Runway visual range deviation evaluation method based on background light brightness and light fusion

Abstract

The embodiment of the invention provides a runway visual range deviation evaluation method based on background light brightness and light fusion. The method comprises the following steps: acquiring a brightness gradient mutation feature of a runway sidelight sequence in a continuous frame video in a V channel of an HSV color space; extracting a preliminary regulation and control parameter of a light pulse duty ratio based on the brightness gradient mutation feature; dynamically correcting the preliminary regulation and control parameters in combination with a spatial Laplace change rate of background sky brightness; and performing frame-by-frame light pulse duty ratio regulation and control according to the corrected regulation and control parameters so as to compensate perceptual drift of human eyes to flickering light fusion frequency. Through the scheme of the embodiment of the invention, the problem that visual range interpretation is lagged due to the fact that human eyes perceive drifting of flickering light fusion frequency under low visibility can be solved.

Description

Technical Field

[0001] This application relates to the field of aviation safety assurance technology, specifically to a method for assessing runway visual range deviation based on the fusion of background brightness and lighting. Background Technology

[0002] In the field of existing aviation navigation and airport navigation lighting technologies, accurate assessment of runway visual range (RVR) is crucial for ensuring flight safety under low visibility conditions. Currently, RVR acquisition mainly relies on point sensors such as forward-scattering or transmissive visibility meters. Although these devices can directly measure meteorological optical visual range, their measurement results reflect the local visibility conditions at the instrument's mounting point, making it difficult to comprehensively and in real-time characterize the actual visibility of the entire runway area, especially the navigation lighting sequence that is directly related to the pilot's visual perception. At the same time, existing navigation lighting dimming systems typically adjust brightness in steps based on preset visibility levels. Or, flicker frequency adjustment is an open-loop control method that fails to consider and compensate in real time for the direct impact of dynamic changes in complex background light (such as fog, haze, and twilight) on the visual perception characteristics of the human eye in low-visibility environments. Specifically, the critical fusion frequency of the human eye for flickering lights will drift with changes in the ambient background brightness, causing pilots to lag or make errors in interpreting the light sequence, which may affect their correct judgment of runway distance and attitude. Therefore, there is an urgent need for an evaluation and control method that can integrate the state of the lights themselves with the visual perception environment of the human eye in real time, so as to dynamically optimize the light signal and improve the accuracy and reliability of visibility judgment. Summary of the Invention

[0003] In view of this, the present disclosure provides a method for evaluating runway visibility deviation based on the fusion of background light intensity and lighting, which at least partially solves the problems existing in the prior art.

[0004] A method for assessing runway visibility deviation based on the fusion of background brightness and lighting includes:

[0005] Obtain the brightness gradient abrupt change characteristics of the runway side light sequence in the V channel of the HSV color space in a continuous frame video;

[0006] Preliminary control parameters for the duty cycle of the light pulse are extracted based on the aforementioned brightness gradient abrupt change characteristics;

[0007] The initial control parameters are dynamically corrected by taking into account the spatial Laplacian rate of change of the background sky brightness.

[0008] The duty cycle of the light pulse is adjusted frame by frame based on the corrected control parameters to compensate for the perceptual drift of the human eye in the fusion frequency of the flickering lights.

[0009] According to one embodiment, the preliminary control parameters for extracting the duty cycle of the light pulse based on the brightness gradient abrupt change feature include:

[0010] Statistically analyze the frequency of pulse on / off states within the region of abrupt changes in brightness gradient;

[0011] The frequency is normalized to the range of [0, 1] and combined with the average pulse frequency over a historical period;

[0012] Weighting coefficients α and β are set to distinguish response priorities in different scenarios;

[0013] The initial duty cycle is calculated using the following formula: DP = α × (f / f_avg) + β × (m / m_total), where f is the pulse frequency of the current frame, f_avg is the historical average frequency, m is the number of abrupt change points, and m_total is the total number of pixels in the current frame.

[0014] According to one embodiment, dynamically correcting the initial control parameters by incorporating the spatial Laplacian rate of change of background sky brightness includes:

[0015] Extract the background sky region from consecutive frame videos and perform grayscale processing;

[0016] Calculate the Laplacian rate of change λ for the background image of each frame, denoted as: ,in It is the second derivative, and N is the total number of pixels;

[0017] An adjustment coefficient k is introduced to represent the relationship between the rate of change of background brightness and the perceived drift of the human eye;

[0018] The correction is performed using the following formula: DP_corrected = DP × e^(kλ), which is used to reduce the impact of background fluctuations on pulse duty cycle control.

[0019] According to one embodiment, the step of implementing frame-by-frame light pulse duty cycle adjustment based on the corrected control parameters to compensate for the perceptual drift of the human eye to the fusion frequency of flickering lights includes:

[0020] Divide DP_corrected into two regions, high and low, and define them as DH and DL respectively;

[0021] The duty cycle output is adjusted using linear interpolation within the DH region;

[0022] Set the dynamic offset δ(t) = a × t + b, where t represents the current time step, and a and b are the scaling and constant factors, respectively.

[0023] The actual output duty cycle DP_output = DP_corrected + δ(t) is adjusted to adapt to changes in human visual adaptation.

[0024] According to one embodiment, obtaining the brightness gradient abrupt change features of the runway side light sequence in the HSV color space V channel in consecutive frame video includes:

[0025] A multi-frame accumulation method is used to perform temporal analysis on consecutive frame images to construct a video motion vector field;

[0026] Hough transform was used to detect the contour of side lights and their brightness distribution trend.

[0027] For each frame, calculate the difference between the maximum and minimum brightness values ​​of the side lights, Max_V and Min_V.

[0028] Calculate the relative mutation index μ = (Max_V-Min_V) / V_avg. If μ > θ, where θ is a pre-set threshold, then it is determined that there is a significant gradient mutation.

[0029] According to one embodiment, the preliminary control parameters for extracting the duty cycle of the light pulse based on the brightness gradient abrupt change feature include:

[0030] Group the pixels within the edge light area by row and analyze their brightness change sequence;

[0031] Calculate the standard deviation σ_i of the brightness fluctuation for each pixel;

[0032] By comparing σ_i with the background standard deviation σ_b, the brightness correlation parameter ω = σ_i / σ_b is constructed;

[0033] The initial duty cycle parameter DP_initial = ∑ω_i × f_i / ∑ω_i is generated for each frame using a weighted moving average method.

