Embedded image enhancement method and system for large-target-surface image sensor

Abstract

The invention provides an embedded image enhancement method and system for a large-target-surface image sensor. According to the method, temperature gradient distribution and environment humidity data of a large-target-surface image sensor are collected, and an environment parameter matrix and a temperature noise mapping model are constructed in an associated mode; and correcting optical distortion of the incident light signal by using the liquid crystal modulator array, and outputting an original image. Then through a double-path module, a main path extracts noise features, and an auxiliary path separates color shift features caused by humidity; and verifying the physical consistency of the characteristics based on a thermodynamic equation and screening effective data. And finally, dynamically optimizing the image contrast, the dynamic range and the target details by the embedded processor, and outputting an enhanced image. Through environmental parameter modeling, dual-path feature separation and physical verification, temperature and humidity interference is effectively suppressed, and detail enhancement and color fidelity of the large-target-surface image sensor are improved.

Description

Technical Field

[0001] The present application relates to the field of embedded image processing technology, and in particular to an embedded image enhancement method and system for a large-area image sensor. Background Art

[0002] In temperature-sensitive scenarios like industrial high-temperature inspection and aerospace remote sensing, large-area image sensors are susceptible to pixel response non-uniformity, optical distortion, and humidity-related color drift due to thermal expansion and drastic fluctuations in ambient temperature and humidity. These scenarios require embedded systems to implement dynamic non-uniformity correction within limited hardware resources, while simultaneously decoupling temperature gradient noise from humidity-induced color distortion and ensuring that enhanced image detail complies with the constraints of thermodynamic physics.

[0003] The current mainstream solution adopts a hybrid framework based on a pre-calibrated lookup table (LUT) and data-driven enhancement: the noise distribution of the sensor under different temperature and humidity combinations is pre-collected to generate a LUT library, and the pixel-level gain compensation coefficient is calculated through temperature interpolation to suppress fixed pattern noise; the compensated image is then input into the convolutional neural network model for detail enhancement, and the image quality is improved through end-to-end learning.

[0004] This solution is highly dependent on static calibration data and is difficult to adapt to complex transient temperature changes, resulting in compensation failure; its single temperature variable control mechanism ignores the nonlinear color shift caused by the synergistic effect of temperature and humidity, resulting in residual color distortion; in addition, the purely data-driven enhancement process of the convolutional neural network lacks thermodynamic constraints and is prone to over-sharpening details in high-temperature areas, thereby destroying the authenticity of the thermal radiation energy distribution. Summary of the Invention

[0005] The present application provides an embedded image enhancement method and system for a large-area image sensor, which is used to solve the problems in the prior art caused by ignoring the cross-interference of temperature and humidity and lack of physical constraints, resulting in image color distortion, residual noise and detail enhancement distortion.

[0006] In a first aspect, the present application provides an embedded image enhancement method for a large-area image sensor, comprising: Collecting the temperature gradient distribution of each pixel of the large-area image sensor, synchronously acquiring the ambient humidity data, correlating the temperature gradient distribution with the ambient humidity data according to spatial coordinates to form an environmental parameter matrix, and constructing a temperature noise mapping model based on the environmental parameter matrix; Processing the incident light signal received by the large-area image sensor through an integrated liquid crystal modulator array, adjusting the polarization direction of each modulation unit of the liquid crystal modulator array according to the temperature gradient distribution, eliminating optical distortion, and outputting a distortion-corrected original image signal; Inputting the original image signal into a dual-path feature extraction module, wherein a primary path in the dual-path feature extraction module extracts temperature-related noise distribution features based on the temperature noise mapping model, and an auxiliary path separates humidity-induced color shift features in combination with the ambient humidity data; Performing a physical consistency check on the noise distribution characteristics and the color shift characteristics, constructing an energy conservation constraint function through thermodynamic equations, verifying the matching degree between the noise distribution characteristics and the color shift characteristics and the physical response characteristics of the sensor, and screening out feature data that conforms to physical laws; An image enhancement operation is performed based on the feature data, the contrast and dynamic range of the image are dynamically adjusted and the detail features of the target area are enhanced through the embedded processor, and an optimized enhanced image is output.

[0007] Optionally, constructing an energy conservation constraint function through a thermodynamic equation, verifying the matching degree between the noise distribution feature, the color shift feature, and the physical response characteristics of the sensor, and screening out feature data that conforms to physical laws, includes: Based on the preset thermodynamic equation, an energy conservation constraint function including noise distribution feature quantity, color shift feature quantity and sensor physical response characteristic parameters is constructed; Calculating the characteristic energy value of the interaction between the noise distribution feature and the color shift feature pixel by pixel according to the energy conservation constraint function; Perform a matching check between the characteristic energy value of each pixel and the pre-stored allowable threshold range of the sensor's physical response characteristics; Pixel feature data whose feature energy values ​​are within the allowable threshold range are screened out as feature data that conforms to physical laws.

[0008] Optionally, dynamically adjusting the contrast and dynamic range of the image and enhancing the detail features of the target area by the embedded processor to output an optimized enhanced image includes: Dynamically generating a contrast adjustment parameter according to a noise distribution characteristic value in the characteristic data by an embedded processor; Dynamically generating a dynamic range extension parameter according to the color shift characteristic value in the characteristic data by the embedded processor; Identifying a target area through the embedded processor and calculating detail enhancement parameters according to pixel detail feature values ​​of the target area; The original image signal is synchronously processed in the embedded processor according to the contrast adjustment parameter, the dynamic range extension parameter and the detail enhancement parameter, and an optimized enhanced image is output.

[0009] Optionally, the processing of the incident light signal received by the large-area image sensor by an integrated liquid crystal modulator array, adjusting the polarization direction of each modulation unit of the liquid crystal modulator array according to the temperature gradient distribution, eliminating optical distortion, and outputting a distortion-corrected original image signal includes: Integrating a liquid crystal modulator array on the surface of the optical component of the large-area image sensor, establishing a positional correspondence between each modulation unit in the liquid crystal modulator array and a pixel point of the image sensor, and generating a modulation unit coordinate mapping table; Determining the polarization adjustment parameter of the corresponding modulation unit through the modulation unit coordinate mapping table according to the temperature data of each pixel point in the temperature gradient distribution; driving each modulation unit in the liquid crystal modulator array to adjust the polarization direction according to the polarization adjustment parameter, so that the incident light signal is projected to the optical component through the adjusted liquid crystal modulator array, thereby forming a compensation light path that eliminates optical distortion; The original image signal after distortion correction is output through the compensation optical path.

