Film exposure uniformity online detection method and system based on image processing

Abstract

The present application belongs to the technical field of image processing, and particularly relates to a film exposure uniformity online detection method and system based on image processing, which comprises the following steps: collecting original transmission images by using external source synchronization pulses, mapping real-time pixel gray values of each pixel point into an optical density matrix based on the Beer-Lambert optical absorption law; calculating the Laplacian operator response of each pixel point and constructing a negative exponential decay term, combining with the deviation degree to calculate the drift potential of each pixel point and constructing a drift potential field; performing local sliding window aggregation calculation in the drift potential field, extracting window non-uniformity based on global dispersion and the ratio of average potential and maximum potential; combining the window non-uniformity with the actual linear speed of the film to calculate the compensation voltage increment, and performing closed-loop correction. The present application realizes decoupling and separation of high-frequency background noise and low-frequency exposure decay signals, reduces false positives and missed detections in the production process of medical films, and realizes high-precision and anti-interference dynamic closed-loop control.

Description

Technical Field

[0001] This invention relates to the field of image processing technology. More specifically, this invention relates to an online detection method and system for film exposure uniformity based on image processing. Background Technology

[0002] In the high-speed production process of medical imaging films (such as DR, CT, MRI and other medical dry films), exposure uniformity is a core technical indicator that determines the final clinical imaging quality, grayscale resolution and diagnostic accuracy; any slight exposure unevenness may cause artifacts during clinical image interpretation, thereby misleading medical diagnosis.

[0003] To monitor exposure quality in real time during the production process, the industry commonly uses machine vision-based transmission-type online inspection systems. Existing mainstream inspection technologies typically use a high-brightness, uniform backlight to illuminate the film, and a linear scan camera on the other side to capture continuous transmission images. Subsequently, the captured images are divided into several local rectangular blocks, and a basic statistical model is used to calculate the average grayscale value and standard deviation of all pixels in each block. Finally, the calculated standard deviation of grayscale is directly compared with a manually set global fixed threshold. If the threshold is exceeded, it is determined to be uneven exposure and an alarm is triggered.

[0004] However, the aforementioned existing technologies face serious challenges of false alarms and missed detections in the complex and dynamic production environment: Medical imaging film, as a high-precision medical polymer composite material, inevitably experiences random fluctuations in micro-density during the manufacturing process. At the same time, it generates tiny mechanical stress ripples under high-speed winding tension. These structural fluctuations of the material background manifest as high-frequency local gray-scale abrupt changes in the transmission image. Such abrupt changes are easily confused with the characteristics of tiny lesions such as calcification points in medical images. Furthermore, the actual uneven exposure caused by the aging of the light source array or the decay of light emission usually exhibits a slow and gradual low-frequency photometric drift characteristic in spatial distribution, which can lead to large-area abnormal gray haze in the background of medical images.

[0005] Existing detection methods directly perform mixed calculations on the raw grayscale data, which is highly coupled with high-frequency material background noise and low-frequency exposure drift signals. This leads to an irreconcilable contradiction in threshold settings: if the threshold is too low, the high-frequency microstructural fluctuations of the material background will be erroneously amplified and trigger a large number of false alarms, resulting in an extremely high rate of medical film scrap; if the threshold is too high, the slowly evolving low-frequency exposure attenuation signal will be masked, causing serious missed detection of real defects, resulting in substandard medical device products with diagnostic risks entering the clinical field. Summary of the Invention

[0006] To address the technical problems of existing technologies being unable to remove interference from complex grayscale images and failing to meet the stringent quality control and high-precision online dynamic closed-loop control requirements in medical imaging film production, this invention provides solutions in the following aspects.