[0034] According to one embodiment, the preliminary control parameters for extracting the duty cycle of the light pulse based on the brightness gradient abrupt change feature include:

[0035] Support vector machines were used to model the brightness mutation characteristics and historical duty cycle control parameters;

[0036] Predict the expected pulse frequency f_pred for each frame using training data;

[0037] A dynamic feedback error E = |f_true-f_pred| is introduced as an adjustment weight parameter;

[0038] Update the control parameters according to the following formula: DP_new = DP_prev × (1-E / f_max), where f_max represents the maximum allowable frequency.

[0039] According to one embodiment, dynamically correcting the initial control parameters by incorporating the spatial Laplacian rate of change of background sky brightness includes:

[0040] Set a reference window in the background sky area and record the average brightness M_sky within the window;

[0041] Compare the current frame brightness M_current with M_sky and calculate the brightness deviation ΔM = |M_current-M_sky|;

[0042] Multiply ΔM by its time derivative term to obtain the brightness change rate factor;

[0043] The corrected control parameter is: DP_robust = DP_initial × (1-ΔM × [d(ΔM) / dt] / σ_sky), where σ_sky represents the standard deviation of background brightness.

[0044] According to one embodiment, the step of implementing frame-by-frame light pulse duty cycle adjustment based on the corrected control parameters to compensate for the perceptual drift of the human eye to the fusion frequency of flickering lights includes:

[0045] The target duty cycle reference D_ref is calculated based on the ambient light intensity L_env;

[0046] The actual duty cycle change rate is adjusted according to the human eye adaptation function H(τ) = 1 e^{τ / τ_c}, where τ is a time variable and τ_c represents the adaptation time constant;

[0047] Introducing a duty cycle buffer factor b enables a smooth transition in regulation;

[0048] The final output duty cycle D_final = D_ref × H(τ) + b × DP_robust ensures dynamic stability.

[0049] This disclosure provides a method for evaluating runway visibility deviation based on background brightness and light fusion, including: acquiring the brightness gradient abrupt change characteristics of the runway side light sequence in the HSV color space V channel of a continuous frame video; extracting preliminary adjustment parameters for the light pulse duty cycle based on the brightness gradient abrupt change characteristics; dynamically correcting the preliminary adjustment parameters by combining the spatial Laplacian rate of change of the background sky brightness; and implementing frame-by-frame light pulse duty cycle adjustment according to the corrected adjustment parameters to compensate for the perceptual drift of the human eye to the fusion frequency of flickering lights. The solution of this disclosure can solve the problem of how to implement frame-by-frame adjustment of the light pulse duty cycle based on the brightness gradient abrupt change characteristics of the runway side light sequence in the HSV color space V channel of a continuous frame video, combined with the spatial Laplacian rate of change of the background sky brightness, to solve the problem of lag in visibility judgment caused by the perceptual drift of the human eye to the fusion frequency of flickering lights under low visibility conditions. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the exemplary embodiments of this disclosure, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of this disclosure and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart of the runway visibility deviation assessment method based on the fusion of background brightness and lighting, according to the present invention.

[0052] Figure 2 This is a flowchart of the preliminary control parameters for the duty cycle of a light pulse extracted based on the abrupt change in brightness gradient features, according to the present invention.

[0053] Figure 3 This is a flowchart of the present invention that dynamically corrects the initial control parameters by combining the spatial Laplace rate of change of background sky brightness;

[0054] Figure 4 This is a flowchart of another embodiment of the present invention for obtaining the brightness gradient abrupt change characteristics of the runway side light sequence in the V channel of the HSV color space in a continuous frame video.

[0055] Figure 5 This is a flowchart of another embodiment of the present invention for extracting preliminary control parameters of the light pulse duty cycle based on the brightness gradient abrupt change features;

[0056] Figure 6 This is a flowchart illustrating another embodiment of the present invention, which describes the implementation of frame-by-frame light pulse duty cycle control based on modified control parameters to compensate for the perceptual drift of the human eye to the fusion frequency of flickering lights. Detailed Implementation

[0057] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0058] Next, refer to Figure 1 The present invention describes a runway visibility deviation assessment method based on the fusion of background brightness and lighting, comprising:

[0059] Step S101: Obtain the brightness gradient abrupt change features of the runway side light sequence in the V channel of the HSV color space in the continuous frame video.

[0060] First, high dynamic range (HLR) cameras deployed at specific locations on the runway (such as the control tower or near the runway centerline) continuously capture video streams containing sequences of runway edge lights. The video stream is input to the processing system at a rate of N frames per second (e.g., N=25 or 30). Instead of processing in the raw RGB color space, the system immediately converts each frame from the RGB color space to the HSV (Hue, Saturation, Lightness) color space. The core reason for choosing the HSV space is that its V (Value, Lightness) channel directly and independently represents the brightness information of a pixel, is highly linearly correlated with the human eye's subjective perception of brightness, and effectively separates the interference of color information (H and S) on brightness analysis.

[0061] After conversion, the system focuses on analyzing the V channel image. To suppress interference such as image sensor noise and minor brightness fluctuations caused by atmospheric turbulence, Gaussian blur filtering is first applied to the V channel image of each frame, for example, by smoothing it using a 3x3 or 5x5 Gaussian convolution kernel.

[0062] Subsequently, to capture the boundaries of rapid brightness changes caused by pulse-driven (e.g., flashing on and off at a specific frequency, such as 3Hz) runway edge lights, the Sobel operator was applied to the smoothed V-channel image for gradient calculation. Specifically, the image was convolved using horizontal Sobel kernels (Gx) and vertical Sobel kernels (Gy) to obtain approximate brightness gradients for each pixel in the horizontal and vertical directions. Then, the gradient magnitude (or gradient intensity) of that pixel was calculated using the formula G = sqrt(Gx^2 + Gy^2), which characterizes the drasticness of the brightness change at that point.