[0010] Optionally, driving each modulation unit in the liquid crystal modulator array to perform polarization direction adjustment according to the polarization adjustment parameter so that an incident light signal is projected to the optical component through the adjusted liquid crystal modulator array to form a compensation light path that eliminates optical distortion includes: Inputting the polarization adjustment parameter into the driving circuit of the liquid crystal modulator array to generate a voltage control signal corresponding to each modulation unit; Changing the molecular arrangement direction of each modulation unit in the liquid crystal modulator array according to the voltage control signal to achieve polarization direction adjustment; The incident light signal is passed through the liquid crystal modulator array after polarization direction adjustment to generate a corresponding polarization deflection amount, thereby generating a polarization-corrected light signal; The polarization-corrected optical signal is projected onto the surface of the optical component to form a compensation optical path for eliminating optical distortion.

[0011] Optionally, the inputting the original image signal into a dual-path feature extraction module, wherein a main path in the dual-path feature extraction module extracts temperature-related noise distribution features based on the temperature noise mapping model, and an auxiliary path separates humidity-induced color shift features in combination with the ambient humidity data, includes: Inputting the original image signal into a dual-path feature extraction module; In the main path of the dual-path feature extraction module, the temperature noise mapping model is called to process the original image signal to extract the noise distribution characteristics related to the temperature of each pixel; In the auxiliary path of the dual-path feature extraction module, the original image signal is synchronously received and combined with the ambient humidity data to separate the color shift feature induced by humidity at each pixel; The noise distribution feature and the color shift feature maintain the same pixel coordinate mapping relationship.

[0012] Optionally, the collecting of the temperature gradient distribution of each pixel of the large-area image sensor and the synchronous acquisition of the ambient humidity data, associating the temperature gradient distribution with the ambient humidity data according to spatial coordinates to form an environmental parameter matrix, and constructing the temperature noise mapping model based on the environmental parameter matrix include: The temperature gradient distribution of each pixel of the large-area image sensor is collected through the temperature sensor array, and the ambient humidity data is synchronously obtained through the humidity sensor; Associating and binding the temperature gradient distribution data corresponding to each pixel with the environmental humidity data according to the two-dimensional spatial coordinates of the pixel, and generating an environmental parameter matrix containing coordinate positions and corresponding data based on the associated data of all pixels; A temperature noise mapping model is directly constructed based on the mapping relationship between the temperature gradient value in the environmental parameter matrix and the image noise.

[0013] In a second aspect, the present application provides an embedded image enhancement system for a large-area image sensor, comprising: An acquisition module is used to acquire the temperature gradient distribution of each pixel of the large-scale image sensor, synchronously obtain environmental humidity data, associate the temperature gradient distribution with the environmental humidity data according to spatial coordinates to form an environmental parameter matrix, and construct a temperature noise mapping model based on the environmental parameter matrix; a processing module, configured to process the incident light signal received by the large-area image sensor through an integrated liquid crystal modulator array, adjust the polarization direction of each modulation unit of the liquid crystal modulator array according to the temperature gradient distribution, eliminate optical distortion, and output a distortion-corrected original image signal; an extraction module, configured to input the original image signal into a dual-path feature extraction module, wherein a primary path extracts temperature-related noise distribution features based on the temperature noise mapping model, and an auxiliary path separates humidity-induced color shift features in combination with the ambient humidity data; a verification module, configured to perform physical consistency verification on the noise distribution characteristics and the color shift characteristics, construct an energy conservation constraint function through thermodynamic equations, verify the matching degree between the noise distribution characteristics and the color shift characteristics and the physical response characteristics of the sensor, and screen out characteristic data that conforms to physical laws; The output module is used to perform image enhancement operations based on the feature data, dynamically adjust the contrast and dynamic range of the image and enhance the detail features of the target area through the embedded processor, and output the optimized enhanced image.

[0014] In a third aspect, the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an embedded image enhancement method for a large-area image sensor as described in the first aspect above.

[0015] In a fourth aspect, the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements the embedded image enhancement method for a large-area image sensor as described in the first aspect.

[0016] In an example of the present application, the temperature gradient distribution of each pixel of a large-area image sensor is collected, and ambient humidity data is simultaneously acquired. The temperature gradient distribution and the ambient humidity data are associated according to spatial coordinates to form an environmental parameter matrix, and a temperature noise mapping model is constructed based on the environmental parameter matrix. The incident light signal received by the large-area image sensor is processed by an integrated liquid crystal modulator array, and the polarization direction of each modulation unit of the liquid crystal modulator array is adjusted according to the temperature gradient distribution to eliminate optical distortion, and an original image signal with distortion correction is output. The original image signal is input into a dual-path feature extraction module. In the dual-path feature extraction module, a main path extracts temperature-related noise distribution features based on the temperature noise mapping model, and an auxiliary path separates humidity-induced color shift features based on the ambient humidity data. A physical consistency check is performed on the noise distribution features and the color shift features. An energy conservation constraint function is constructed using thermodynamic equations to verify the match between the noise distribution features and the color shift features and the physical response characteristics of the sensor, and feature data that conforms to physical laws is selected. An image enhancement operation is performed based on the feature data. An embedded processor dynamically adjusts the contrast and dynamic range of the image and enhances the detail features of the target area, and an optimized enhanced image is output.

[0017] The technical solution of this application has the following beneficial effects: This application constructs an environmental parameter matrix and a temperature noise mapping model related to temperature and humidity, and combines it with a liquid crystal modulator array to dynamically correct optical distortion; uses a dual-path feature extraction module to synchronously separate temperature noise features and humidity-induced color shift features, and performs physical consistency verification based on thermodynamic equations to screen valid data; and finally implements adaptive adjustment of contrast and dynamic range and enhancement of target details in an embedded processor, achieving a synergistic optimization effect of synchronously suppressing temperature and humidity interference, maintaining the physical authenticity of the image, and enhancing high-fidelity in key areas.

[0018] The physical consistency verification process is further defined: based on thermodynamic equations, an energy conservation constraint function is constructed that includes noise distribution features, color offset features, and sensor physical response parameters; the characteristic energy value of the interaction between the two types of features is calculated pixel by pixel; the energy value is matched and verified with the allowable threshold range of the sensor physical response; and feature data within the threshold range is selected as valid output. Through pixel-by-pixel quantitative verification of the energy conservation constraint function, distortion features that violate thermodynamic laws are accurately eliminated, ensuring that the enhancement process strictly adheres to the sensor's physical response characteristics. This fundamentally avoids image thermal radiation distortion and energy distribution anomalies caused by environmental interference, thereby improving the physical credibility and scene adaptability of the enhancement results.

[0019] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0021] Figure 1 A flow chart showing an embedded image enhancement method for a large-area image sensor provided by the present application is shown; Figure 2 A scene diagram showing an embedded image enhancement method for a large-area image sensor provided by the present application; Figure 3 A schematic structural diagram of an embedded image enhancement system for a large-area image sensor provided by the present application is shown; Figure 4 A schematic structural diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0023] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0024] Research shows that the current enhancement scheme for large-format image sensors in temperature-sensitive environments has three major restrictive defects due to its reliance on static temperature and humidity calibration lookup tables (LUTs) and a purely data-driven convolutional neural network enhancement framework: First, the pre-stored calibration lookup tables are difficult to cover dynamic temperature change scenarios such as transient thermal shock, resulting in the failure of non-uniformity correction; second, the single temperature variable control mechanism ignores the nonlinear color shift caused by the synergistic effect of temperature and humidity, resulting in residual regional color cast in the image after correction; third, the convolutional neural network enhancement process lacks the constraints of thermodynamic laws. When details are over-sharpened in high-temperature areas, the authenticity of the thermal radiation energy distribution will be destroyed, resulting in distortion of physical information in industrial inspection and remote sensing scenarios.

