Automatic dark level correction system

Abstract

The application discloses a dark level automatic correction system and relates to the technical field of image communication. The dark level automatic correction system comprises an accuracy calibration module, a uniformity calibration module and an accuracy evaluation module. The accuracy calibration module is used for dividing an image into a first region and a second region based on each regional pixel of the image obtained by real-time monitoring of a camera during a dark level calibration process, and performing image accuracy calibration according to the pixel number of the first region and each pixel value of the second region. The uniformity calibration module is used for performing pixel compensation calibration based on a second region uniformity evaluation result and dark current data. The pixel compensation calibration improves image uniformity by dynamically adjusting a basic temperature compensation coefficient and a sensor bias voltage. The accuracy evaluation module is used for performing calibration accuracy evaluation and feedback based on each regional pixel of a camera image after the dark level calibration is completed.

Description

Technical Field

[0001] This invention relates to the field of image communication technology, and more particularly to an automatic dark level correction system. Background Technology

[0002] Automatic dark level correction is a core technology in the fields of image sensors and video signal processing. Its core lies in eliminating dark current noise caused by the physical characteristics of devices (such as semiconductor thermal excitation, manufacturing process deviations, analog circuit offsets, etc.). Under no-light conditions, sensor pixels will still generate charge due to thermal noise, resulting in a non-zero baseline offset. If not corrected, this will cause the dark areas of the image to appear grayish, discolored, or have fixed pattern noise, significantly reducing the signal-to-noise ratio and dynamic range.

[0003] Existing automatic dark level correction methods mainly acquire the original signal values ​​of the optical black area of ​​the sensor or dedicated dark pixels under no-light conditions, calculate the baseline offset using statistical algorithms or dynamic models, and then perform real-time subtraction correction or clamping processing on the effective pixel signals through digital circuits or image signal processors. Some advanced systems also introduce machine learning to optimize noise modeling to eliminate dark current interference caused by thermal noise, process deviations, etc., and finally achieve accurate normalization of zero offset in the dark areas of the image.

[0004] For example, the image sensor dark level correction structure and method disclosed in patent application CN114205541A includes: S1, reading the voltage values ​​of pixel signals and ramp signals and inputting them into a comparator; S2, during the dark pixel row period, setting a reference voltage, and after the comparator flips twice, the output value is the CDS value of the dark pixel signal; S3, the digital module converts the output value of step S2 into a level value ΔV and feeds the dark level signal back to the reference voltage generator; S4, during the effective pixel row period, setting the reference voltage again, and after the comparator flips twice, the output value is the CDS value of the effective pixel signal; S5, the digital module directly outputs the effective pixel signal from step S4 through the output module.

[0005] For example, patent application CN104333717A discloses an algorithm and system for dark level correction of a CMOS image sensor, comprising: S1, setting a dark level upper limit and output data offset in the background system; S2, inputting an image with dark level information rows; S3, correcting defective pixels in the dark level information rows using a median filtering algorithm; S4, calculating the average dark level of each pixel channel in the dark level information rows after filtering and correction; S5, adjusting the output data offset; S6, performing dark level correction on the input image data; and S7, outputting a normal image. This system includes a defective pixel correction module, a correction information calculation module, and a dark level correction module. The correction information calculation module includes an average dark level calculation module and a data output offset adjustment module.

[0006] However, in the process of implementing the inventive technical solution in the embodiments of this application, it was found that the above-mentioned technology has at least the following technical problems:

[0007] In existing technologies, during the automatic dark level correction of cameras under no-light conditions, traditional time series models assume a fixed data distribution and cannot capture in real time the data distribution shifts caused by short-term noise changes such as sensor temperature fluctuations and dynamic accumulation of dark current. This results in insufficient adaptability of the model to non-stationary environments, leading to deviations between the correction parameters and the actual dark level values, and thus low accuracy of adaptive dark level correction. Summary of the Invention

[0008] This application provides an automatic dark level correction system, which solves the problem in the prior art where, during the automatic dark level correction of a camera under no-light conditions, the traditional time series model assumes a fixed data distribution and cannot capture in real time the data distribution shift caused by short-term noise changes such as sensor temperature fluctuations and dynamic accumulation of dark current. This results in insufficient adaptability of the model to non-stationary environments, leading to deviations between the correction parameters and the actual dark level values, and low accuracy of adaptive dark level correction. The system achieves rapid adaptation of the dark level correction system to short-term noise changes in non-stationary environments.