[0063] To identify regions of significant brightness change caused by the instantaneous "on" or "off" of light pulses (i.e., "brightness gradient abrupt change features"), the system sets a gradient amplitude threshold γ. This threshold γ can be adaptively calibrated based on typical gradient values ​​of the runway edge light region in historical video data. For each pixel in the image, a gradient amplitude statistic (e.g., the difference between the maximum and minimum values, or the standard deviation) is calculated within its surrounding neighborhood (e.g., a 3x3 neighborhood). When this statistic exceeds the preset threshold γ, the pixel or its neighborhood is identified as a "brightness gradient abrupt change point." By scanning the entire frame or a predefined region of interest (i.e., the approximate area where the runway edge lights are located), a brightness gradient abrupt change matrix composed of all abrupt change points can be obtained. This matrix quantifies the spatial brightness change pattern caused by the light pulses in the current frame.

[0064] Step S102: Extract preliminary control parameters for the duty cycle of the light pulse based on the brightness gradient abrupt change features.

[0065] The goal of this step is to transform the spatial brightness variation pattern obtained in step S101 into a preliminary parameter (DP) for adjusting the time proportion (i.e., duty cycle) of the next cycle's light pulse "on". First, within the region comprised of identified brightness gradient abrupt change points, the brightness values ​​at these points are analyzed as a sequence of changes over time (i.e., across consecutive frames). By detecting the jumps in brightness values ​​between high and low levels, the currently observed light pulse frequency f can be statistically determined. Simultaneously, the system maintains the average pulse frequency f_avg over a historical time window (e.g., the past 10 seconds) as a baseline reference for perceiving drift.

[0066] Secondly, the number of brightness gradient abrupt change points m in the current frame is calculated and compared with the total number of pixels m_total in the current frame image or region of interest to obtain a proportion representing the relative size of the brightness gradient abrupt change region.

[0067] Finally, two weighting coefficients, α and β, are set. α assigns importance to the current instantaneous pulse frequency f relative to the historical average f_avg, used for rapid response to instantaneous changes in pulse frequency; β assigns weight to the size of the brightness abrupt change region, used to adjust the control intensity when the light is partially obscured (e.g., by non-uniform fog distribution). The initial duty cycle control parameter DP is calculated using the formula DP = α × (f / f_avg) + β × (m / m_total). For example, when the observation frequency f is higher than the historical average f_avg, (f / f_avg) > 1, and the initial parameter DP tends to increase the duty cycle to attempt to compensate for the possible increase in human visual frequency perception due to ambient brightness, thereby maintaining the subjective flickering sensation of the light.

[0068] In one implementation, the specific process of extracting preliminary control parameters based on the brightness gradient abrupt change features is as follows: First, the system aligns the connected regions where the identified brightness gradient abrupt change points are located in the time dimension to form a brightness change time series. Next, the standard deviation σ_i of this series at each pixel is calculated to quantify the severity of the brightness fluctuation at that point. Simultaneously, the standard deviation σ_b of the brightness values ​​in the non-light-lit background sky region of the video frame is calculated as a benchmark for inherent environmental fluctuations. Subsequently, a brightness correlation parameter ω = σ_i / σ_b is constructed for each abrupt change pixel. This parameter characterizes the significance of the fluctuation at that point relative to the environmental background; a larger ω value indicates that the brightness change at that point is more likely to originate from active light pulses rather than environmental noise. Finally, the system weights all abrupt changes in the current frame according to their ω_i values ​​and combines this with their instantaneous brightness state (e.g., the normalized brightness value f_i, where high brightness corresponds to f_i close to 1 and low brightness corresponds to f_i close to 0). The system then uses the weighted moving average formula DP_initial = Σ(ω_i × f_i) / Σω_i to calculate the initial duty cycle parameter DP_initial. This method automatically filters out feature points strongly correlated with the light pulse through the weight ω, making the extraction of the initial parameter more resistant to environmental interference.

[0069] In another implementation, the initial control parameters are extracted using a prediction method based on a machine learning model. Specifically, the system pre-collects a large amount of historical video data and its corresponding optimal lighting control parameters to form a training dataset. The model uses the brightness gradient abrupt change features of the previous and current frames (such as the spatial distribution density of abrupt change points, average gradient magnitude, and area ratio of abrupt change regions) as input feature vectors. During training, a support vector machine regression model is used to learn the mapping relationship from the input features to the desired pulse frequency f_pred. In online applications, for real-time video streams, the system first extracts the brightness gradient abrupt change feature vector of the current frame, inputs it into the pre-trained SVR model, and directly predicts the desired pulse frequency f_pred for the current frame. Then, a dynamic feedback mechanism is introduced: the frequency f_pred predicted by the model is compared with the actual instantaneous pulse frequency f_true detected through inter-frame analysis, and the absolute error E = |f_true - f_pred| is calculated. This error E reflects the reliability of the model's prediction in the current environment. Finally, the initial control parameters are updated according to the formula DP_new = DP_prev × (1 - E / f_max), where DP_prev is the control parameter of the previous frame and f_max is the maximum frequency allowed by the system. This embodiment utilizes a model to learn complex mappings from historical data and can self-correct through real-time errors, making it more adaptable.

[0070] Step S103: Dynamically correct the preliminary control parameters by combining the spatial Laplace rate of change of the background sky brightness.

[0071] The human eye's perception of flicker is influenced not only by the light source itself but also, and more significantly, by the overall brightness of the background environment and its rate of change. To quantify this background effect, the system segments the background sky region in each frame (typically located at the top of the image, determined by color features or a pre-defined mask). This region is then converted to grayscale, and its spatial Laplacian rate of change λ is calculated. The Laplacian operator (…) () is a second-order differential operator used to measure the second-order changes or curvature of an image. It is very sensitive to rapid spatial changes in intensity in an image (such as cloud edges or brightness gradients).

[0072] In the specific calculation, a Laplacian convolution kernel is applied to the grayscale image of the background sky to obtain the Laplacian response value of each pixel. Then, the absolute values ​​of the response values ​​of all pixels are taken and summed, and then divided by the total number of pixels N to obtain the spatial Laplacian rate of change of the sky brightness of the background in that frame. The larger the λ value, the more uneven and drastic the spatial distribution of background brightness (e.g., crepuscular light, moving fog edges). This dynamic background noise can interfere with the human eye's stable perception of foreground light.