[0025] To address the above issues, this application proposes an embedded image enhancement method for large-area image sensors. Its core lies in eliminating optical distortion by constructing an environmental parameter matrix related to temperature and humidity to drive a liquid crystal modulator. It also utilizes dual-path feature extraction to decouple temperature noise and humidity color shift features. Furthermore, it combines the thermodynamic energy conservation constraint function to filter physically compliant data pixel by pixel, ultimately enabling dynamic enhancement via an embedded processor. This method replaces static LUTs with dynamic modeling to address transient temperature change failures, eliminates residual color shift through dual-path separation, and relies on physical verification to maintain thermal radiation authenticity. This method achieves high-fidelity enhancement of key details while ensuring compliance with the physical laws of the image.

[0026] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0027] Figure 1 A flowchart of an embedded image enhancement method for a large-area image sensor is provided in an embodiment of the present application. Figure 1 As shown, the method includes: 101. Collect the temperature gradient distribution of each pixel of the large-area image sensor, and simultaneously obtain environmental humidity data, associate the temperature gradient distribution with the environmental humidity data according to spatial coordinates to form an environmental parameter matrix, and construct a temperature noise mapping model based on the environmental parameter matrix; Optionally, step 101 may specifically include the following steps: 1011. Collect the temperature gradient distribution of each pixel of the large-surface image sensor through the temperature sensor array, and synchronously obtain the ambient humidity data through the humidity sensor; 1012. Associating and binding the temperature gradient distribution data corresponding to each pixel with the environmental humidity data according to the two-dimensional spatial coordinates of the pixel, and generating an environmental parameter matrix including coordinate positions and corresponding data based on the associated data of all pixels; 1013. Directly construct a temperature noise mapping model based on the mapping relationship between the temperature gradient value in the environmental parameter matrix and the image noise.

[0028] In the above scheme, the temperature gradient distribution refers to the temperature change rate data of each pixel position on the surface of the large-scale image sensor, including the temperature difference and spatial change trend between adjacent pixels, which is used to quantify the pixel response difference caused by thermal expansion.

[0029] Ambient humidity data refers to the absolute humidity of the sensor's environment, reflecting the interference intensity of water molecule concentration in the air on the optical path. The environmental parameter matrix is ​​a structured dataset that binds the two-dimensional coordinates, temperature gradient, and ambient humidity value H of each pixel point, establishing a mapping relationship between spatial location and environmental parameters. The temperature noise mapping model is an expression that describes the mathematical relationship between temperature gradient and image noise intensity and is used to predict the pixel-level noise distribution caused by temperature changes.

[0030] In the embodiment of the present application, first, step 1011 uses a high-precision temperature sensor array, such as a micron-scale thermocouple grid, to collect temperature data from each pixel on the surface of a large-scale image sensor to obtain a temperature gradient distribution. Simultaneously, a humidity sensor integrated within the sensor module simultaneously acquires ambient humidity data. The temperature sensor array covers the sensor target surface in a grid format, with each sensing unit corresponding to a pixel position, directly measuring the temperature at that point. The humidity sensor continuously outputs ambient absolute humidity data. For example, if pixel (10, 20) measures 52.1°C, the adjacent pixel (10, 21) measures 53.0°C, and the ambient absolute humidity is 65% RH.

[0031] Then, in step 1012, based on the temperature gradient distribution and ambient humidity data output in step 1011, the temperature change rate of each pixel and its adjacent pixels is calculated to generate a temperature gradient value. , the calculation formula is as follows: ,in The current pixel The temperature value, For adjacent pixels The temperature value of the sensor is d, and the physical distance between the two pixels is determined by the sensor design specifications. Then the two-dimensional coordinates of each pixel are , the corresponding temperature gradient value The environmental parameter matrix is ​​finally integrated to form a structured data unit. Each row of the matrix stores the environmental parameter of a single pixel, and establishes an accurate mapping relationship between the spatial position and the environmental parameter. For example, the pixel (10,20) corresponds to a temperature of 52.1°C, and the pixel (10,21) corresponds to a temperature of 53.0°C. The temperature gradient between these two pixels is , combined with the corresponding ambient humidity value H of 65, the generated environmental parameter matrix is ​​[x=10,y=20,ΔT=0.9,H=65].

[0032] Finally, step 1013 uses the temperature gradient value data in the environmental parameter matrix and the pre-stored historical noise database to perform correlation analysis. The temperature gradient value is fitted by the linear regression algorithm. The mathematical relationship between the noise intensity N and the temperature noise mapping model is constructed. The specific process is: extract all temperature gradient values ​​in the matrix and the historical noise measured value at the corresponding position, and use the least square method to solve the coefficients to obtain the form The model will be used to predict the temperature-induced noise intensity of any pixel point. For example, the temperature gradient value The historical noise intensity measured value N = 15dB, k = 0.03, b = 0.5 are obtained through linear regression fitting, and the noise mapping model is established. .

[0033] In practical applications, in the surface temperature monitoring of industrial equipment A, when executing step 1011, the temperature sensor measures the temperature of pixel (100,200) as 51.8°C, the temperature of the adjacent pixel (100,201) as 54.3°C, and the humidity sensor outputs 70%RH; step 1012 calculates the gradient at this position. , bind coordinates (100,200), =2.5, H=70 to generate matrix entries; Step 1013 calls the historical database, records N=0.56dB when ΔT=2.0℃ / mm, and establishes a model through linear regression , substitute =2.5, so the predicted noise intensity N=0.03×2.5+0.5=0.575dB.

[0034] The above-mentioned 101 overall solution achieves precise spatial correlation of temperature and humidity data through synchronous data acquisition of the temperature sensor array and humidity sensor, pixel-level temperature gradient calculation and environmental parameter matrix construction; further based on historical noise data regression modeling, it generates a quantitative mapping model that can dynamically predict temperature noise, providing an adaptive computing basis for noise suppression in complex temperature change scenarios, and significantly improving the accuracy and robustness of non-uniformity correction.