[0009] This application provides an automatic dark level correction system, including: an accuracy calibration module, a uniformity calibration module, and an accuracy evaluation module. The accuracy calibration module is used to divide the image into a first region and a second region based on the pixels of each region of the image obtained through real-time monitoring by the camera during the dark level calibration process. It then performs image accuracy calibration based on the number of pixels in the first region and the pixel values ​​in the second region. Image accuracy calibration includes dynamic adjustment of the mixing coefficient and evaluation of the uniformity of the second region. The uniformity calibration module is used to perform pixel compensation calibration based on the uniformity evaluation results of the second region and dark current data. Pixel compensation calibration improves image uniformity by dynamically adjusting the base temperature compensation coefficient and the sensor bias voltage. The accuracy evaluation module is used to evaluate and provide feedback on the calibration accuracy based on the pixels of each region of the camera image after the dark level calibration is completed.

[0010] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0011] 1. By monitoring the pixel distribution of images in real time and performing intelligent region segmentation, the system accurately identifies the characteristic differences between high-response and low-response regions. Simultaneously, it implements customized calibration for different region characteristics by dynamically adjusting the mixing coefficient and hierarchical compensation strategy. Furthermore, it establishes a multi-dimensional compensation model by combining dark current characteristics and ambient temperature changes for collaborative parameter optimization. Finally, it constructs a complete working cycle through a closed-loop verification mechanism and adaptive fine-tuning function. This enables dynamic modeling of image noise distribution, intelligent suppression of global noise and accurate compensation for local non-uniformity, dynamic adaptation and precise control of sensor operating status, and long-term stability and reliability of the system. As a result, the camera system can obtain accurate and stable dark level correction effects in various complex environments, comprehensively improving imaging quality and environmental adaptability.

[0012] 2. By comparing image pixels with a preset threshold to divide the image into a first region and a second region, and calculating the proportion of the first region as the noise ratio and comparing it with the threshold, a quantitative assessment of the image noise level is achieved. An optimal mixing coefficient is obtained through a preset mapping table and real-time noise ratio calculation, thereby constructing an adaptive filtering weight adjustment system. A closed-loop noise suppression mechanism is formed by calculating the difference value based on time-series noise data and dynamically adjusting the mixing coefficient. Adaptive gain coefficient configuration in different operating modes ensures optimal correction effects in various scenarios. This achieves accurate identification and classification of image noise, intelligent filtering control in dynamic environments, real-time monitoring and early warning of abnormal states, and adaptive dark level optimization correction across all scenarios, significantly improving image uniformity and system robustness.

[0013] 3. By dividing the second region into non-uniform and uniform regions and adopting differentiated compensation strategies, accurate identification of regions with different response characteristics is achieved. By dividing the dark current response curve of the non-uniform region into low-illuminance and high-illuminance segments and implementing incremental and attenuation compensation, a piecewise linear compensation model is established. By dynamically increasing the base temperature compensation coefficient in the low-illuminance range to suppress noise fluctuations and dynamically decreasing the base temperature compensation coefficient in the high-illuminance range to optimize energy efficiency, an adaptive temperature compensation adjustment mechanism is formed. By using mean matching and base temperature compensation coefficient lookup for the uniform region, a standardized compensation process is constructed, thereby achieving fine correction of sensor non-uniformity, adaptive adjustment of compensation parameters under dynamic environments, intelligent optimization of system energy efficiency, and consistent dark level output quality across the entire region, significantly improving image uniformity and system adaptability. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the structure of the automatic dark level correction system provided in the embodiments of this application.

