detection result of the blind pixel is output.

[0031] Specifically, the difference in pixel response values ​​under different temperature conditions in high-illuminance and dark-field scenarios is represented as the difference between the high-illuminance response value and the dark-field response value, and is expressed as... Wherein, the unit is DN, equivalent to response rate. The global mean of the difference is expressed as: This reflects the overall level of response difference. Global standard deviation of the difference: This reflects the degree of dispersion of the response difference; Upper limit threshold: ; Lower threshold: ; Judgment: If , marked as blind element.

[0032] The proportional threshold for the response rate criterion is fixed at 0.5, i.e. ±50%.

[0033] In this embodiment, the temporal noise criterion is specifically to generate a third threshold interval based on the mean and standard deviation of the noise standard deviation under different temperature scenarios such as dark field, medium illumination and high illumination. If the temporal noise value is greater than the maximum value of the third threshold interval or less than the minimum value of the third threshold interval, the detection result of the blind cell is output.

[0034] Specifically, the calculations of time-domain noise values ​​for dark fields, medium illumination, and high illumination are as follows: Noise mean: ,in This is a dataset of temporal noise standard deviations for the current scene, i.e., dark field, medium illumination, and high illumination, reflecting the overall noise level, in units of DN.

[0035] Noise standard deviation: This reflects the degree of dispersion of the noise; Upper limit threshold: ; Lower threshold: ; The determination method is as follows: if the time-domain noise value or , marked as blind element.

[0036] The proportional threshold for the time-domain noise criterion (fixed at 1.0, i.e. ±100%). In this embodiment, the linearity criterion is specifically to generate a nonlinear deviation value based on the slope and intercept obtained by fitting the pixel response values ​​under different temperature scenarios of high illumination and dark fields. If the nonlinear deviation value is greater than the fourth threshold, the detection result of the blind pixel is output.

[0037] Specifically, the parameters of the fitted line include the slope and the intercept, where: slope ,intercept ; Indicates the response value for high illumination. This represents the response value in the dark field. This indicates the high illuminance temperature (35℃). This indicates the dark field temperature (20℃).

[0038] Calculate the theoretical response under high illumination and dark field conditions: ; Nonlinear deviation: ; Represents a cell At temperature The actual response value is given in DN. The determination method is: if , marked as blind element.

[0039] The proportional threshold (fixed at 25DN) represents the linearity criterion and is set based on engineering experience to cover an acceptable range of nonlinearity.

[0040] In this embodiment, the G-parameter criterion is specifically that if the correction gain parameter is greater than the maximum value of the fifth threshold interval or less than the minimum value of the fifth threshold interval, then the detection result of the blind cell is output.

[0041] The specific determination method is as follows: If the two-point correction gain parameter , marked as blind element.

[0042] This indicates that the value can be 0.5, which is a threshold that can be set manually.

[0043] It should be noted that the above thresholds are generally set based on the characteristics of infrared detector pixels.

[0044] Step S4. For blind pixels, average interpolation is performed using the adaptive gain and offset parameters of normal pixels in the 3×3 neighborhood to achieve correction.

[0045] In this embodiment, the process of using the correction parameters of normal pixels in the neighborhood to perform average interpolation to achieve correction includes the following formula: In the above formula, Represents a blind pixel At temperature The corrected signal voltage; N represents the number of normal pixels in the 3×3 neighborhood. Represents a blind pixel The coordinates of normal pixels within a 3×3 neighborhood. Represents the normal pixels in the neighborhood The corrected signal voltage at temperature T. 3×3 neighborhood average interpolation balances "accuracy" and "computational complexity". The corrected values ​​of normal pixels in the neighborhood have compensated for non-uniformity. After interpolation, the transition between the blind pixel region and the surrounding image is natural and without obvious abruptness. Based on the position information of the blind pixel table, blind pixels are processed in a targeted manner to ensure the uniformity and continuity of the entire image and complete the final corrected output.

[0046] In this embodiment, for normal pixels, the correction is performed using the correction response curve based on pixel fitting using nonlinear deviation, as shown below: in, Represents a cell Adaptive gain parameters Represents a cell The adaptive offset parameters, Represents a cell At temperature The corrected signal voltage; Represents a cell At temperature The original signal voltage. Pixels whose parameters deviate too far from 1 indicate that their response characteristics are very different from those of most pixels. Even after two-point correction, the non-uniformity cannot be compensated, and they are essentially blind pixels.

[0047] It should be noted that a calibration curve is fitted separately for each pixel, solving the problem of poor wide-temperature adaptability of traditional two-point calibration; unlike traditional fixed-parameter calibration, adaptive gain... and offset It can compensate for individual differences (non-uniformity) and nonlinear bias of pixels; adaptive gain (G(x,y)) and offset (O(x,y)) ensure that the response of all pixels is consistent across a wide temperature range; this embodiment links feature extraction with correction execution to achieve a closed loop of accurate feature-adaptive correction.