[0035] 102. Processing the incident light signal received by the large-area image sensor through an integrated liquid crystal modulator array, adjusting the polarization direction of each modulation unit of the liquid crystal modulator array according to the temperature gradient distribution, eliminating optical distortion, and outputting a distortion-corrected original image signal; Optionally, step 102 may specifically include the following steps: 1021. Integrate a liquid crystal modulator array on the surface of the optical component of the large-area image sensor, establish a positional correspondence between each modulation unit in the liquid crystal modulator array and a pixel point of the image sensor, and generate a modulation unit coordinate mapping table; 1022. Determine a polarization adjustment parameter of a corresponding modulation unit using the modulation unit coordinate mapping table according to the temperature data of each pixel in the temperature gradient distribution; 1023. Drive each modulation unit in the liquid crystal modulator array to perform polarization direction adjustment according to the polarization adjustment parameter, so that the incident light signal is projected to the optical component through the adjusted liquid crystal modulator array, thereby forming a compensation optical path that eliminates optical distortion; Among them, step 1023 may specifically include the following processes: inputting the polarization adjustment parameters into the driving circuit of the liquid crystal modulator array to generate a voltage control signal corresponding to each modulation unit; changing the molecular arrangement direction of each modulation unit in the liquid crystal modulator array according to the voltage control signal to achieve polarization direction adjustment; allowing the incident light signal to generate a corresponding polarization deflection amount through the liquid crystal modulator array after polarization direction adjustment to generate a polarization-corrected light signal; and projecting the polarization-corrected light signal onto the surface of the optical component to form a compensation light path to eliminate optical distortion.

[0036] 1024. Output the original image signal after distortion correction through the compensation optical path.

[0037] In the above scheme, the liquid crystal modulator array refers to an adjustable optical device integrated into the optical surface of the image sensor. It is composed of a matrix of micron-scale liquid crystal cells. Each cell can change the internal molecular arrangement direction through a voltage signal, thereby dynamically adjusting the polarization angle of the transmitted light to correct the optical path offset caused by thermal distortion. The modulation unit coordinate mapping table refers to an index data set that records the one-to-one correspondence between each modulation unit in the liquid crystal array and the spatial position of the image sensor pixel. It is generated through a coordinate calibration algorithm to ensure the precise association between temperature data and optical modulation units. The polarization adjustment parameter refers to the amount of adjustment of the polarization angle of the liquid crystal unit calculated through a physical model based on the temperature gradient value of the pixel point, which is used to drive the directional deflection of the liquid crystal molecules to offset thermal expansion distortion. The compensation optical path refers to the corrective optical path formed by the polarization direction of the incident light signal, eliminating the geometric deformation caused by the uneven temperature distribution on the sensor surface, and ultimately outputting the original image signal with a true geometric structure.

[0038] In this embodiment, a liquid crystal modulator array, composed of tens of thousands of micron-sized liquid crystal cells, is first mounted in close contact with the optical lens surface of the image sensor in step 1021. A laser positioning system scans the physical coordinates of each liquid crystal cell and sensor pixel, and a coordinate matching algorithm is run to generate a coordinate mapping table for the modulator cells. For example, the physical location of liquid crystal cell number M20 is precisely bound to the sensor pixel region (100, 200), and M21 is bound to (100, 201), forming a one-to-one "liquid crystal cell-pixel" location database.

[0039] Next, in step 1022, the liquid crystal unit number corresponding to each pixel is searched according to the modulation unit coordinate mapping table, and the temperature gradient value of each pixel point in the temperature gradient distribution is calculated. , adjust the liquid crystal unit corresponding to each pixel, and the polarization adjustment parameter is calculated by the formula Calculate, where According to the calibration experiment, 0.3° / ℃ / mm is usually used. The corresponding temperature gradient value for each pixel. For example, the temperature gradient at pixel (100,200) =2.5℃ / mm, the corresponding liquid crystal unit is M20, and its polarization adjustment parameter is calculated as .

[0040] Then, in step 1023, the calculated polarization adjustment parameter is input into the driving circuit of the liquid crystal modulator array, and the voltage control signal of the modulation unit is generated through a preset voltage conversion model. The conversion formula is as follows: ,in =5 is the voltage conversion coefficient; the voltage acts on the liquid crystal unit electrode, causing the liquid crystal molecules inside the modulation unit to rotate in a directional manner under the action of the electric field, synchronously changing the polarization direction of the transmitted light; at this time, the incident light passes through the rotated liquid crystal layer and undergoes polarization deflection, accurately offsetting the thermal expansion optical path deviation caused by the temperature gradient in the area. For example, the pixel (100,200) corresponds to the polarization adjustment parameter , the voltage control signal is generated by the formula =3.75, the modulation unit is controlled to rotate 0.75° to offset the temperature gradient in this area The corrected optical signal is re-projected perpendicularly to the imaging surface of the optical component, forming a compensating optical path that eliminates geometric distortion, making the metal edge that was originally bent due to thermal deformation appear straight on the sensor target surface.

[0041] Finally, the compensation light path generated by the liquid crystal modulator array is projected onto the target surface of the image sensor to eliminate the optical distortion caused by the temperature gradient. The sensor converts the corrected light signal into the original image electrical signal output without any enhancement processing. This signal completely retains the geometric structure and radiation information of the real scene, providing a high-fidelity input basis for the subsequent dual-path feature extraction module.

[0042] In a practical application, during the overheating component inspection of industrial equipment B, the image sensor detected a temperature gradient of ΔT = 3.2°C / mm at the target pixel (150, 250). A pre-generated modulation unit coordinate mapping table confirmed that this area was managed by liquid crystal unit M75. The polarization adjustment parameters were calculated based on the physical model: ; Input the θ value into the drive circuit and calculate the required voltage according to the voltage conversion formula: , applied to the M75 unit to rotate its liquid crystal molecules by 0.96°, and synchronously correct the polarization direction of the incident light to offset the equivalent optical path deflection caused by thermal expansion; finally, the compensation light path is projected to the sensor to output the corrected image. The edges of the heat sink fins that were originally distorted due to high temperature distortion are restored to a straight shape, forming an original image signal with a true geometric structure.

[0043] The above-mentioned 102 overall solution offsets the optical distortion caused by thermal expansion through dynamic adjustment of liquid crystal polarization driven by temperature data, ensuring that the large-area sensor captures the original image of the true geometric structure in a high-temperature environment, providing high-fidelity input for subsequent feature extraction.

[0044] 103. Input the original image signal into a dual-path feature extraction module, wherein a primary path in the dual-path feature extraction module extracts temperature-related noise distribution features based on the temperature noise mapping model, and an auxiliary path separates humidity-induced color shift features in combination with the ambient humidity data; Optionally, step 103 may specifically include the following steps: 1031. Input the original image signal into a dual-path feature extraction module; 1032. In the main path of the dual-path feature extraction module, call the temperature noise mapping model to process the original image signal and extract the noise distribution feature related to the temperature of each pixel; 1033. In the auxiliary path of the dual-path feature extraction module, synchronously receive the original image signal and combine it with the ambient humidity data to separate the color shift feature induced by humidity at each pixel; 1034. The noise distribution feature and the color shift feature maintain the same pixel coordinate mapping relationship.