[0015] Figure 2This is a flowchart of the dark level calibration process provided in an embodiment of this application. Detailed Implementation

[0016] This application provides an automatic dark level correction system, which solves the problem in the prior art where, during automatic dark level correction of cameras under no-light conditions, traditional time series models assume a fixed data distribution and cannot capture in real time the data distribution shifts caused by short-term noise changes such as sensor temperature fluctuations and dynamic accumulation of dark current. This results in insufficient adaptability of the model to non-stationary environments, leading to deviations between the correction parameters and the actual dark level values, and low accuracy of adaptive dark level correction. The overall approach is as follows:

[0017] First, the image pixel distribution is analyzed in real time, and the image is intelligently divided into a first region and a second region. The temporal filtering weight is optimized by dynamically adjusting the mixing coefficient, while the uniformity of the second region is evaluated. Next, based on the uniformity evaluation results, hierarchical compensation is implemented for non-uniform regions. The bias voltage and base temperature compensation coefficient are enhanced in the low-illuminance range, and the compensation intensity is reduced in the high-illuminance range. For uniform regions, a baseline compensation combined with a smooth correction of the temperature change rate is used. Finally, the number of abnormal pixel regions after calibration is counted, triggering pixel-level fine-tuning or early warning, realizing the rapid adaptation of the dark level correction system to short-term noise changes in non-stationary environments.

[0018] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0019] like Figure 1 The diagram shows the structure of the automatic dark level correction system provided in this embodiment of the application. The automatic dark level correction system includes: an accuracy calibration module, a uniformity calibration module, and an accuracy evaluation module. The accuracy calibration module is used to divide the image into a first region and a second region based on the pixels of each region of the image obtained from real-time camera monitoring during the dark level calibration process. It then performs image accuracy calibration based on the number of pixels in the first region and the pixel values ​​in the second region. Image accuracy calibration includes dynamic adjustment of the mixing coefficient and evaluation of the uniformity of the second region. The uniformity calibration module is used to perform pixel compensation calibration based on the uniformity evaluation results of the second region and dark current data. Pixel compensation calibration improves image uniformity by dynamically adjusting the base temperature compensation coefficient and the sensor bias voltage. The accuracy evaluation module is used to evaluate and provide feedback on the calibration accuracy based on the pixels of each region of the camera image after the dark level calibration is completed.

[0020] In this embodiment, the accuracy calibration module analyzes the image pixel distribution in real time and achieves adaptive noise suppression by dynamically adjusting the mixing coefficient and uniformity assessment; the uniformity calibration module dynamically adjusts the base temperature compensation coefficient and bias voltage based on the assessment results and combined with dark current data to effectively eliminate sensor non-uniformity; the accuracy assessment module verifies the calibration results at the pixel level and provides feedback for optimization. This invention breaks through the limitations of traditional fixed parameters, adapts to temperature fluctuations and noise changes through real-time dynamic adjustment, and the partition calibration strategy significantly improves dark field uniformity, making it particularly suitable for high-precision imaging scenarios.

[0021] like Figure 2 The diagram shows a flowchart of the dark level calibration process provided in this application embodiment. In the diagram, the threshold represents the noise ratio threshold, the bias voltage represents the sensor bias voltage, the temperature coefficient represents the base temperature compensation coefficient, non-uniformity represents a non-uniform region, and uniformity represents a uniform region. Specifically, the step of dividing the image into a first region and a second region based on the pixels of each region of the image obtained from real-time camera monitoring during the dark level calibration process includes: comparing the pixels of each region of the image with a preset pixel threshold; if the pixels of a certain region of the image are not lower than the preset pixel threshold, then the pixels of that region are recorded as the first region; if the pixels of a certain region of the image are lower than the preset pixel threshold, then the pixels of that region are recorded as the second region.

[0022] The step of calibrating image accuracy based on the number of pixels in the first region and the pixel values ​​in the second region includes: recording the ratio of the number of pixels in the first region to the total number of pixels in the image as the noise ratio; comparing the noise ratio with a preset noise ratio threshold; if the noise ratio exceeds the preset noise ratio threshold, dynamically adjusting the mixing coefficient based on the current noise ratio and the noise ratio of adjacent time monitoring points; if the noise ratio does not exceed the noise ratio threshold, determining whether the pixel value range of the second region exceeds a preset pixel range threshold; if so, marking the second region as a non-uniform region; otherwise, marking the second region as a uniform region; the mixing coefficient is used to adjust the weight distribution ratio of the current frame data and the historical frame data during the temporal filtering process to achieve an optimized balance between image noise suppression and image quality preservation; temporal filtering is a technical framework for cameras to acquire continuous time series image frames and perform image processing using inter-frame correlation.