[0007] In a first aspect, the present invention provides an online detection method for film exposure uniformity based on image processing, comprising: synchronously acquiring an original transmission image of the film surface using an external sync pulse; mapping the real-time pixel grayscale values ​​of each pixel in the original transmission image to optical density values ​​of each pixel based on Beer-Lambert's optical absorption law and in combination with pre-acquired dark background grayscale and empty background grayscale, thereby constructing an optical density matrix; calculating the Laplacian operator response of each pixel in the optical density matrix; constructing a negative exponential decay term using the absolute value of the Laplacian operator response of each pixel, and combining the Laplacian operator response of each pixel... The deviation of the optical density value from the pre-calibrated reference density is used to calculate the drift potential energy of each pixel, and a drift potential energy field for separating high-frequency texture and low-frequency attenuation is constructed. In the drift potential energy field, a local sliding window aggregation calculation that maps to the physical size of the underlying high-brightness uniform backlight is performed. Based on the global dispersion of the drift potential energy of each pixel in the window, and the ratio of the average potential energy to the maximum potential energy of each pixel in the window, the window non-uniformity is extracted. Combining the window non-uniformity with the actual linear velocity of the film, the compensation voltage increment is calculated, and the compensation voltage increment is sent to the lighting control driver to perform closed-loop correction.

[0008] This invention precisely maps transmitted grayscale to an optical density matrix and constructs a drift potential energy field with a negative exponential decay term using the Laplace operator response. By utilizing the second-order differential properties, it penalizes and suppresses high-frequency physical noise generated by fluctuations in the microstructure of the film background, while completely preserving the low-frequency signal caused by the actual attenuation of the light source, thus achieving decoupling and separation of physical field signals. Combined with the subsequently extracted window non-uniformity and sinusoidal feedforward dynamic model, this invention overcomes the traditional threshold setting dilemma, reduces false alarms and missed detections in the medical imaging film production process, and achieves high-precision, anti-interference exposure closed-loop control under dynamic conditions.

[0009] Preferably, the method for obtaining the dark background grayscale and the unloaded background grayscale is as follows: during the system initialization phase, the bottom high-brightness uniform backlight is turned off, and the line scan camera is controlled to perform multiple continuous scans and take the average value in a completely dark state to obtain the dark background grayscale; the bottom high-brightness uniform backlight is turned on to the rated power, and in an unloaded state without film obstruction, the line scan camera is controlled to directly photograph the bottom high-brightness uniform backlight and perform multiple scans and take the average value to obtain the unloaded background grayscale.

[0010] Preferably, the step of mapping the real-time pixel grayscale values ​​of each pixel in the original transmission image to optical density values ​​based on Beer-Lambert's optical absorption law and combined with pre-acquired dark background grayscale and empty background grayscale, and constructing an optical density matrix, includes: In the formula, The coordinates of the target pixel; The optical density value of the target pixel; This is a commonly used logarithmic function with base 10; The real-time pixel grayscale value of the target pixel; Grayscale background; The background grayscale is empty; is the function to maximize the value; is a preset minimal positive constant to prevent logarithmic overflow; construct an optical density matrix containing the optical density values ​​of all pixels.

[0011] This invention utilizes Beer-Lambert's optical absorption law and the logarithmic ratio operator to upgrade the real-time pixel grayscale value of the representation to optical density, thereby removing stray interference from the nonlinear response and thermal noise of the linear array camera sensor from the underlying data stream and restoring the true medium penetration characteristics.

[0012] Preferably, calculating the Laplacian operator response of each pixel in the optical density matrix includes: reading the reference density of the film from the production batch archive; and constructing a matrix containing the center point and its eight neighboring points based on the obtained optical density matrix. Local spatial convolution mask; the local spatial convolution mask is used to slide and traverse the entire optical density matrix, and the second-order gradient difference is calculated between the optical density value of each pixel and the optical density value of the adjacent pixels in the local spatial convolution mask to obtain the Laplacian operator response of each pixel.

[0013] Preferably, a negative exponential decay term is constructed using the absolute value of the Laplacian operator response of each pixel, and the drift potential energy of each pixel is calculated by combining the deviation of the optical density value of each pixel from a pre-calibrated reference density. The calculation formula is as follows: In the formula, The coordinates of the target pixel; The drift potential energy of the target pixel; It is the natural logarithm function; The optical density value of the target pixel; As the baseline density; It is an exponential function with the natural constant as its base; The absolute value of the Laplacian operator response at the target pixel.

[0014] The potential energy calculation formula constructs an adaptive filter through a negative exponential decay shielding mechanism. Only extremely gradual spatial brightness decay, i.e., light source anomalies, can retain high potential energy, while high-frequency oscillations caused by any material scratches or blemishes will cause the exponential term to approach zero, thus shielding complex background physical pseudo-defects.