[0045] In the above solution, the dual-path feature extraction module refers to an embedded architecture for parallel processing of image signals, which includes independent main and auxiliary paths to achieve physical decoupling of temperature and humidity interference. The noise distribution feature refers to the pixel-level signal distortion data directly caused by the temperature gradient, including the charge migration intensity and spatial distribution pattern caused by thermal expansion, which is used to characterize the non-uniformity of the sensor response. The color offset feature refers to the color channel deviation caused by the difference in the optical path caused by humidity changes, including the intensity offset value and spatial distribution of a specific band, which is used to identify humidity-induced color distortion. The pixel coordinate mapping relationship refers to the spatial correspondence rule that strictly binds the feature data output by the main / auxiliary paths to the pixel position of the original image to ensure the positioning consistency of subsequent physical verification.

[0046] In this embodiment of the present application, the corrected original image signal is first synchronously transmitted to the primary and secondary paths of the dual-path feature extraction module in step 1031. This module uses a hardware-level parallel processing architecture to ensure that the primary and secondary paths simultaneously receive completely identical image data streams and retain the spatial coordinate information of the original pixels. For example, the RGB value (120, 85, 110) of the pixel (150, 250) is synchronously obtained by both paths, providing a unified input basis for subsequent feature separation.

[0047] Next, step 1032 calls the temperature noise mapping model generated in step 101 through the main path of the dual-path feature extraction module to process the original signal pixel by pixel. First, the pixel is located and the temperature gradient value from the environmental parameter matrix is ​​read. Then, the noise intensity is calculated by substituting the formula of the temperature noise mapping model to characterize the intensity of thermally induced charge migration. Finally, the noise distribution characteristic value of the pixel is output as 0.596dB. For example, the temperature gradient value corresponding to the pixel point (150,250) is ΔT=3.2℃ / mm. Substituting it into the formula, the noise intensity of the pixel is calculated. The above operation is performed on all pixels in the original image signal to generate noise distribution features.

[0048] Then, step 1033 receives the original image signal synchronously through the auxiliary path of the dual-path feature extraction module, and calls the preset humidity color shift model in combination with the global environmental humidity data. ,in The humidity sensitivity coefficient is calibrated individually for the sensor. Is the original channel value. Pixel-by-pixel analysis of color shift: First, read the original value of the specified color channel of the current pixel, calculate the channel offset based on the humidity data, output the humidity-induced color shift feature value of the pixel, and generate a global pixel-level color shift feature map for subsequent physical verification. For example, the original value of the green channel at pixel (300,400) is G=110, the ambient humidity value is H=75%RH, and the humidity sensitivity coefficient is 0.01. Substituting it into the model, the color shift feature is calculated: , indicating that the green channel is enhanced by 82.5 units due to humidity.

[0049] Finally, step 1034 strictly binds the noise distribution features output by the primary path and the color shift features output by the auxiliary path to the same pixel coordinates, generating a feature pair with exact spatial matching. For example, if the noise feature value obtained by the primary path is 0.596dB and the color shift feature value obtained by the auxiliary path is ΔB = 77, these two features are bound to the pixel (150, 250), generating a matching feature pair {coordinates (150, 250): noise = 0.596dB, color shift = ΔB = 77}. This process achieves global pixel-level matching through a coordinate index table, ensuring that subsequent physical verification can accurately locate the temperature and humidity interference components at each location.

[0050] In practical applications, when extracting features from the original image in the welding point monitoring of industrial equipment D, the RGB value (100, 120, 90) at the pixel (200, 300) in the corrected original image is synchronously input into the dual-path module; the main path queries the temperature gradient at this position. =3.5℃ / mm, call the noise model Calculate the noise signature: , add 0.5 to get The auxiliary path is combined with the ambient humidity H=80%RH, the original value of the green channel G=120, and the color shift model Calculation: 0.01×80=0.8, then multiply by 120 to get ; The noise characteristic value and color shift characteristic value Bind coordinates (200, 300), output feature pairs {coordinates (200, 300), noise: , color cast: } for subsequent verification.

[0051] The above-mentioned 103 overall solution simultaneously extracts temperature noise distribution and humidity color offset features through a dual-path parallel architecture: the main path quantifies thermally induced signal distortion based on a temperature noise mapping model, and the auxiliary path analyzes moisture-induced channel imbalance in combination with ambient humidity data, thereby achieving physical separation of temperature and humidity interference; at the same time, it strictly maintains the pixel-level spatial alignment of feature data, providing independent and accurate input data for subsequent thermodynamic constraint verification, fundamentally solving the feature confusion problem caused by temperature and humidity coupling in traditional solutions, and significantly improving the interference suppression accuracy in complex environments.

[0052] 104. Performing a physical consistency check on the noise distribution feature and the color shift feature, constructing an energy conservation constraint function using a thermodynamic equation, verifying the matching degree between the noise distribution feature and the color shift feature and the physical response characteristics of the sensor, and screening out feature data that conforms to physical laws; Optionally, step 104 may specifically include the following steps: 1041. Construct an energy conservation constraint function based on a preset thermodynamic equation, including noise distribution feature quantities, color shift feature quantities, and sensor physical response characteristic parameters; 1042. Calculate, pixel by pixel, a characteristic energy value of the interaction between the noise distribution feature and the color shift feature according to the energy conservation constraint function; 1043. Perform a matching check between the characteristic energy value of each pixel and the pre-stored sensor physical response characteristic allowable threshold range; 1044. Filter out pixel feature data whose feature energy values ​​are within the allowable threshold range as feature data that conforms to physical laws.

[0053] In the above scheme, the energy conservation constraint function is a mathematical expression constructed based on the first law of thermodynamics. It includes noise distribution characteristics, color shift characteristics, and sensor physical response characteristic parameters. By quantifying the compliance of signal energy under temperature and humidity interference, it is used to assess whether the feature data conforms to physical laws. The characteristic energy value is the comprehensive energy evaluation value generated by the interaction between noise and color shift characteristics within the thermodynamic framework. It is calculated pixel by pixel using the constraint function and reflects the physical plausibility of the temperature and humidity interference at that location. The allowable threshold range is the boundary interval of acceptable characteristic energy values ​​within the physical limits of the sensor. It is determined by calibration of the sensor material's heat capacity and photoelectric response law and is used to filter out distorted data that violates energy conservation. Physically compliant feature data refers to a set of features that pass threshold verification. Its noise distribution and color shift strictly adhere to the sensor's thermodynamic response mechanism, ensuring that subsequent image enhancement does not violate physical authenticity.

[0054] In the embodiment of the present application, first, in step 1041, an energy conservation constraint function is constructed based on the first law of thermodynamics and the physical characteristic parameters of the sensor: ,in is the noise energy conversion coefficient, and its calibration value is , is the color deviation energy conversion coefficient, and its calibration value is , The calibration value of the temperature and humidity weight ratio is 0.5. is the noise characteristic value, This function quantifies the energy compliance of the interaction between the noise eigenvalue and the color shift feature, converting abstract physical laws into a calculable characteristic energy value E. For example, the interaction energy of a noise eigenvalue of 0.6dB and a color shift feature of ΔB=75 will be quantified according to this formula.