[0073] An adjustment coefficient k is introduced, pre-determined through psychophysical experiments, characterizing the empirical relationship between the spatial rate of change of background brightness λ and the perceptual drift rate of the human eye in response to flickering light. An exponential function is used to correct the initial parameter DP: DP_corrected = DP × e^(kλ). When the background is stable (λ≈0), the correction coefficient is approximately 1, having no effect. When the background fluctuates drastically (λ is large), if k is negative (indicating background interference causing a decrease in perceptual sensitivity), then e^(kλ) < 1, thus lowering DP_corrected. The system will tend to use a more conservative (e.g., slightly lower) duty cycle to counteract background interference and avoid over-adjustment. This step ensures the adaptability of the control strategy to complex ambient light dynamics.

[0074] Step S104: Implement frame-by-frame light pulse duty cycle adjustment according to the corrected control parameters to compensate for the perceptual drift of the human eye to the fusion frequency of the flashing lights.

[0075] The corrected control parameter DP_corrected is sent to the lighting control system. To achieve smooth, abrupt control and further simulate the adaptation process of human vision, DP_corrected undergoes final processing.

[0076] First, the possible range of DP_corrected values ​​is divided into two regions: a high duty cycle region DH (e.g., 0.7–1.0) and a low duty cycle region DL (e.g., 0–0.7). Different interpolation strategies are used in different regions, such as linear mapping in the DL region and non-linear (e.g., logarithmic) mapping in the DH region, to finely adjust the subjective perception under high brightness.

[0077] Secondly, a dynamic offset δ(t) is introduced. This is a function that changes slowly over time, for example, set as δ(t) = a × t + b, where t is the current system operating time step or cumulative time, and a and b are scaling factors and constants set according to the adaptation characteristics of the human eye. This offset is used to simulate and compensate for the slow adaptation effect produced by the human eye when exposed to specific light environments for a long time.

[0078] Finally, the duty cycle of the actual output light pulse is calculated: DP_output = [interpolation result based on DP_corrected] + δ(t). This DP_output value is sent in real time to the runway edge light dimming controller via a fieldbus (such as DMX512) or a dedicated dimming interface. The controller adjusts the duration of the "on" state in the next one or several pulse cycles based on the DP_output value.

[0079] Through frame-by-frame, closed-loop execution of the above four steps, the method of this invention can perceive the state of the light signal itself (steps S101 and S102) and its dynamic optical environment in real time (step S103), and generate the optimal light control command by combining the two (step S104). This essentially endows the traditional navigation lighting system with "visual-feedback" capability, enabling its output light signal to actively adapt to environmental changes and the physiological characteristics of the human eye, thereby effectively stabilizing the pilot's perception of the runway edge light flashing frequency under low visibility conditions, reducing interpretation lag, and improving the accuracy of vision-based runway visual range assessment and flight safety margin.

[0080] In one embodiment of the present invention, obtaining the brightness gradient abrupt change characteristics of the runway side light sequence in the V channel of the HSV color space in a continuous frame video may further include the following steps.

[0081] First, each frame of the video is Gaussian blurred to eliminate noise interference. This step aims to reduce false detections caused by sensor noise or other interference by smoothing the image. The parameter of the Gaussian filter is typically the standard deviation σ, ranging from 0.5 to 3, where σ=1 provides good denoising for minute details in the image while preserving edge information. For example, in runway visual range assessment, airport lighting images may contain noise due to uneven ambient lighting; applying Gaussian blur in this case helps improve the accuracy of subsequent brightness gradient analysis.

[0082] Secondly, the Sobel operator is used to calculate the horizontal and vertical brightness gradients in the V channel. The Sobel operator is an edge detection method, and its formula is: This method is used to detect the rate of change of brightness of pixels in an image along the x and y directions. This process can capture the local characteristics of brightness changes, making it easier to identify edges or boundaries. For example, in the edge lighting area of ​​an airport runway, the boundary between the light and the surrounding background can be distinguished by calculating the gradient at different pixel locations.

[0083] Finally, the direction and intensity of brightness changes are determined based on the magnitude of the brightness gradient, and a brightness gradient abrupt change matrix is ​​established. This matrix can characterize the brightness change trend in each region of the image, thereby revealing the distribution and stability information of the edge light sequence. A gradient threshold γ is set; when the gradient difference between adjacent regions of a pixel exceeds γ, it is determined to be a brightness gradient abrupt change point. The setting of this threshold γ usually needs to be combined with the difference between the image background brightness and the actual scene brightness, generally ranging from 5 to 20, with an optimal value of 12. For example, during nighttime operation, if the brightness fluctuation of the environment around the runway is large, appropriately increasing the value of γ can avoid misjudging abrupt changes in non-critical areas.

[0084] This technical solution can effectively identify abnormal brightness in the runway edge light sequence caused by ambient light interference or equipment failure, enhance the reliability and real-time performance of runway visual range assessment, and provide important basis for flight safety.

[0085] Next, refer to Figure 2 The present invention describes the steps for extracting preliminary control parameters of the light pulse duty cycle based on the abrupt change in brightness gradient features, specifically including:

[0086] S201: Statistically count the frequency of pulse on / off in the region of abrupt changes in brightness gradient;

[0087] The frequency is normalized to the range of [0, 1] and combined with the average pulse frequency over a historical period;

[0088] S202: Set weighting coefficients α and β to distinguish response priorities in different scenarios;

[0089] S203: Calculate the initial duty cycle according to the following formula: DP = α × (f / f_avg) + β × (m / m_total), where f is the pulse frequency of the current frame, f_avg is the historical average frequency, m is the number of abrupt change points, and m_total is the total number of pixels in the current frame.