[0015] Preferably, the method for obtaining the local sliding window that maps to the physical size of the underlying high-brightness uniform backlight includes: setting a local sliding window that maps to the physical size based on the physical projected area of ​​a single light-emitting array unit in the underlying high-brightness uniform backlight.

[0016] Preferably, the formula for calculating the window non-uniformity is: In the formula, For window non-uniformity; This represents the total number of pixels in the window. For the traversal index of the pixel; For the first partial sliding window The drift potential energy of each pixel; The average potential energy of the window; It is the natural logarithm function; This represents the maximum potential energy of the window.

[0017] This invention effectively balances the breadth and depth of defects by constructing a window nonuniformity mechanism that includes a multiplicative coupling mechanism of global variance reflecting large-area uniformity deterioration and extreme logarithmic function reflecting abrupt changes in local light source dead zones. This ensures that the system has high detection sensitivity for both gradual and collapsed exposure anomalies.

[0018] Preferably, the compensation voltage increment is calculated by combining the window non-uniformity with the actual linear velocity of the film, and the compensation voltage increment is sent to the lighting control driver to perform closed-loop correction. This includes: continuously monitoring the window non-uniformity of all local sliding window outputs during system operation; when the window non-uniformity is less than a preset alarm threshold, the system maintains the closed-loop correction mode and calculates the compensation voltage increment sent to the lighting control driver of the corresponding area. The calculation formula is as follows: In the formula, To compensate for voltage increments; This is the voltage regulation step constant; For window non-uniformity; This is the alarm threshold; It is a sine function; Pi is a constant. This represents the actual linear velocity of the film. The absolute physical rated maximum linear velocity; the lighting control driver adjusts the driving current of the light-emitting array of the underlying high-brightness uniform backlight according to the received compensation voltage increment, forming a closed-loop control of physical execution.

[0019] This invention incorporates the actual linear velocity of the film as the independent variable of the sine operator into the compensation logic. This dynamic model generates damping at low linear velocities, effectively preventing over-compensation and overexposure during low-speed operation of the equipment. At high speeds, it releases the full compensation command to counteract the shortening of the integration time during high-speed operation, ensuring the physical inertia matching of compensation under cross-speed conditions.

[0020] Preferably, the method further includes: continuously monitoring the window non-uniformity of all local sliding window outputs during system operation; when the window non-uniformity of any local sliding window is greater than or equal to a preset alarm threshold, triggering an audible and visual alarm and controlling the system to stop.

[0021] In a second aspect, the present invention provides an online detection system for film exposure uniformity based on image processing, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-described online detection method for film exposure uniformity based on image processing is implemented.

[0022] By adopting the above technical solution, the above-mentioned online detection method for film exposure uniformity based on image processing is generated into a computer program and stored in a memory so that it can be loaded and executed by a processor. In this way, a terminal device can be made based on the memory and the processor for convenient use.

[0023] The beneficial effects of this invention are as follows:

[0024] This invention precisely maps transmitted grayscale to an optical density matrix and constructs a drift potential energy field with a negative exponential decay term using the Laplace operator response. By utilizing the second-order differential properties, it penalizes and suppresses high-frequency physical noise generated by fluctuations in the microstructure of the film background, while completely preserving the low-frequency signal caused by the actual attenuation of the light source, thus achieving decoupling and separation of physical field signals. Combined with the subsequently extracted window non-uniformity and sinusoidal feedforward dynamic model, this invention overcomes the traditional threshold setting dilemma, reduces false alarms and missed detections in the medical imaging film production process, and achieves high-precision, anti-interference exposure closed-loop control under dynamic conditions. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the online detection method for film exposure uniformity based on image processing in this invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0028] This invention discloses an online detection method for film exposure uniformity based on image processing, referring to... Figure 1 This includes steps S1-S4:

[0029] S1. The original transmission image is synchronously acquired using an external synchronous pulse. Based on Beer-Lambert's optical absorption law and combined with the pre-acquired dark background grayscale and empty background grayscale, the real-time pixel grayscale value of each pixel in the original transmission image is mapped to the optical density value of each pixel to construct an optical density matrix.