[0055] Next, in step 1042, based on the energy conservation constraint function constructed in step 1041, the characteristic energy value is processed pixel by pixel: the noise distribution characteristic value and color shift characteristic value of the current pixel are read and substituted into the energy conservation constraint function. Perform numerical calculations, and the calculation process strictly follows the arithmetic priority: first solve the noise components separately Color cast , and then sum to get the dimensionless characteristic energy value E, forming the energy distribution map of the whole image. For example, read the dual-path characteristic data of pixel (100,200): noise characteristic value N=0.6dB and color offset characteristic ΔB=75. Substitute into the function calculation: .

[0056] Then, in step 1043, the characteristic energy value E calculated in step 1042 is compared with the pre-stored sensor physical response allowable threshold range. A comparison check is performed: If the characteristic energy value E is within the specified range, the pixel's feature data is considered to comply with thermodynamic laws; if the characteristic energy value E exceeds the threshold, it is considered to violate physical laws. The check results generate a binary marker map, where 0 indicates violation and 1 indicates compliance, which is used to guide feature data screening. For example, if a pixel's characteristic energy value E = 6.75 is compared to the sensor's physical response threshold range of [5.0, 7.0], the result is compliance and is marked as 1 in the marker map. However, if a pixel's characteristic energy value E = 8.6 is exceeded, it is marked as a violation and is marked as 0 in the marker map.

[0057] Finally, step 1044 is used to filter the pixel feature data that have passed the verification according to the marking map of step 1043. For the pixels marked as compliant, their noise characteristics and color shift characteristics are retained and packaged as Data pairs; for pixels marked as illegal, the feature data is directly removed. Finally, the full-image physical compliance feature dataset is output and transmitted to the image enhancement module for subsequent processing.

[0058] In practical applications, in the heat sink temperature monitoring of industrial equipment D, dual-path feature data is input for pixel (500, 600): noise distribution characteristics , color shift characteristics ;According to the physical parameters of the sensor Constructing the energy function ; Substitute into numerical calculation: first find the noise component , and then calculate the color shift component ,final ;Will Compare with the pre-stored threshold [5.0, 7.0], determine compliance, mark 1 in the generated binary label map; retain the feature data . For the abnormal pixel (600,700) , calculated , exceeds the threshold [5.0, 7.0], perform feature removal, and mark 0 in the binary label map. Filter the pixel feature data that pass the verification in the label map, package them into data pairs, and output the full-image physical compliance feature dataset.

[0059] The above-mentioned 104 overall solutions use thermodynamic energy conservation constraint functions to verify the physical compliance of noise and color deviation features pixel by pixel, accurately screen out data that violates the physical response laws of the sensor, and ensure that subsequent image enhancement is performed only based on thermodynamically compatible feature data. This fundamentally eliminates image radiation distortion and geometric distortion caused by environmental interference, and significantly improves the physical authenticity and reliability of images in industrial inspection and remote sensing scenarios.

[0060] 105. Perform an image enhancement operation based on the feature data, dynamically adjust the contrast and dynamic range of the image and enhance the detail features of the target area through the embedded processor, and output an optimized enhanced image.

[0061] Optionally, step 105 may specifically include the following steps: 1051. Dynamically generate a contrast adjustment parameter according to the noise distribution characteristic value in the characteristic data by an embedded processor; 1052. Dynamically generate a dynamic range extension parameter according to the color shift characteristic value in the characteristic data by the embedded processor; 1053. Identify a target area by the embedded processor, and calculate detail enhancement parameters according to pixel detail feature values ​​of the target area; 1054. Synchronously process the original image signal in the embedded processor according to the contrast adjustment parameter, the dynamic range extension parameter, and the detail enhancement parameter, and output an optimized enhanced image.

[0062] In the above scheme, the contrast adjustment parameter refers to the image brightness and darkness difference adjustment coefficient dynamically generated based on the noise distribution eigenvalue, including the brightness compensation intensity and spatial distribution pattern of the signal distortion area, which is used to optimize the image layer performance. The dynamic range expansion parameter refers to the brightness level expansion coefficient calculated based on the color offset eigenvalue, including the exposure compression rate and channel balance factor of high dynamic scenes, which is used to restore shadow / highlight details. The detail enhancement parameter refers to the texture enhancement coefficient for the target area, including the local gradient response intensity and the sharpening convolution kernel weight, which is used to improve the recognition of key structures. The synchronous processing mechanism refers to the operating architecture of the embedded processor to execute contrast adjustment, dynamic range expansion and detail enhancement in parallel, and realize the synergy of the three types of parameters through hardware acceleration to ensure that the output image strengthens the target information while suppressing environmental interference.

[0063] In the embodiment of the present application, first, step 1051 reads the noise distribution characteristic value in the physical compliance feature data screened in step 104 through the embedded processor, calls the preset noise contrast mapping model, and calculates and generates the contrast adjustment parameter of the area. The calculation formula is as follows: ,in is the contrast adjustment parameter, is the adjustment coefficient, is the upper noise limit, is the characteristic value of noise distribution. The contrast adjustment parameter will be used to improve the brightness and darkness difference performance of low noise area. For example, the upper limit of noise is 1.0, and the noise distribution characteristic value at pixel (300,400) is , then the pixel contrast adjustment parameter is calculated .

[0064] Next, in step 1052 , based on the color shift characteristic value in the compliance characteristic data of the same pixel, the embedded processor applies the color shift dynamic range model to calculate and generate the dynamic range extension parameter. The calculation formula is as follows: ,in Reference gain, is the color shift threshold, is the color shift characteristic value, is the dynamic range expansion parameter. The dynamic range expansion parameter is used to restore the level details of the high humidity area. For example, the color shift characteristic value of the same pixel is , first calculate the color cast ratio , then calculate 1-0.7=0.3, and finally get the pixel dynamic range expansion parameter .

[0065] Then, in step 1053, the embedded processor scans the original image using the Sobel edge detection algorithm, locates the target area and outputs pixel detail feature values, such as edge point pixel detail feature values. The detail enhancement formula is then called , based on the pixel detail feature value calculation to generate detail enhancement parameters, where k is the sharpening coefficient and is set to 0.02, is the pixel detail feature value of the edge point. The detail enhancement parameter will specifically enhance the key structural texture. For example, the pixel detail feature value of the bolt edge gradient is , substituting into the formula to calculate the detail enhancement parameter is .

[0066] Finally, step 1054 starts the parallel computing unit through the embedded processor, and the first unit adjusts the parameters according to the ratio Perform histogram stretching to adjust contrast; the second unit expands the parameters according to the dynamic range Layered compression of brightness to expand dynamic range; the third unit applies detail enhancement parameters to the target area The intensity sharpening convolution kernel enhances details. The three-way processing results are fused to output an enhanced image that eliminates temperature and humidity interference and clearly defines key structures.