[0023] The time series model is developed by preprocessing historical noise-proportion data, including outlier removal, missing data imputation, and smoothing and denoising. This is then combined with ADF (Augmented Dickey-Fuller Test) and KPSS (Kwiatkowski-Phillips-Schmidt) algorithms. The ShinTest method is used to complement the stationarity of historical data. The ADF test assumes the sequence is non-stationary, while the KPSS test assumes it is stationary. If the ADF test result matches the hypothesis, the KPSS test is used for further verification. If the ADF test result does not match the hypothesis, the historical noise proportion data is stationary. If the KPSS test result does not match the preset, the historical noise proportion data is stabilized through differencing or transformation methods. If the KPSS test result matches the preset, the historical noise proportion data is stationary. Next, the classical decomposition method is used to extract trend, periodic, and random data, and the autocorrelation function and partial autocorrelation function are used to identify the model order. Then, the maximum likelihood estimation method is used to generate predicted values. Finally, the rolling window validation algorithm is used to evaluate the effect of the predicted values, and new noise proportion data is continuously uploaded to update the model parameters to ensure that the model can accurately capture the evolution of noise.

[0024] The mixing coefficient is a weighting parameter used to control the mixing of multi-frame image data. It determines the proportion of contribution of the current frame and historical frames to the output result. The larger the mixing coefficient, the higher the weight of the current frame and the weaker the filtering effect. Conversely, the smaller the mixing coefficient, the higher the weight of historical frames and the stronger the noise reduction, but it may cause motion blur. The current frame is the image frame that the camera is processing during the current detection time period. It contains the latest information of the scene but is greatly affected by noise. In temporal filtering, the current frame usually has a high weight. Historical frames are a series of old frame images captured or processed before the current frame. They provide redundant information in the temporal dimension and are used to suppress random noise. In static scenes, historical frames can significantly improve the signal-to-noise ratio.

[0025] The steps for dynamically adjusting the mixing coefficient based on the current noise ratio and the noise ratio of adjacent time monitoring points include: pre-establishing a mapping table between the noise ratio and the mixing coefficient, which covers the mixing coefficient values ​​corresponding to different noise ratio ranges under different camera operating modes; calculating the current noise ratio in real time, and using the real-time calculated noise ratio as an index parameter according to the current camera operating mode, obtaining the corresponding optimal mixing coefficient from the mapping table; when there is no mixing coefficient item in the mapping table that matches the combination of the current noise ratio and the current camera operating mode, calculating the mixing coefficient based on the current noise ratio and the noise ratio of adjacent time monitoring points.

[0026] The formula for calculating the mixing coefficient is: In the formula, This represents the adjusted mixing coefficient. Indicates the current mixing coefficient. The adjustment gain coefficient determines the sensitivity of the noise ratio to the adjustment of the mixing coefficient. The larger the adjustment gain coefficient, the more aggressive the response to noise changes; the smaller the adjustment gain coefficient, the more conservative the adjustment. Under static, dynamic, and extreme noise scenarios, the acquisition quality scores corresponding to different adjustment gain coefficients are tested. Based on the test data, a Gaussian process surrogate model is constructed to establish the mapping relationship between the acquisition quality score and the test data. The expected value is estimated using Monte Carlo sampling, and the adjustment gain coefficient is iteratively selected through the expected improvement function. When the image quality score improves by more than a preset score improvement threshold after five consecutive iterations, the optimal adjustment gain coefficient is output. n is the nonlinear adjustment exponent, which controls the nonlinear relationship between the noise ratio and the adjustment amount of the mixing coefficient. When n=1, it is linear adjustment; when n>1, the adjustment is more drastic in high-noise areas; when n<1, it is smooth adjustment. This can be directly derived using the Poisson-Gaussian mixed noise model. p is the noise ratio difference value.

[0027] The steps for calculating the mixing coefficient based on the current noise ratio and the noise ratio of adjacent time monitoring points include: obtaining the noise ratio difference value based on the current noise ratio and the noise ratio of adjacent time monitoring points;

[0028] The mixing coefficient is dynamically adjusted based on the noise ratio difference value, the adjustment gain coefficient, and the nonlinear adjustment index; if the noise ratio still exceeds the preset noise ratio threshold after the number of consecutive adjustments exceeds the preset adjustment number threshold, a correction anomaly warning will be issued.