[0030] It should be noted that due to the microscopic displacement of medical imaging film during high-speed transmission, and the non-linear energy absorption rate of the photoelectric conversion phenomenon of a line scan camera, existing technologies that directly use the real-time pixel grayscale values ​​of each pixel are prone to introducing spatiotemporal misalignment and nonlinear physical errors. Furthermore, on high-speed production lines, the actual linear speed of medical imaging film is extremely high and exhibits minute dynamic fluctuations. If asynchronous image acquisition is directly relied upon using the free clock within the line scan camera, severe motion blur and vertical spatial scale distortion will occur, compromising the geometric basis of subsequent physical field analysis. The real-time pixel grayscale values ​​output by the line scan camera do not have a non-linear relationship with light energy and cannot directly reflect the film's true physical absorption characteristics of light. Since Beer-Lambert's optical absorption law can reduce the transmission phenomenon to the energy absorbed by the medium itself... In terms of optical conversion characteristics, this invention substitutes the external synchronization pulse and the real-time pixel grayscale value of each pixel into the logarithmic ratio operator, so that the grayscale of the dark background of the linear scan camera itself and the grayscale of the empty background are used as physical constant benchmarks to establish a ratio relationship. Combined with the underlying truncation constant, a lower limit constraint is established to prevent calculation overflow. The actual transmission appearance is nonlinearly mapped by the logarithmic function as the molecular end. This mechanism suppresses stray interference from the optical environment and preserves the true physical absorption capacity of the film, thereby solving the problems of spatiotemporal distortion and nonlinear aliasing, and constructing a physical optical density foundation. By connecting a high-frequency pulse encoder connected to the end of the guide roller shaft, strict spatiotemporal alignment is achieved, and Beer-Lambert's optical absorption law is used to convert the appearance signal into an optical density matrix that characterizes the essential properties of the material, laying a real physical data foundation for decoupling analysis.

[0031] Specifically, during the system initialization phase, the underlying high-brightness uniform backlight is turned off, and the line scan camera is controlled to perform multiple continuous scans and take the average value under a completely dark state to obtain the grayscale of the dark background; the underlying high-brightness uniform backlight is turned on to its rated power, and under an unloaded state without film obstruction, the line scan camera is controlled to directly photograph the underlying high-brightness uniform backlight multiple times and take the average value to obtain the grayscale of the unloaded background; during the production phase, an external synchronization pulse is output through a high-frequency pulse encoder, and the line scan camera triggers a line scan exposure when it receives each external synchronization pulse to obtain the original transmission image of the film surface.

[0032] Furthermore, for each pixel in the original transmission image, this pixel is taken as the target pixel, and the real-time pixel gray value of the target pixel is extracted. Combined with the gray values ​​of the dark background, the gray values ​​of the empty background, and the underlying truncation constant, the optical density value of the target pixel is calculated, thereby constructing an optical density matrix containing the optical density values ​​of all pixels.

[0033] The formula for calculating the optical density value of the target pixel is:

[0034]

[0035] In the formula, The coordinates of the target pixel; The optical density value of the target pixel; This is a commonly used logarithmic function with base 10; The real-time pixel grayscale value of the target pixel; Grayscale background; The background grayscale is empty; is the function for maximizing the logarithm; is a preset minimum positive constant to prevent logarithmic overflow, in this embodiment, is 0.01.

[0036] The calculation formula accurately restores the absolute optical absorption capability of the film at the corresponding spatial coordinates by subtracting the dark background grayscale of the line scan camera itself and taking the negative logarithm after calculating the ratio of the actual transmitted light intensity to the unobstructed incident light intensity. This eliminates the interference of the nonlinear response of the line scan camera, enabling the optical density matrix to reflect the comprehensive physical superposition state of the material distribution inside the film and the external illumination field.

[0037] S2. Calculate the Laplacian operator response of each pixel in the optical density matrix, construct a negative exponential decay term using the absolute value of the Laplacian operator response of each pixel, and calculate the drift potential energy of each pixel by combining the deviation of the optical density value of each pixel from the pre-calibrated reference density, and construct the drift potential energy field.