[0067] In practical applications, in industrial valve surface detection, the noise characteristics of the compliant pixels (350,450) are and color shift characteristics At the same time, the area is identified as the target structure by Sobel edge detection, that is, the thread edge gradient ; Then the noise characteristics Input the corresponding formula of the noise contrast model to calculate the contrast enhancement parameters ; Then based on the color shift feature Generate dynamic range extension parameters by calculating with color cast dynamic range model ; Then based on the edge detection and identification of the thread edge gradient , according to the detail enhancement formula, the sharpening intensity parameter is calculated ; Finally, the embedded processor performs image enhancement synchronously: contrast enhancement parameters Stretching histogram and expanding parameters according to dynamic range Layered compression brightness, and according to the sharpening intensity parameter The thread area is sharpened with a 1.9x intensity; the valve thread geometry in the output image is clear and undistorted, and the rust microscopic details are visible.

[0068] The above-mentioned 105 overall solution is based on physical compliance feature data. It dynamically generates contrast adjustment, dynamic range expansion and target detail enhancement parameters through an embedded processor, and simultaneously performs adaptive image optimization processing. While suppressing temperature and humidity interference, it significantly improves the structural clarity and texture details of key areas, and outputs high-fidelity enhanced images that meet industrial inspection requirements.

[0069] The following is a complete example for steps 101 to 105. Figure 2 As shown in the figure, in the surface temperature monitoring of the industrial reactor, the temperature data of the pixel points on the surface of the image sensor is first collected through the micron-level temperature sensor array. The temperature of the pixel (200,300) is 51.8℃, and the temperature of the adjacent pixel (200,301) is 54.3℃. The ambient humidity is 70%RH at the same time. The temperature gradient between the two pixels is calculated. ; Bind the coordinates (200,300), ΔT=2.5, and H=70 to generate the environmental parameter matrix entries; and fit the temperature noise mapping model based on the historical noise data , predict the noise intensity at this location .

[0070] Confirm that the pixel (200,300) corresponds to the liquid crystal unit M50 through the pre-generated modulation unit coordinate mapping table; based on the temperature polarization model Calculate polarization adjustment angle ; The driving circuit is converted according to the voltage formula The M50 unit is driven to rotate 0.75° to correct the polarization direction of the incident light; the compensation light path is projected and a correction image is output to eliminate the bolt thread deformation caused by high temperature on the surface of the reactor.

[0071] Next, the corrected original image is input into the dual-path module, and features are extracted for each pixel. For example, the pixel (200, 300) corresponds to RGB = (95, 110, 130) and is input into the dual-path module. The main path calls the noise model to obtain the corresponding noise feature value. ; The auxiliary path is combined with H=70%RH, and the corresponding color shift characteristic value is obtained according to the color shift model Bind pixel coordinates (200, 300) and output the feature pair {noise: 0.575dB, color shift: ΔB=91}.

[0072] Then, construct the energy conservation function ; Substitute into the eigenvalue calculation ; Compare with the threshold [5.0,7.0] in the threshold range and determine compliance; retain the feature data {noise: 0.575dB, color deviation: ΔB=91}.

[0073] Finally, the contrast parameters are generated by the embedded processor , dynamic range parameters With the gradient based on the reactor flange edge Obtained detailed parameters Referring to the above parameters, the image enhancement operation is performed simultaneously, the histogram stretching increases the contrast by 21.25%, the layered compression expands the dynamic range by 18%, the flange edge is sharpened by 1.8 times, and the final enhanced image is output. The microstructure of the bolt meshing surface is clearly visible in the enhanced image, and the high-temperature color spots are completely eliminated.

[0074] Figure 3 The present invention provides a schematic structural diagram of an embedded image enhancement system for a large-area image sensor. Figure 3 As shown, the system includes: An acquisition module 31 is configured to acquire the temperature gradient distribution of each pixel of the large-area image sensor, simultaneously acquire ambient humidity data, associate the temperature gradient distribution with the ambient humidity data according to spatial coordinates to form an environmental parameter matrix, and construct a temperature noise mapping model based on the environmental parameter matrix; a processing module 32 for processing the incident light signal received by the large-area image sensor through an integrated liquid crystal modulator array, adjusting the polarization direction of each modulation unit of the liquid crystal modulator array according to the temperature gradient distribution, eliminating optical distortion, and outputting a distortion-corrected original image signal; an extraction module 33 configured to input the original image signal into a dual-path feature extraction module, wherein a primary path extracts temperature-related noise distribution features based on the temperature noise mapping model, and an auxiliary path separates humidity-induced color shift features in combination with the ambient humidity data; a verification module 34 for performing physical consistency verification on the noise distribution characteristics and the color shift characteristics, constructing an energy conservation constraint function through thermodynamic equations, verifying the matching degree between the noise distribution characteristics and the color shift characteristics and the physical response characteristics of the sensor, and screening out characteristic data that conforms to physical laws; The output module 35 is used to perform image enhancement operations based on the feature data, dynamically adjust the contrast and dynamic range of the image and enhance the detail features of the target area through the embedded processor, and output an optimized enhanced image.

[0075] Figure 3 The embedded image enhancement system of a large image sensor can perform Figure 1 The implementation principles and technical effects of the embedded image enhancement method for a large image sensor described in the illustrated embodiment are not further elaborated. The specific manner in which the various modules and units of the embedded image enhancement system for a large image sensor in the aforementioned embodiment perform their operations has been described in detail in the relevant embodiments of the method and will not be further elaborated here.

[0076] In one possible design, Figure 3 The embedded image enhancement system of a large image sensor of the embodiment shown can be implemented as a computing device, such as Figure 4 As shown, the computing device may include a storage component 41 and a processing component 42; The storage component 41 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 42 .

[0077] The processing component 42 is used for the above Figure 1 The embodiment provides an embedded image enhancement method for a large-area image sensor.

[0078] The processing component 42 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0079] The storage component 41 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0080] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0081] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0082] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0083] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0084] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The embodiment shown is an embedded image enhancement method for a large-area image sensor.

[0085] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0086] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0087] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer or server) to execute the methods described in each embodiment or certain portions of the embodiments.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An embedded image enhancement method for a large-area image sensor, characterized in that: include: Collecting the temperature gradient distribution of each pixel of the large-area image sensor, synchronously acquiring the ambient humidity data, correlating the temperature gradient distribution with the ambient humidity data according to spatial coordinates to form an environmental parameter matrix, and constructing a temperature noise mapping model based on the environmental parameter matrix; Processing the incident light signal received by the large-area image sensor through an integrated liquid crystal modulator array, adjusting the polarization direction of each modulation unit of the liquid crystal modulator array according to the temperature gradient distribution, eliminating optical distortion, and outputting a distortion-corrected original image signal; Inputting the original image signal into a dual-path feature extraction module, wherein a primary path in the dual-path feature extraction module extracts temperature-related noise distribution features based on the temperature noise mapping model, and an auxiliary path separates humidity-induced color shift features in combination with the ambient humidity data; Performing a physical consistency check on the noise distribution characteristics and the color shift characteristics, constructing an energy conservation constraint function through thermodynamic equations, verifying the matching degree between the noise distribution characteristics and the color shift characteristics and the physical response characteristics of the sensor, and screening out feature data that conforms to physical laws; An image enhancement operation is performed based on the feature data, the contrast and dynamic range of the image are dynamically adjusted and the detail features of the target area are enhanced through the embedded processor, and an optimized enhanced image is output.