[0029] The noise ratio difference value is obtained as follows: the first noise ratio influence parameter is obtained by multiplying the noise ratio of the previous time monitoring point with its corresponding first weighting coefficient; the second noise ratio influence parameter is obtained by multiplying the noise ratio of the next time monitoring point with its corresponding second weighting coefficient; the first noise ratio influence parameter and the second noise ratio influence parameter are coupled to obtain the noise ratio influence parameter; the absolute difference between the current noise ratio and the noise ratio influence parameter is recorded as the noise ratio difference value.

[0030] The gain adjustment factor is dynamically configured according to the camera's operating mode, which includes static scene mode, dynamic scene mode, and extreme noise mode. The gain adjustment factor range for static scene mode is [0.8, 1.2], for dynamic scene mode it is [0.3, 0.6], and for extreme noise mode it is [1.5, 2.0].

[0031] In this embodiment, the present invention achieves precise dark level calibration through intelligent partitioning and dynamic adjustment. First, image pixels are divided into high-response (first region) and low-response (second region) based on a preset threshold. Global noise adaptive control is achieved through noise ratio quantization evaluation, breaking through the limitations of traditional fixed filter coefficients. If the threshold is not exceeded, local uniformity is detected by the pixel range of the second region. Regional range detection can identify local aging or damage of the sensor, making temperature and bias voltage compensation more targeted. Ultimately, the noise suppression ratio is improved in static scenes, and the ghosting effect of traditional time-domain filtering is avoided in dynamic scenes. The calibration speed and accuracy far exceed those of traditional methods.

[0032] Furthermore, the pixel compensation calibration steps based on the uniformity evaluation results of the second region and the dark current data include: if the second region is a non-uniform region, the dark current response curve is divided into a low-illuminance segment and a high-illuminance segment, and the bias voltage increment compensation is performed on the low-illuminance segment based on the preset bias voltage compensation amount, and the bias voltage attenuation compensation is performed based on the preset bias voltage compensation amount; if the second region is a uniform region, no additional operation is performed.

[0033] Pixel compensation calibration also includes: if the second region is a non-uniform region, the base temperature compensation coefficient is dynamically increased based on the fluctuation amplitude of dark current data in the low-illuminance range, and dynamically decreased based on the stability of dark current data in the high-illuminance range; if the second region is a uniform region, the average dark current data is calculated, and the average dark current data is matched with a preset mapping table corresponding to the average dark current data and the base temperature compensation coefficient. During the smoothing correction of the base temperature compensation coefficient based on the rate of change of ambient temperature, the system first uses a temperature sensor to monitor the changes in ambient temperature in real time and calculates the rate of temperature change per unit time. To eliminate instantaneous fluctuation noise in the temperature data, the system uses an exponentially weighted moving average algorithm to filter the rate of temperature change. Next, the system compares the filtered rate of temperature change with a preset rate correction coefficient lookup table to obtain the dynamic correction factor. When the temperature is rising, the correction factor is adjusted positively according to the temperature rise coefficient; when the temperature is falling, the correction factor is adjusted negatively according to the temperature fall coefficient. The temperature rise and fall coefficients are set separately for different temperature ranges: below 0℃ (low temperature range), the temperature rise coefficient is 0.15 and the temperature fall coefficient is 0.10; within the range of (0, 40℃), the temperature rise coefficient is 0.12 and the temperature fall coefficient is 0.08; above 40℃ (high temperature range), the temperature rise coefficient is 0.10 and the temperature fall coefficient is 0.05. This asymmetric configuration is determined by the characteristics of silicon-based sensors. Experimental data shows that the dark current growth rate during heating is about 1.5 times faster than the decay rate during cooling. Then, the system multiplies the base temperature compensation coefficient by the dynamic correction factor to calculate the final compensation coefficient. To ensure system stability, the final compensation coefficient is limited to a certain range of the base temperature compensation coefficient, preventing both undercompensation and overcompensation.