[0038] It should be noted that the transformed optical density matrix still mixes the high-frequency structural fluctuations of the film with the low-frequency exposure drift caused by light source defects. Normal film structural fluctuations have drastic changes in the second derivative in space, while real exposure anomalies are usually slow gradient spread. Therefore, this invention constructs a drift potential energy field containing a negative exponential penalty term of the spatial second-order differential operator. By utilizing the characteristic that the Laplacian operator response of the pixel is extremely sensitive to high-frequency abrupt changes, it adaptively and strongly suppresses the inherent background texture of the film. This makes the drastic differential response representing the high-frequency physical abrupt changes of the film converge rapidly exponentially after being substituted into the negative exponent, and the interference term is penalized. The gentle spatial gradient representing the real low-frequency attenuation of the light source is completely protected by the exponential model due to the absolute value of the Laplacian operator response of the extremely small pixel, and retains its original physical energy characteristics. Thus, extremely weak low-frequency exposure distortion features are purified from complex background noise, solving the problem of frequency domain aliasing.

[0039] Specifically, the reference density of the film is read from the production batch file. This reference density is obtained by averaging multiple physical samples of the defect-free main strip of the batch using an offline high-precision transmission spectrophotometer. Based on the obtained optical density matrix, a matrix is ​​constructed containing the center point and its eight surrounding adjacent points. The local spatial convolution mask is used to slide through the entire optical density matrix. The second-order gradient difference between the optical density value of each pixel and the optical density value of the corresponding adjacent pixels in the local spatial convolution mask is calculated to obtain the Laplacian operator response of each pixel.

[0040] Furthermore, by combining the Laplacian operator response of each pixel, the optical density value of each pixel, and the reference density, the drift potential energy of each pixel is calculated, and the drift potential energy of all pixels constitutes a complete drift potential energy field.

[0041] The formula for calculating the drift potential energy of the target pixel is:

[0042]

[0043] In the formula, The coordinates of the target pixel; The drift potential energy of the target pixel; It is the natural logarithm function; The optical density value of the target pixel; As the baseline density; It is an exponential function with the natural constant as its base; The absolute value of the Laplacian operator response at the target pixel.

[0044] In this calculation formula, the first term, deviation, characterizes the absolute magnitude of the deviation of the current point from the ideal exposure state. The further the optical density value of the pixel deviates from the reference density, the larger the value of this logarithmic term becomes, serving as a reference for signal strength. The second term, exponential decay, constitutes a physical shielding layer. The absolute value of the Laplacian operator response of the pixel acts as the independent variable. When the underlying data stream sweeps over film scratches or texture fluctuations, the drastic second-order micro-change causes the absolute value of the Laplacian operator response of the pixel to increase, resulting in a sharp decay of the exponential term approaching zero. This eliminates the potential energy contribution generated by the high-frequency structure and blocks the upward transmission of interference noise. When there is exposure distortion caused by the slow decay of the light source in a local area, its second derivative is extremely small, and the exponential term approaches 1, thus completely preserving the potential energy response of low-frequency drift and achieving precise removal of physical interference.

[0045] S3. Perform local sliding window aggregation calculation in the drift potential energy field that maps to the physical size of the underlying high-brightness uniform backlight. Extract the window non-uniformity based on the global dispersion of the drift potential energy of each pixel in the window and the ratio of the average potential energy to the maximum potential energy.

[0046] It should be noted that the aforementioned drift potential energy field has been decoupled from the continuous physical photometric distortion space. In order to map the microscopic and continuous physical field into industrial control commands, it is necessary to introduce statistical operators with spatiotemporal convergence characteristics. The real exposure defects in physical appearance simultaneously include large-area gradual diffusion and local extreme light decay collapse, which makes it impossible to accurately measure the overall defect degree using a single variance or extreme value. Therefore, this invention constructs a comprehensive evaluation system that integrates the overall discreteness and the logarithmic amplification of extreme values. The global variance of the drift potential energy field is used as the basic product term to quantify the physical extension range of defect diffusion. At the same time, the logarithmic operator is used to use the extreme value deviation as a nonlinear gain coefficient, so that the tiny local light physical collapse can generate a nonlinear gain response in the mathematical space. The discrete microscopic potential energy manifestation is upgraded and aggregated into a statistical evaluation index with macroscopic guiding significance, taking into account both the breadth and depth characteristics of defects.