[0090] After obtaining the brightness gradient abrupt change matrix, the system focuses on the pixel regions marked as abrupt change points for temporal analysis. For each pixel in this region, its V channel brightness value is tracked in consecutive frames (e.g., the last 5 frames). By detecting the number of brightness transitions between preset high and low thresholds, the instantaneous flicker frequency f of the light pulse in this region in the current frame is calculated. Simultaneously, the system maintains a pulse frequency sequence within a sliding time window (e.g., the past 30 seconds) and calculates its arithmetic mean f_avg as a baseline for perceptual drift. The parameter m represents the total number of brightness gradient abrupt change points in the current frame, and m_total is the total number of pixels in the current frame analysis area set by the system (usually the rectangular sensing area where the runway edge lights are located). Weighting coefficients α and β are used to distinguish the response priority under different scenarios: α is usually set to a larger value (e.g., 0.7), giving the current frequency f a higher response priority, enabling the system to respond quickly to sudden changes in visibility; β is set to a smaller value (e.g., 0.3), used to smooth abnormal fluctuations in the number of abrupt change points m caused by image noise or local occlusion (e.g., a bird briefly flying by). Finally, the initial duty cycle DP is calculated using the formula DP = α × (f / f_avg) + β × (m / m_total). For example, when decreased visibility causes the actual observation frequency f to drop below the historical average f_avg, (f / f_avg) < 1. This formula will output a lower DP value, thus tending to increase the pulse duty cycle (i.e., increase the proportion of "bright" time) in subsequent adjustments to compensate for the reduced fusion frequency perception of the human eye due to environmental darkness, thereby maintaining the necessary flashing warning effect.

[0091] Next, refer to Figure 3 The present invention describes the steps of dynamically correcting preliminary control parameters by incorporating the spatial Laplace rate of change of background sky brightness, which include:

[0092] S301: Extract the background sky region from a continuous frame video and perform grayscale processing;

[0093] S302: Calculate the Laplacian rate of change λ for the background image of each frame, denoted as... ,in It is the second derivative, and N is the total number of pixels;

[0094] S303: Introduce the adjustment coefficient k to represent the relationship between the rate of change of background brightness and the perceived drift of the human eye;

[0095] S304: Corrected according to the following formula: DP_corrected = DP × e^(kλ), used to reduce the impact of background fluctuations on pulse duty cycle control.

[0096] Specifically, in each frame of the image, the background sky region is accurately extracted and converted into a grayscale image I(x,y) by color thresholding (filtering out high-saturation, low-value sky regions) combined with edge detection. Then, a 3x3 Laplacian convolution kernel (with a center value of +4 and values ​​of -1 on all sides) is applied to the grayscale image to obtain the Laplacian response value for each pixel. This value reflects the degree of "bumping" or "unevenness" of the brightness at that point relative to its surrounding area. Next, the Laplace rate of change λ for the entire sky region is calculated using the formula: , where N is the total number of pixels in the sky region. The absolute values ​​of the intensity of changes in brightness space were summed and then divided by N to obtain the average intensity of change. The adjustment coefficient k is a key parameter calibrated experimentally: under typical dawn / dusk or rapidly moving fragmented cloud conditions, the impact of background fluctuations on human eye flicker perception was determined through subjective evaluation experiments. k is usually negative (e.g., k = -0.05), indicating that the more drastic the background spatial changes, the greater the interference with stable human eye perception, requiring more conservative control. The correction formula DP_corrected = DP × e^(kλ) utilizes the properties of an exponential function. When the background is stable and uniform (λ≈0), e^(kλ)≈1, and DP is almost unaffected; when the background fluctuates drastically (λ is large), e^(kλ) becomes less than 1, thus reducing DP_corrected compared to DP. This correction effectively suppresses potential malfunctions in light control caused by dynamic background noise such as the rapid movement of light through cloud gaps.

[0097] In one embodiment of the present invention, the frame-by-frame light pulse duty cycle control based on the modified control parameters specifically includes:

[0098] Divide DP_corrected into two regions, high and low, and define them as DH and DL respectively;

[0099] The duty cycle output is adjusted using linear interpolation within the DH region;

[0100] Set the dynamic offset δ(t) = a × t + b, where t represents the current time step, and a and b are the scaling and constant factors, respectively.

[0101] The actual output duty cycle DP_output = DP_corrected + δ(t) is adjusted to adapt to changes in human visual adaptation.

[0102] Specifically, the corrected duty cycle parameter DP_corrected (with a value range of [0, 1]) is input into the final output mapping module. First, a segmentation threshold D_th (e.g., 0.65) is set to divide the parameter domain into a high duty cycle region DH (D_th, 1.0) and a low duty cycle region DL [0, D_th]. Within the DL region, linear mapping is used for output, i.e., output duty cycle = DP_corrected; within the DH region, considering that the human eye is more sensitive to duty cycle changes under high brightness, linear interpolation is used for adjustment, for example, output duty cycle = D_th + (DP_corrected - D_th). S, where S is a scaling factor (e.g., S=0.8), makes the changes in the high zone smoother. Simultaneously, to simulate and compensate for visual adaptations (such as dark or light adaptation) that occur during prolonged viewing, the system introduces a dynamic offset δ(t) that changes slowly over time. Its calculation formula is δ(t) = a × t + b, where t is the cumulative time (in minutes) since the system's self-regulation began, a is a scaling factor (e.g., -0.01 / minute, indicating a slow adjustment of the baseline over time), and b is the initial offset constant (e.g., +0.02). Finally, the actual output duty cycle sent to the lighting controller is DP_output = the segmented mapped value + δ(t). This value is limited to a safe range (e.g., 0.1 to 0.95) and then transmitted in real-time via a communication protocol, ensuring that the lighting system's output can simultaneously respond to immediate environmental changes and the slow physiological adaptation process of the human eye.

[0103] Next, refer to Figure 4 The following describes a step in obtaining the brightness gradient abrupt change characteristics of a runway side light sequence in the V channel of the HSV color space in a continuous frame video, according to another embodiment of the present invention:

[0104] S401: Employs a multi-frame accumulation method to perform temporal analysis on consecutive frame images and constructs a video motion vector field;

[0105] S402: Detecting the contour and brightness distribution trend of side lights using Hough transform;

[0106] S403: For each frame, calculate the difference between the maximum and minimum brightness values ​​of the side lights, Max_V and Min_V.

[0107] S404: Calculate the relative mutation index μ = (Max_V-Min_V) / V_avg. If μ > θ, where θ is a pre-set threshold, then it is determined that there is a significant gradient mutation.