[0034] In this embodiment, the correction direction is opposite to the temperature change rate. In non-uniform regions, the adjustment range of the bias voltage is synchronized with the dynamic change of the base temperature compensation coefficient. In uniform regions, the adjustment range of the bias voltage is proportional to the corrected base temperature compensation coefficient. This invention employs an intelligent hierarchical compensation strategy to achieve accurate dark level correction. For detected non-uniform regions, the system intelligently divides the dark current response curve into low-illuminance and high-illuminance segments, implementing differentiated compensation for each, significantly improving the accuracy of dark current correction, especially improving the signal-to-noise ratio in low-illuminance regions. For uniform regions, after matching the base compensation value through a preset mapping table, smooth fine-tuning is performed in conjunction with the real-time temperature change rate, avoiding over-correction or under-correction caused by traditional fixed compensation values, thus improving the consistency of dark level in non-uniform regions and reducing power consumption.

[0035] Furthermore, the steps for evaluating and providing feedback on the calibration accuracy based on the pixels in each region of the camera image after dark level calibration include: counting the number of regions in the camera image after dark level calibration whose pixel values ​​are lower than a preset pixel threshold; if the number of regions exceeds the preset threshold, the corresponding region will start pixel-level fine-tuning mode and issue a warning of abnormal calibration parameters; if the number of regions does not exceed the preset threshold, no additional processing will be performed.

[0036] In this embodiment, the pixel-level fine-tuning mode is a technical mode used in the dark-level automatic correction system to finely adjust the response of individual or local pixels of the image sensor. It locates stable abnormal pixels using inter-frame difference, then generates a compensation lookup table by combining a dark current reference frame with real-time data. Based on an adaptive step-size strategy and a temperature-voltage coupling model, a DAC (Digital-to-Analog Converter Circuit) is used to compensate the target pixel with a millivolt-level bias voltage, achieving precise calibration of the single-pixel dark-level response. This invention ensures the final calibration quality through a closed-loop evaluation mechanism. When an abnormal area exceeds a safety threshold, the pixel-level fine-tuning mode is immediately triggered for precise local correction, and a parameter warning is issued. If the detection result is within the tolerance range, the current calibration parameters are maintained. Quantitative evaluation achieves objective verification of the calibration effect, avoiding the subjective judgment errors of traditional methods. The rapid location and fine-tuning function of abnormal areas further reduces residual non-uniformity. The warning mechanism provides clear indicators for system maintenance, improves the efficiency of preventative maintenance, and significantly enhances the long-term stability of dark-level calibration.

[0037] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0038] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1A device that provides the functions specified in one or more boxes.

[0039] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0040] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0041] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0042] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A dark level automatic correction system, characterized by, The method comprises an accuracy calibration module, a uniformity calibration module and an accuracy evaluation module. The accuracy calibration module is configured to divide the image into a first region and a second region based on the pixels of the image monitored by the camera in real time during the dark level calibration, and to perform image accuracy calibration according to the number of pixels in the first region and the pixel values in the second region, wherein the image accuracy calibration comprises dynamic adjustment of a mixing coefficient and evaluation of the uniformity of the second region. The step of dividing the image into the first region and the second region based on the pixels of the image monitored by the camera in real time during the dark level calibration comprises: comparing the pixels of each region of the image with a preset pixel threshold value, and if the pixels of a certain region of the image are not lower than the preset pixel threshold value, the pixels of the image region are recorded as the first region; if the pixels of a certain region of the image are lower than the preset pixel threshold value, the pixels of the image region are recorded as the second region. The step of performing image accuracy calibration according to the number of pixels in the first region and the pixel values in the second region comprises: recording the ratio of the number of pixels in the first region to the total number of pixels of the image as a noise ratio; comparing the noise ratio with a preset noise ratio threshold value, if the noise ratio exceeds the preset noise ratio threshold value, dynamically adjusting the mixing coefficient based on the current noise ratio and the noise ratio of the adjacent time monitoring point; if the noise ratio does not exceed the noise ratio threshold value, determining whether the range of pixel values in the second region exceeds a preset pixel range threshold value, if yes, marking the second region as a non-uniform region; otherwise, marking the second region as a uniform region; the mixing coefficient is used to adjust the weight distribution ratio of the current frame data and the historical frame data in the time domain filtering process, so as to realize the optimized balance between image noise suppression and image quality maintenance; the time domain filtering is a technical framework in which the camera acquires image frames in a continuous time sequence and processes the images by using the inter-frame correlation. The uniformity calibration module is configured to perform pixel compensation calibration based on the evaluation result of the uniformity of the second region and the dark current data, wherein the pixel compensation calibration improves the image uniformity by dynamically adjusting a basic temperature compensation coefficient and a sensor bias voltage. The accuracy evaluation module is configured to perform calibration accuracy evaluation and feedback based on the pixels of the image of the camera after the dark level calibration is completed.