[0048] To more clearly illustrate the content of this embodiment, such as Figure 2 As shown, a mid-wave infrared external array camera (640×512 resolution) is used as an example: (1) Data acquisition: 20 frames of images were acquired at 20℃ (dark field), 20℃ (low illumination), 30℃ (medium illumination), and 35℃ (high illumination); (2) Preprocessing: Calculate the value of each pixel The average value of multiple frames is obtained. This is used to suppress time-domain noise; (3) Feature extraction: Response rate RIJ: the difference between the signal voltage at 35℃ and 20℃ divided by the temperature difference; Noise voltage VNIJ: Standard deviation at 20℃ in a dark environment; Linearity: The maximum deviation is calculated by fitting a straight line to the response at 20℃, 30℃, and 35℃.

[0049] (4) Blind Pixel Detection: Pixels with spatial response deviation > ±20%, response rate deviation > ±50%, temporal noise value > 2 times the mean, nonlinear deviation value > 25DN, and G parameter deviation > ±50% are marked as blind pixels.

[0050] (5) Multi-point calibration: For normal pixels, the response curves at 35℃, 30℃, and 20℃ are fitted to obtain the correction parameters G and O; For blind pixels, G and O interpolation of 3×3 neighboring normal pixels is used to fill them; Output: The non-uniformity of the corrected image decreased from 15% to 4%. Figure 2 (a) is the original image. Figure 2 (b) is the output result diagram, from Figure 2 As can be seen in (b), there are no obvious defects in the blind pixel region and the target edge is clear.

[0051] As an example, this embodiment also proposes an apparatus for multi-point correction of infrared images to perform the method as described above, such as... Figure 3 As shown, the device 30 includes: The image acquisition module 301 is used to acquire infrared images of more than 10 frames in real time under four temperature scenarios: dark field, low illumination, medium illumination, and high illumination. It performs multi-frame averaging on each frame to suppress temporal noise and generate an image to be analyzed. The feature extraction module 302 is used to extract pixel feature parameters from the image to be analyzed and generate blind pixel detection parameters based on the response rate, noise voltage, noise equivalent temperature difference and linearity features of each pixel. The blind pixel labeling module 303 is used to label the blind pixels and normal pixels of the image to be analyzed according to the constructed blind pixel table based on blind pixel detection parameters. The blind pixel correction module 304 is used to perform average interpolation filling for blind pixels by using adaptive gain and offset parameters of normal pixels in a 3×3 neighborhood.

[0052] This embodiment also discloses an electronic device, including a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, it causes the processor to perform the steps of a method for multi-point correction of infrared images.

[0053] Specifically, the electronic device includes a processor and a memory connected via a system bus. The processor provides computing and control capabilities to support the operation of the entire electronic device. The memory may include non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. These computer programs can be executed by the processor to implement an image multi-point correction method provided in the following embodiments. The internal memory provides a cached runtime environment for the operating system computer programs in the non-volatile storage media.

[0054] The various modules in the calibration device provided in this application embodiment can be implemented in the form of a computer program. This computer program can run on a terminal or server. The program modules constituted by this computer program can be stored in the memory of the terminal or server. When the computer program is executed by a processor, it implements the steps of the method described in the embodiments of this application.

[0055] This embodiment also provides a computer-readable storage medium. One or more non-volatile computer-readable storage media containing computer-executable instructions, which, when executed by one or more processors, cause the processors to perform the steps of a method for multi-point correction of an infrared image. A computer program product containing instructions, which, when run on a computer, causes the computer to perform image correction.

[0056] Any references to memory, storage, database, or other media used in the embodiments of this application may include non-volatile and / or volatile memory. Suitable non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which is used as external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM).

[0057] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0058] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for multi-point correction of infrared images, characterized in that, Specifically, the steps include the following: Step S1. Real-time acquisition of infrared images of 10 or more frames under four temperature scenarios: dark field, low illumination, medium illumination, and high illumination. Multi-frame averaging is performed on each frame to suppress temporal noise and generate the image to be analyzed. Step S2. Extract pixel feature parameters from the image to be analyzed to generate blind pixel detection parameters based on the response rate, noise voltage, noise equivalent temperature difference, and linearity features of each pixel; Step S3. Mark the blind pixels and normal pixels of the image to be analyzed according to the constructed blind pixel table based on blind pixel detection parameters; Step S4. For blind pixels, average interpolation is performed using the adaptive gain and offset parameters of normal pixels in the 3×3 neighborhood to achieve correction.