[0047] Specifically, based on the physical projection area of ​​a single light-emitting array unit in the underlying high-brightness uniform backlight, a local sliding window for physical size mapping is set, and this local sliding window is used to perform sliding traversal in the drift potential energy field.

[0048] Furthermore, for a local sliding window that has been slid to any position, the total number of pixels within the local sliding window is counted, the drift potential energy of all pixels within the local sliding window is extracted, the average potential energy and the maximum potential energy of the window corresponding to the local sliding window are calculated, and then the non-uniformity of the window corresponding to the local sliding window is calculated.

[0049] The formula for calculating window non-uniformity is:

[0050]

[0051] In the formula, For window non-uniformity; This represents the total number of pixels in the window. For the traversal index of the pixel; For the first partial sliding window The drift potential energy of each pixel; The average potential energy of the window; It is the natural logarithm function; This represents the maximum potential energy of the window.

[0052] It should be noted that when the actual optical density of all pixels within this local sliding window... All are equal to the ideal reference density In the calculation results of the pixel's drift potential energy This leads to the drift potential energy of each pixel within the local sliding window being equal to 0. At this point, the average potential energy of the window is... A value of 0 represents an ideal physical state: within this sliding window area, the light source experiences no attenuation, the exposure is uniform, and the film is free from interference noise. In this case, the window non-uniformity is equal to 0.

[0053] The calculation formula captures the overall discrete aggregation degree of the drift potential energy of pixels within a local sliding window as it spatially distributes through the forward discreteness term, reflecting the gradual non-uniformity phenomenon over a large area. The value of this term increases linearly as the fluctuation amplitude of the drift potential energy of pixels increases. At the same time, the backward logarithmic term forms an extreme distortion amplifier. When there is an extremely severe local light source failure dead zone within the local sliding window, causing the maximum potential energy of the window to rise abnormally, the maximum potential energy of the window is much greater than the average potential energy of the window. The logarithmic term produces a significant nonlinear gain amplification effect. After multiplying this gain term with the aforementioned physical reference multiplier, it increases the final response value of the window non-uniformity in the underlying data stream. This multiplicative coupling mechanism ensures that the window non-uniformity can provide a sensitive abnormal response, whether it is a slow overall deterioration or a sudden local collapse.

[0054] S4. Combine the window non-uniformity with the actual linear velocity of the film obtained from the winding drive system controller, calculate the compensation voltage increment, and send the compensation voltage increment to the lighting control driver to perform closed-loop correction.

[0055] It should be noted that if a rigid fixed voltage is directly applied for compensation after confirming an exposure anomaly, it will cause severe light intensity fluctuations and over-compensation problems because the physical dynamic integration law in the high-speed transmission process is ignored. More importantly, the actual linear velocity of the film directly determines its integration time for receiving exposure energy under the underlying high-brightness uniform backlight. The higher the actual linear velocity of the film, the more sensitive the demand for supplementary light. Since the sine function has a dynamic easing characteristic of monotonically increasing and derivative decaying in the first quadrant, this invention constructs a feedforward closed-loop compensation model that integrates the dynamic characteristics of film operation. The window non-uniformity and the actual linear velocity of the film are substituted into the dynamic equation based on sinusoidal feedforward. The actual linear velocity of the film is mapped as the phase variable of the sine operator, so that the output of the compensation voltage increment is mathematically damped to prevent overexposure during the low-speed operation phase of the production line. During the high-speed stable operation range, the damping effect is weakened and approaches full output, solving the dynamic control instability problem caused by fixed constant feedback and ensuring the dynamic stability of the production process.

[0056] Specifically, the actual linear speed of the film is read in real time from the winding drive system controller, and the absolute physical rated maximum linear speed of the drive motor as specified on the factory nameplate is obtained, as well as the voltage regulation step constant determined by the digital-to-analog converter of the underlying hardware of the lighting control driver, while setting alarm thresholds.

[0057] Furthermore, during system operation, the window non-uniformity of all local sliding window outputs is continuously monitored. When the window non-uniformity of any local sliding window is greater than or equal to the alarm threshold, an audible and visual alarm is triggered and the system is shut down. When the window non-uniformity is less than the alarm threshold, the system maintains a closed-loop correction mode and calculates the compensation voltage increment sent to the lighting control driver in the corresponding area.