2. The method according to claim 1, characterized in that The energy conservation constraint function is constructed by using thermodynamic equations, the matching degree between the noise distribution characteristics, the color shift characteristics and the physical response characteristics of the sensor is verified, and characteristic data that conforms to physical laws is screened out, including: Based on the preset thermodynamic equation, an energy conservation constraint function including noise distribution feature quantity, color shift feature quantity and sensor physical response characteristic parameters is constructed; Calculating the characteristic energy value of the interaction between the noise distribution feature and the color shift feature pixel by pixel according to the energy conservation constraint function; Perform a matching check between the characteristic energy value of each pixel and the pre-stored allowable threshold range of the sensor's physical response characteristics; Pixel feature data whose feature energy values ​​are within the allowable threshold range are screened out as feature data that conforms to physical laws.

3. The method according to claim 1, characterized in that The method of dynamically adjusting the contrast and dynamic range of an image and enhancing the detail features of a target area by using an embedded processor to output an optimized enhanced image includes: Dynamically generating a contrast adjustment parameter according to a noise distribution characteristic value in the characteristic data by an embedded processor; Dynamically generating a dynamic range extension parameter according to the color shift characteristic value in the characteristic data by the embedded processor; Identifying a target area through the embedded processor and calculating detail enhancement parameters according to pixel detail feature values ​​of the target area; The original image signal is synchronously processed in the embedded processor according to the contrast adjustment parameter, the dynamic range extension parameter and the detail enhancement parameter, and an optimized enhanced image is output.

4. The method according to claim 1, wherein The method processes the incident light signal received by the large-area image sensor through the integrated liquid crystal modulator array, adjusts the polarization direction of each modulation unit of the liquid crystal modulator array according to the temperature gradient distribution, eliminates optical distortion, and outputs the original image signal after distortion correction, including: Integrating a liquid crystal modulator array on the surface of the optical component of the large-area image sensor, establishing a positional correspondence between each modulation unit in the liquid crystal modulator array and a pixel point of the image sensor, and generating a modulation unit coordinate mapping table; Determining the polarization adjustment parameter of the corresponding modulation unit through the modulation unit coordinate mapping table according to the temperature data of each pixel point in the temperature gradient distribution; driving each modulation unit in the liquid crystal modulator array to adjust the polarization direction according to the polarization adjustment parameter, so that the incident light signal is projected to the optical component through the adjusted liquid crystal modulator array, thereby forming a compensation light path that eliminates optical distortion; The original image signal after distortion correction is output through the compensation optical path.

5. The method according to claim 4, characterized in that The method of driving each modulation unit in the liquid crystal modulator array to adjust the polarization direction according to the polarization adjustment parameter so that the incident light signal is projected to the optical component through the adjusted liquid crystal modulator array to form a compensation light path that eliminates optical distortion includes: Inputting the polarization adjustment parameter into the driving circuit of the liquid crystal modulator array to generate a voltage control signal corresponding to each modulation unit; Changing the molecular arrangement direction of each modulation unit in the liquid crystal modulator array according to the voltage control signal to achieve polarization direction adjustment; The incident light signal is passed through the liquid crystal modulator array after polarization direction adjustment to generate a corresponding polarization deflection amount, thereby generating a polarization-corrected light signal; The polarization-corrected optical signal is projected onto the surface of the optical component to form a compensation optical path for eliminating optical distortion.

6. The method according to claim 1, characterized in that The original image signal is input into a dual-path feature extraction module, in which a main path extracts temperature-related noise distribution features based on the temperature noise mapping model, and an auxiliary path separates humidity-induced color shift features in combination with the ambient humidity data, including: Inputting the original image signal into a dual-path feature extraction module; In the main path of the dual-path feature extraction module, the temperature noise mapping model is called to process the original image signal to extract the noise distribution characteristics related to the temperature of each pixel; In the auxiliary path of the dual-path feature extraction module, the original image signal is synchronously received and combined with the ambient humidity data to separate the color shift feature induced by humidity at each pixel; The noise distribution feature and the color shift feature maintain the same pixel coordinate mapping relationship.

7. The method according to claim 1, characterized in that The method collects the temperature gradient distribution of each pixel of the large-area image sensor, synchronously obtains the ambient humidity data, associates the temperature gradient distribution with the ambient humidity data according to spatial coordinates to form an environmental parameter matrix, and constructs a temperature noise mapping model based on the environmental parameter matrix, including: The temperature gradient distribution of each pixel of the large-area image sensor is collected through the temperature sensor array, and the ambient humidity data is synchronously obtained through the humidity sensor; Associating and binding the temperature gradient distribution data corresponding to each pixel with the environmental humidity data according to the two-dimensional spatial coordinates of the pixel, and generating an environmental parameter matrix containing coordinate positions and corresponding data based on the associated data of all pixels; A temperature noise mapping model is directly constructed based on the mapping relationship between the temperature gradient value in the environmental parameter matrix and the image noise.

8. An embedded image enhancement system for a large-area image sensor, characterized in that: include: An acquisition module is used to acquire the temperature gradient distribution of each pixel of the large-scale image sensor, synchronously obtain environmental humidity data, associate the temperature gradient distribution with the environmental humidity data according to spatial coordinates to form an environmental parameter matrix, and construct a temperature noise mapping model based on the environmental parameter matrix; a processing module, configured to process the incident light signal received by the large-area image sensor through an integrated liquid crystal modulator array, adjust the polarization direction of each modulation unit of the liquid crystal modulator array according to the temperature gradient distribution, eliminate optical distortion, and output a distortion-corrected original image signal; an extraction module, configured to input the original image signal into a dual-path feature extraction module, wherein a primary path extracts temperature-related noise distribution features based on the temperature noise mapping model, and an auxiliary path separates humidity-induced color shift features in combination with the ambient humidity data; a verification module, configured to perform physical consistency verification on the noise distribution characteristics and the color shift characteristics, construct an energy conservation constraint function through thermodynamic equations, verify the matching degree between the noise distribution characteristics and the color shift characteristics and the physical response characteristics of the sensor, and screen out characteristic data that conforms to physical laws; The output module is used to perform image enhancement operations based on the feature data, dynamically adjust the contrast and dynamic range of the image and enhance the detail features of the target area through the embedded processor, and output the optimized enhanced image.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an embedded image enhancement method for a large-area image sensor as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the embedded image enhancement method for a large-area image sensor according to any one of claims 1 to 7 is implemented.

Images