2. The automatic dark level correction system of claim 1, wherein The step of dynamically adjusting the mixing coefficient based on the current noise ratio and the noise ratio of the adjacent time monitoring point comprises: a mapping relationship table of the noise ratio and the mixing coefficient is established in advance, wherein the mapping relationship table covers the mixing coefficient values corresponding to different noise ratio ranges in different camera operation modes; the current noise ratio is calculated in real time, and the optimal mixing coefficient corresponding to the noise ratio calculated in real time is obtained from the mapping relationship table as an index parameter according to the current camera operation mode; when there is no mixing coefficient item in the mapping relationship table that matches the combination of the current noise ratio and the current camera operation mode, the mixing coefficient is calculated based on the current noise ratio and the noise ratio of the adjacent time monitoring point.

3. The automatic dark level correction system of claim 2, wherein The step of calculating the mixing coefficient based on the current noise ratio and the noise ratio of the adjacent time monitoring point comprises: Obtaining a noise ratio difference value based on the current noise ratio and the noise ratio of the adjacent time monitoring point; Dynamically adjusting the mixing coefficient based on the noise ratio difference value, an adjustment gain coefficient, and a nonlinear adjustment index; When the number of continuous adjustments exceeds a preset adjustment number threshold, and the noise ratio still exceeds a preset noise ratio threshold, issuing a correction abnormality warning.

4. The automatic dark level correction system of claim 3, wherein The noise ratio difference value is obtained in the following manner: Obtaining a first noise ratio influence parameter based on the noise ratio of the previous time monitoring point and a corresponding first weighting coefficient; Obtaining a second noise ratio influence parameter based on the noise ratio of the next time monitoring point and a corresponding second weighting coefficient; Coupling the first noise ratio influence parameter and the second noise ratio influence parameter to obtain a noise ratio influence parameter; The absolute difference between the current noise ratio and the noise ratio influence parameter is recorded as the noise ratio difference value.

5. The automatic dark level correction system of claim 3, wherein The adjustment gain coefficient is dynamically configured according to the camera working mode, and the camera working mode comprises a static scene mode, a dynamic scene mode, and an extreme noise mode.

6. The automatic dark level correction system of claim 1, wherein The step of performing pixel compensation calibration based on the second area uniformity evaluation result and the dark current data comprises: If the second area is a non-uniform area, dividing the dark current response curve into a low-illumination section and a high-illumination section, performing bias voltage increment compensation on the low-illumination section based on a preset bias voltage compensation amount, and performing bias voltage attenuation compensation based on the preset bias voltage compensation amount; If the second area is a uniform area, no additional operation is performed.

7. The automatic dark level correction system of claim 6, wherein The pixel compensation calibration further comprises: If the second area is a non-uniform area, dynamically increasing the basic temperature compensation coefficient according to the fluctuation amplitude of the dark current data in the low-illumination interval, and dynamically decreasing the basic temperature compensation coefficient according to the stability degree of the dark current data in the high-illumination interval; If the second area is a uniform area, calculating the mean value of the dark current data, matching the mean value of the dark current data with a preset mapping relationship table corresponding to the mean value of the dark current data-basic temperature compensation coefficient, obtaining the basic temperature compensation coefficient, and smoothing the basic temperature compensation coefficient based on the environmental temperature change rate.

8. The automatic dark level correction system of claim 1, wherein, The step of evaluating and feeding back the calibration accuracy of each area pixel of the camera image after the dark level calibration comprises: Counting the number of regions in the camera image after the dark level calibration whose pixel values are lower than a preset pixel threshold value; If the number of regions exceeds a preset number threshold, starting a pixel-level fine-tuning mode in the corresponding region, and issuing a calibration parameter abnormality warning; If the number of regions does not exceed the preset number threshold, no additional processing is performed.

Images