2. The method for multi-point correction of infrared images according to claim 1, characterized in that, The process of averaging multiple frames of each image is implemented using the following formula: Where F is the number of frames. For the f-th frame pixel The signal voltage.

3. The method for multi-point correction of infrared images according to claim 1, characterized in that, The calculation formulas for the response rate, noise voltage, noise equivalent temperature difference, and linearity characteristics generated in step S2 are expressed as follows: In the above formula, Representing pixels at different temperatures Response rate , Pixels at temperature T in a high-illuminance scene and at temperature T0 in a low-illuminance scene, respectively. Signal voltage; Represents pixels in dark fields noise voltage, Represents pixels in dark fields The response voltage, This represents the pixel in the f-th frame at temperature T0 in a low-light scene. The signal voltage; Indicates the noise equivalent temperature difference; Nonlinear deviation representing linearity characteristics Represents a pixel The original signal voltage at temperature T under high illumination conditions. , Represents a pixel The coefficients for fitting the response curve.

4. The method for multi-point correction of infrared images according to claim 3, characterized in that, The steps for constructing a blind pixel table based on blind pixel detection parameters specifically include fusing spatial response criteria, response rate criteria, temporal noise criteria, linearity criteria, and G-parameter criteria, and generating a blind pixel table through a logical OR operation. When blind pixel labeling of the image to be analyzed is performed using the blind pixel table, if any performance parameter of the spatial response criteria, response rate criteria, temporal noise criteria, linearity criteria, or G-parameter criteria is abnormal, the detection result of the blind pixel is output.

5. The method for multi-point correction of infrared images according to claim 4, characterized in that, The spatial response criterion is specifically to generate a first threshold interval based on the global mean and global standard deviation of the pixel response values ​​under different temperature scenarios such as dark field, medium illumination and high illumination. If the pixel response value is greater than the maximum value of the first threshold interval or less than the minimum value of the first threshold interval, the detection result of the blind cell is output. The response rate criterion is specifically to generate a second threshold interval based on the mean and standard deviation of the pixel response value difference under different temperature scenarios of high illumination and dark field. If the pixel response value difference is greater than the maximum value of the second threshold interval or less than the minimum value of the second threshold interval, the detection result of the blind cell is output. The temporal noise criterion is specifically generated by generating a third threshold interval based on the mean and standard deviation of the noise standard deviation under different temperature scenarios such as dark field, medium illumination and high illumination. If the temporal noise value is greater than the maximum value of the third threshold interval or less than the minimum value of the third threshold interval, the detection result of the blind cell is output. The linearity criterion is specifically to generate a nonlinear deviation value based on the slope and intercept obtained by fitting the pixel response values ​​under different temperature scenarios of high illumination and dark field. If the nonlinear deviation value is greater than the fourth threshold, the detection result of the blind cell is output. The G-parameter criterion is specifically that if the correction gain parameter is greater than the maximum value of the fifth threshold interval or less than the minimum value of the fifth threshold interval, then the detection result of the blind cell is output.

6. The method for multi-point correction of infrared images according to claim 5, characterized in that, For blind pixels, the correction process, which uses adaptive gain and offset parameters of normal pixels in a 3×3 neighborhood for average interpolation, includes the following formula: In the above formula, Represents a blind pixel At temperature The corrected signal voltage; N represents the number of normal pixels in the 3×3 neighborhood. Represents a blind pixel The coordinates of normal pixels within a 3×3 neighborhood. Represents the normal pixels in the neighborhood The corrected signal voltage at temperature T.

7. The method for multi-point correction of infrared images according to claim 6, characterized in that, The method further includes correcting for normal pixels using a correction response curve based on nonlinear deviation pixel fitting, expressed as: in, Represents a pixel Adaptive gain parameters Represents a pixel The adaptive offset parameters, Represents a pixel At temperature The corrected signal voltage; Represents a pixel At temperature The original signal voltage.

8. An apparatus for multi-point correction of infrared images, for performing the method as described in any one of claims 1-7, characterized in that, The device includes: The image acquisition module is used to acquire infrared images of 10 or more frames in real time under four temperature scenarios: dark field, low illumination, medium illumination, and high illumination. Each frame is averaged to suppress temporal noise and generate an image to be analyzed. The feature extraction module is used to extract pixel feature parameters from the image to be analyzed and generate blind pixel detection parameters based on the response rate, noise voltage, noise equivalent temperature difference and linearity features of each pixel. The blind pixel labeling module is used to label the blind pixels and normal pixels of the image to be analyzed according to the constructed blind pixel table based on blind pixel detection parameters. The blind pixel correction module is used to fill blind pixels with average interpolation parameters of normal pixels in a 3×3 neighborhood to achieve correction.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method as described in any one of claims 1 to 7.

Images