[0058] The formula for calculating the compensation voltage increment is as follows:

[0059]

[0060] In the formula, To compensate for voltage increments; This is the voltage regulation step constant; For window non-uniformity; This is the alarm threshold; It is a sine function; Pi is a constant. This represents the actual linear velocity of the film. This is the absolute physical maximum linear velocity.

[0061] The calculation formula achieves flexible damping adjustment by using a proportional control term as a linear negative feedback proportional gain term, which characterizes the physical urgency of the current deviation from the limit state. When the window non-uniformity is closer to the alarm threshold, the output adjustment force is greater, while fine-tuning is maintained during low-level drift to avoid repeated oscillations of the system. Since the actual linear velocity of the film is rigidly limited by the absolute physical rated maximum linear velocity, the calculation formula uses the absolute physical rated maximum linear velocity as a constant to strictly normalize the dynamic working condition within the range of zero to one. At the same time, the speed feedforward compensation logic is embedded through a sine factor, and a dynamic damping mechanism is constructed by using the easing characteristics of the first quadrant of the sine function. When the film is running in the low-speed crawling stage, the suppression effect of the sine term makes the compensation action tend to be smooth and automatically suppresses the high-risk supplementary light command in the low-speed state to prevent overexposure. When the film is in a high-speed stable production state, the sine term approaches 1, thereby releasing the full amount of compensation voltage increment to counteract the energy deficit under extremely short exposure time, ensuring that the output compensation voltage increment matches the physical inertia of the equipment transmission, and realizing precise steady-state control across working conditions.

[0062] Furthermore, the lighting control driver adjusts the driving current of the light-emitting array of the underlying high-brightness uniform backlight according to the received compensation voltage increment, forming a closed-loop control of physical execution.

[0063] Further, regarding the manual calibration of the alarm threshold, during the continuous testing phase simulating different attenuation levels, multiple sets of independently increasing threshold variables were set. The number of compensation interventions issued by the system and the corresponding sampling yield of the offline film were recorded under each set of independent threshold variables. The independent threshold variable corresponding to the physical inflection point where the sampling yield first reaches the highest qualified standard, the number of compensation interventions issued by the system reaches a minimum value, and the lighting control driver is in a stable, unsaturated state is locked as the final core indicator of the alarm threshold.

[0064] This invention also discloses an online film exposure uniformity detection system based on image processing, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the online film exposure uniformity detection method based on image processing according to this invention.

[0065] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. An online detection method for film exposure uniformity based on image processing, characterized in that, include: The original transmission image of the film surface is acquired synchronously using an external synchronous pulse. Based on Beer-Lambert's optical absorption law and combined with the pre-acquired dark background grayscale and empty background grayscale, the real-time pixel grayscale value of each pixel in the original transmission image is mapped to the optical density value of each pixel to construct an optical density matrix. The Laplacian operator response of each pixel in the optical density matrix is ​​calculated. A negative exponential decay term is constructed using the absolute value of the Laplacian operator response of each pixel. The drift potential energy of each pixel is calculated by combining the deviation of the optical density value of each pixel from the pre-calibrated reference density. A drift potential energy field for separating high-frequency texture and low-frequency decay is constructed. In the drift potential energy field, a local sliding window aggregation calculation is performed that maps to the physical size of the underlying high-brightness uniform backlight. Based on the global dispersion of the drift potential energy of each pixel in the window, and the ratio of the average potential energy to the maximum potential energy of each pixel in the window, the non-uniformity of the window is extracted. By combining the window non-uniformity with the actual linear velocity of the film, the compensation voltage increment is calculated, and the compensation voltage increment is sent to the lighting control driver to perform closed-loop correction.

2. The online detection method for film exposure uniformity based on image processing according to claim 1, characterized in that, The method for obtaining the grayscale values ​​of the dark background and the empty background is as follows: During the system initialization phase, the underlying high-brightness uniform backlight is turned off, and the line scan camera is controlled to perform multiple continuous scans and take the average value in a completely dark state to obtain the grayscale of the dark background. Turn on the bottom high-brightness uniform backlight to its rated power. In an unloaded state without film obstruction, control the line scan camera to directly capture the bottom high-brightness uniform backlight multiple times and take the average value to obtain the unloaded background grayscale.