[0108] To enhance the robustness of detection, this embodiment employs a multi-frame information fusion method. The system performs temporal analysis on the V-channel images of T consecutive frames (e.g., T=3), constructing a video motion vector field using optical flow or block matching methods to distinguish between real light flickering and brightness changes caused by moving objects (e.g., airplanes). Next, in each frame, Hough transform is used to detect straight lines or preset geometric arrangements of light groups, accurately identifying and locating the contour regions of runway edge lights and eliminating interference from other similar light sources. Within each light contour region, the maximum value (Max_V) and minimum value (Min_V) of all pixel V values ​​within that frame are calculated, and the average brightness of the region (V_avg) is calculated. Then, the relative abrupt change index μ = (Max_V - Min_V) / V_avg is calculated. This index μ reflects the relative contrast of brightness within a single light region. The system sets an empirical threshold θ (e.g., θ=0.6). When μ > θ, it is determined that the light has a significant brightness gradient abrupt change within the current frame, which typically corresponds to the transition edge of a light pulse (the instant of turning on or off). By statistically analyzing the proportion of all side lights in a frame that satisfy μ > θ and fusing it with the spatial gradient detection results, a more reliable and accurate comprehensive index describing the brightness abrupt changes of the entire runway side light sequence can be constructed, providing higher quality input for subsequent parameter extraction.

[0109] Next, refer to Figure 5 The following describes the steps of extracting preliminary control parameters of the light pulse duty cycle based on the brightness gradient abrupt change features according to another embodiment of the present invention, specifically including:

[0110] S501: Group the pixels within the edge light area by row and analyze their brightness change sequence;

[0111] S502: Calculate the standard deviation σ_i of the brightness fluctuation of each pixel;

[0112] S503: Compare σ_i with the background standard deviation σ_b to construct the brightness correlation parameter ω = σ_i / σ_b;

[0113] S504: The initial duty cycle parameter DP_initial = ∑ω_i× f_i / ∑ω_i is generated for each frame using the weighted moving average method.

[0114] First, the pixels within each identified runway edge light contour region (such as a circular or rectangular light spot) are grouped by image rows (or columns). For each group of pixels, the brightness value sequence in consecutive frames (e.g., the most recent 8 frames) is tracked, and a thresholding method is used to determine whether each pixel is in a "bright" (1) or "off" (0) state in each frame, thereby constructing a two-dimensional brightness / off state matrix for each group.

[0115] Next, the standard deviation σ_i of the brightness fluctuation of each pixel within the group over its own brightness time series is calculated. This σ_i value quantifies the dispersion of the brightness change of a single pixel over time: for ideal, clear flickering light, its σ_i will be high; if the light is uniformly attenuated by a thin fog, σ_i will be low; if the point is actually a background pixel, then σ_i will be low. Simultaneously, in a pure background sky region far from all lights, the standard deviation σ_b of the brightness values ​​of all pixels in this region within the same time window is calculated as a baseline level for environmental background noise.

[0116] Then, a brightness correlation parameter ω_i is constructed for each pixel, with the formula ω_i = σ_i / σ_b. This parameter is a key filter: when σ_i is much greater than σ_b (ω_i >> 1), it indicates that the brightness fluctuation of this point is significantly stronger than the ambient background noise, and it is very likely to be a valid light signal point; conversely, if ω_i is close to or less than 1, the point is likely noise or static background, and its data should be suppressed. Finally, the initial duty cycle parameter for the current frame is generated using a weighted moving average method. The calculation formula is DP_initial = Σ (ω_i × f_i) / Σ ω_i. Where f_i is the normalized brightness state of the pixel in the current frame (for example, if the brightness is greater than the threshold, then f_i = 1; otherwise, f_i = 0). The essence of this formula is to use the correlation parameter ω_i as the weight to perform a weighted average of the instantaneous states of all suspected signal points. The weight ω_i automatically assigns higher weight to strong signal points while weakening the contribution of noise points. This makes the extracted DP_initial parameters more robust to interference such as uneven distribution of fog and haze and slight camera shake, and can more realistically reflect the collective state of light pulses.

[0117] In another embodiment, offline model training is first performed. The system collects a large number of historical video data segments covering different weather conditions (sunny, foggy, rainy, hazy) and different time periods (day, night, dusk), and records the optimal (usually evaluated by experts or inferred from visibility meter data) light pulse duty cycle control parameters corresponding to each data segment, forming a labeled dataset. For each frame of training data, a set of multi-dimensional brightness mutation feature vectors are extracted as input to the model, according to the method described in claim 10. The feature vectors mainly include: 1) the mean of wavelet high-frequency energy; 2) the proportion of edge light contours with significant mutations to the total number of contours; 3) the average color consistency rate change value ΔCR of all mutated contours; 4) the average amplitude of the brightness gradient within the mutated contour region. Using these features and the corresponding historical optimal expected pulse frequency f_pred_label as output, a support vector machine regression model is trained. A radial basis function is used as the kernel function, and the penalty parameter C and kernel coefficient gamma are optimized through grid search and cross-validation, enabling the model to learn the nonlinear mapping relationship between complex visual features and ideal frequencies.

[0118] During the online application phase, for real-time input video frames, the system extracts the same multidimensional brightness abrupt change feature vector as described above and inputs it into the pre-trained SVR model. The model directly outputs a predicted expected pulse frequency value f_pred for the current frame. This predicted value incorporates historical experience and represents the pulse frequency that, theoretically, best maintains an effective visual alert under the current visual characteristics.

[0119] Next, a dynamic feedback error stage is introduced. The system uses independent time-domain signal analysis methods (such as Fourier transform of the brightness sequence of a specific light area) to measure the actual light pulse frequency f_true in the current frame. The absolute error between the predicted and measured values ​​is calculated as E = |f_true - f_pred|. This error E has a dual meaning: on the one hand, it reflects the degree of difference between the current environmental conditions and the historical conditions of the model training; on the other hand, it also reflects the reliability of the model's prediction in the current scene. The larger the error E, the lower the reliability of the model's prediction, and the more the system should rely on the adjustment results of the previous frame for minor adjustments, rather than trusting the current prediction.

[0120] Finally, the control parameters are cautiously updated based on the feedback error. The update formula is DP_new = DP_prev × (1 - E / f_max). Here, DP_prev is the control parameter used in the previous frame, and f_max is the maximum safe pulse frequency allowed by the system (e.g., 5Hz, to avoid discomfort caused by high-frequency flicker). This formula achieves a smooth and robust recursive update: when the prediction error E is small, (1 - E / f_max) is close to 1, DP_new is close to DP_prev, and the system remains stable; when the prediction error E increases, (1 - E / f_max) decreases, and DP_new will be adjusted downwards by a certain proportion based on DP_prev. This is a conservative strategy to prevent aggressive adjustments that could lead to system instability when model predictions are unreliable. This approach, combining machine learning prediction with real-time feedback error correction, significantly improves the system's adaptability and control robustness under unseen and atypical weather conditions.

[0121] Next, we describe another embodiment of the present invention, which describes the step of dynamically correcting the initial control parameters by incorporating the spatial Laplacian rate of change of the background sky brightness, including:

[0122] Set a reference window in the background sky area and record the average brightness M_sky within the window;

[0123] Compare the current frame brightness M_current with M_sky and calculate the brightness deviation ΔM = |M_current-M_sky|;

[0124] Multiply ΔM by its time derivative term to obtain the brightness change rate factor;

[0125] The corrected control parameter is: DP_robust = DP_initial × (1-ΔM × [d(ΔM) / dt] / σ_sky), where σ_sky represents the standard deviation of background brightness.

[0126] First, the average brightness values ​​of these reference windows in the initial calibration frame or the previous period are recorded as the reference brightness mean M_sky. For the current processing frame, the average brightness M_current within the same reference window is calculated. Next, the absolute brightness deviation ΔM = |M_current - M_sky| is calculated, which directly reflects the magnitude of change in the overall background brightness.

[0127] To capture the dynamic characteristics of the changes, the temporal derivative of the brightness deviation is further calculated. The system records the M_current sequence of the most recent frames (e.g., 5 frames) and estimates the current brightness change rate using numerical differentiation (e.g., the central difference method), expressed as d(ΔM) / dt. Multiplying the absolute deviation ΔM by its change rate d(ΔM) / dt yields a brightness change rate factor R = ΔM × [d(ΔM) / dt]. This factor R combines the magnitude and speed of change: large, rapid changes (such as the sun suddenly being obscured by thick clouds) will produce a large R value, indicating that the background light environment is changing drastically, causing the greatest interference to human visual perception.

[0128] Finally, this factor is used to robustly correct the initial parameters. The correction formula is DP_robust = DP_initial × (1 - R / σ_sky), where σ_sky is the standard deviation of the brightness of the background reference window over the statistical time period, used to normalize R. When the background changes drastically (large R value), the correction term (1 - R / σ_sky) decreases, thus actively lowering the duty cycle parameter DP_robust. The logic is that in a chaotic environment with rapidly changing background light, overly aggressive adjustment of the light duty cycle may be counterproductive and exacerbate visual confusion. In this case, the system adopts a more conservative and gradual control strategy, waiting for the environment to stabilize before making fine adjustments, thereby improving the overall stability and reliability of the system under complex weather conditions.

[0129] Next, refer to Figure 6 The present invention describes a step in another embodiment of the present invention to perform frame-by-frame light pulse duty cycle control based on modified control parameters to compensate for the perceptual drift of the human eye to the fusion frequency of flickering lights.

[0130] S601: Calculate the target duty cycle reference D_ref based on ambient light intensity L_env;

[0131] S602: Adjust the actual duty cycle change rate according to the human eye adaptation function H(τ) = 1 e^{τ / τ_c}, where τ is a time variable and τ_c represents the adaptation time constant;

[0132] S603: Introducing a duty cycle buffer factor b to make the adjustment smoother;

[0133] S604: The final output duty cycle D_final = D_ref × H(τ) + b × DP_robust ensures dynamic stability.

[0134] First, the system acquires the ambient light intensity L_env in real time using an independent ambient light sensor or by utilizing statistical values ​​from non-lit areas in a video image. Based on a pre-calibrated lookup table or empirical formula, a baseline target duty cycle D_ref is determined from L_env. For example, in extremely dark environments, D_ref might be set to a lower value (e.g., 0.3) to maintain a noticeable flicker; in brighter environments, D_ref might be set to a higher value (e.g., 0.8) to ensure sufficient apparent brightness of the light.

[0135] Subsequently, a time function H(τ) simulating the human eye's adaptation process is introduced. This function is defined as H(τ) = 1 - e^{-τ / τ_c}, where τ is the time elapsed since the last abrupt change in ambient light intensity, and τ_c is the human eye's visual adaptation time constant (e.g., dark adaptation τ_c is approximately 10-30 minutes, while light adaptation is shorter). This function describes the exponential process by which human eye sensitivity changes from abrupt change to gradual stabilization. Multiplying D_ref by H(τ) means that the final output baseline value will adjust slowly as the human eye adapts, rather than jumping immediately.

[0136] To eliminate the slight fluctuations that may arise from frame-by-frame parameter calculation, a duty cycle buffer factor b (e.g., b = 0.2) is introduced. This factor determines the contribution weight of the corrected parameter DP_robust calculated in the current frame to the final output. The final output duty cycle is synthesized by the following formula: D_final = D_ref × H(τ) + b × DP_robust. In this formula, D_ref × H(τ) constitutes a slowly changing, smooth "baseline" signal, ensuring that the overall trend of regulation is synchronized with the physiological adaptation of the human eye; while b × DP_robust is a "ripple" signal superimposed on the baseline, which responds quickly but with limited amplitude to the instantaneous, subtle changes detected by video analysis. The sum of the two ensures both the system's agility in responding to sudden environmental changes and avoids drastic oscillations in the output value, ensuring the dynamic stability and visual comfort of the light dimming command, thereby accurately and smoothly compensating for the perceptual drift of the human eye.