3. The online detection method for film exposure uniformity based on image processing according to claim 1, characterized in that, The method, based on Beer-Lambert's optical absorption law and combined with pre-acquired dark background grayscale and empty background grayscale, maps the real-time pixel grayscale values ​​of each pixel in the original transmission image to the optical density values ​​of each pixel, constructing an optical density matrix, including: ； In the formula, The coordinates of the target pixel; The optical density value of the target pixel; This is a commonly used logarithmic function with base 10; The real-time pixel grayscale value of the target pixel; Grayscale background; The background grayscale is empty; This is the function for finding the maximum value; 𝑞 is a preset minimum positive constant to prevent logarithmic overflow; Construct an optical density matrix containing the optical density values ​​of all pixels.

4. The online detection method for film exposure uniformity based on image processing according to claim 1, characterized in that, Calculating the Laplacian operator response of each pixel in the optical density matrix includes: Read the reference density of the film from the production batch file; Based on the obtained optical density matrix, a structure is constructed containing the center point and its eight neighboring points. Local spatial convolution mask; The local spatial convolution mask is used to slide through the entire optical density matrix. The second-order gradient difference between the optical density value of each pixel and the optical density value of the adjacent pixels in the local spatial convolution mask is calculated to obtain the Laplacian operator response of each pixel.

5. The online detection method for film exposure uniformity based on image processing according to claim 1, characterized in that, A negative exponential decay term is constructed using the absolute value of the Laplacian operator response of each pixel. Combined with the deviation of the optical density value of each pixel from a pre-calibrated reference density, the drift potential energy of each pixel is calculated. The formula is as follows: ； In the formula, The coordinates of the target pixel; The drift potential energy of the target pixel; It is the natural logarithm function; The optical density value of the target pixel; As the baseline density; It is an exponential function with the natural constant as its base; The absolute value of the Laplacian operator response at the target pixel.

6. The online detection method for film exposure uniformity based on image processing according to claim 1, characterized in that, The method for obtaining the local sliding window that maps to the physical size of the underlying high-brightness uniform backlight includes: Based on the physical projection area of ​​a single light-emitting array unit in the underlying high-brightness uniform backlight, a local sliding window for physical size mapping is set.

7. The online detection method for film exposure uniformity based on image processing according to claim 1, characterized in that, The formula for calculating the window non-uniformity is: ； In the formula, For window non-uniformity; This represents the total number of pixels in the window. For the traversal index of the pixel; For the first partial sliding window The drift potential energy of each pixel; The average potential energy of the window; It is the natural logarithm function; This represents the maximum potential energy of the window.

8. The online detection method for film exposure uniformity based on image processing according to claim 1, characterized in that, Combining the window non-uniformity with the actual linear velocity of the film, the compensation voltage increment is calculated, and the compensation voltage increment is sent to the lighting control driver to perform closed-loop correction, including: During system operation, the window non-uniformity of all local sliding window outputs is continuously monitored. When the window non-uniformity is less than the preset alarm threshold, the system maintains a closed-loop correction mode and calculates the compensation voltage increment sent to the lighting control driver in the corresponding area. The calculation formula is as follows: ； In the formula, To compensate for voltage increments; This is the voltage regulation step constant; For window non-uniformity; This is the alarm threshold; It is a sine function; Pi is a constant. This represents the actual linear velocity of the film. This is the absolute physical maximum linear velocity; The lighting control driver adjusts the driving current of the light-emitting array of the underlying high-brightness uniform backlight according to the received compensation voltage increment, forming a closed-loop control of physical execution.

9. The online detection method for film exposure uniformity based on image processing according to claim 1, characterized in that, The method further includes: During system operation, the window non-uniformity of all local sliding window outputs is continuously monitored. When the window non-uniformity of any local sliding window is greater than or equal to the preset alarm threshold, an audible and visual alarm is triggered and the system is shut down.

10. An online detection system for film exposure uniformity based on image processing, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement the online film exposure uniformity detection method based on image processing according to any one of claims 1-9.

Images