[0137] The runway visual range (VAR) deviation assessment method based on background brightness and light fusion of the present invention includes the following steps: First, by acquiring the brightness gradient abrupt change characteristics of the runway edge light sequence in the V channel of the HSV color space in continuous frame video, the change pattern of the runway lights is identified. Next, based on these brightness gradient abrupt change characteristics, preliminary control parameters for the light pulse duty cycle are extracted, thereby preliminarily estimating the human eye's ability to fuse light flicker. Then, combined with the spatial Laplacian rate of change of the background sky brightness, the preliminarily obtained light pulse duty cycle is dynamically corrected to adapt to visual requirements in different environments. Finally, based on the corrected control parameters, the light pulse duty cycle is controlled frame-by-frame for each image, so that the lighting effect can match the human eye's perception of changes in flicker frequency in real time. This method effectively solves the problem of VAR interpretation lag caused by the perceptual drift of the human eye's perception of flicker light fusion frequency in low visibility environments, improving the accuracy and real-time performance of runway VAR assessment, and contributing to improved flight safety and operational efficiency.

[0138] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0139] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0140] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0141] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process. Furthermore, the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functionality involved.

[0142] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0143] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. All or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware, the program being stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiments.

[0144] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. This storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0145] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for evaluating runway visibility deviation based on the fusion of background brightness and lighting, characterized in that, include: Obtain the brightness gradient abrupt change characteristics of the runway side light sequence in the V channel of the HSV color space in a continuous frame video; Preliminary control parameters for the duty cycle of the light pulse are extracted based on the aforementioned brightness gradient abrupt change characteristics; The initial control parameters are dynamically corrected by taking into account the spatial Laplacian rate of change of the background sky brightness. The duty cycle of the light pulse is adjusted frame by frame based on the corrected control parameters to compensate for the perceptual drift of the human eye in the fusion frequency of the flickering lights.

2. The method for evaluating runway visibility deviation based on the fusion of background brightness and lighting as described in claim 1, characterized in that, The preliminary adjustment parameters for extracting the duty cycle of the light pulse based on the brightness gradient abrupt change features include: Statistically analyze the frequency of pulse on / off states within the region of abrupt changes in brightness gradient; The frequency is normalized to the range of [0, 1] and combined with the average pulse frequency over a historical period; Weighting coefficients α and β are set to distinguish response priorities in different scenarios; The initial duty cycle is calculated using the following formula: DP = α × (f / f_avg) + β × (m / m_total), where f is the pulse frequency of the current frame, f_avg is the historical average frequency, m is the number of abrupt change points, and m_total is the total number of pixels in the current frame.

3. The method for evaluating runway visibility deviation based on the fusion of background brightness and lighting as described in claim 1, characterized in that, The dynamic correction of the initial control parameters by incorporating the spatial Laplacian rate of change of background sky brightness includes: Extract the background sky region from consecutive frame videos and perform grayscale processing; Calculate the Laplacian rate of change λ for the background image of each frame, denoted as: ,in It is the second derivative, and N is the total number of pixels; An adjustment coefficient k is introduced to represent the relationship between the rate of change of background brightness and the perceived drift of the human eye; The correction is performed using the following formula: DP_corrected = DP × e^(kλ), which is used to reduce the impact of background fluctuations on pulse duty cycle control.

4. The method for evaluating runway visibility deviation based on the fusion of background brightness and lighting as described in claim 3, characterized in that, The step of implementing frame-by-frame light pulse duty cycle adjustment based on the corrected control parameters to compensate for the perceptual drift of the human eye to the fusion frequency of flickering lights includes: Divide DP_corrected into two regions, high and low, and define them as DH and DL respectively; The duty cycle output is adjusted using linear interpolation within the DH region; Set the dynamic offset δ(t) = a × t + b, where t represents the current time step, and a and b are the scaling and constant factors, respectively. The actual output duty cycle DP_output = DP_corrected + δ(t) is adjusted to adapt to changes in human visual adaptation.

5. The method for evaluating runway visibility deviation based on the fusion of background brightness and lighting as described in claim 1, characterized in that, The acquisition of the brightness gradient abrupt change features of the runway side light sequence in the HSV color space V channel in consecutive frame video includes: A multi-frame accumulation method is used to perform temporal analysis on consecutive frame images to construct a video motion vector field; Hough transform was used to detect the contour of side lights and their brightness distribution trend. For each frame, calculate the difference between the maximum and minimum brightness values ​​of the side lights, Max_V and Min_V. Calculate the relative mutation index μ = (Max_V-Min_V) / V_avg. If μ > θ, where θ is a pre-set threshold, then it is determined that there is a significant gradient mutation.

6. The method for evaluating runway visibility deviation based on the fusion of background brightness and lighting as described in claim 5, characterized in that, The preliminary adjustment parameters for extracting the duty cycle of the light pulse based on the brightness gradient abrupt change features include: Group the pixels within the edge light area by row and analyze their brightness change sequence; Calculate the standard deviation σ_i of the brightness fluctuation for each pixel; By comparing σ_i with the background standard deviation σ_b, the brightness correlation parameter ω = σ_i / σ_b is constructed; The initial duty cycle parameter DP_initial = ∑ω_i × f_i / ∑ω_i is generated for each frame using a weighted moving average method.

7. The method for evaluating runway visibility deviation based on the fusion of background brightness and lighting as described in claim 6, characterized in that, The preliminary adjustment parameters for extracting the duty cycle of the light pulse based on the brightness gradient abrupt change features include: Support vector machines were used to model the brightness mutation characteristics and historical duty cycle control parameters; Predict the expected pulse frequency f_pred for each frame using training data; A dynamic feedback error E = |f_true-f_pred| is introduced as an adjustment weight parameter; Update the control parameters according to the following formula: DP_new = DP_prev × (1-E / f_max), where f_max represents the maximum allowable frequency.

8. The method for evaluating runway visibility deviation based on the fusion of background brightness and lighting as described in claim 7, characterized in that, The dynamic correction of the initial control parameters by incorporating the spatial Laplacian rate of change of background sky brightness includes: Set a reference window in the background sky area and record the average brightness M_sky within the window; Compare the current frame brightness M_current with M_sky and calculate the brightness deviation ΔM = |M_current-M_sky|; Multiply ΔM by its time derivative term to obtain the brightness change rate factor; The corrected control parameter is: DP_robust = DP_initial × (1-ΔM × [d(ΔM) / dt] / σ_sky), where σ_sky represents the standard deviation of background brightness.

9. The method for evaluating runway visibility deviation based on the fusion of background brightness and lighting as described in claim 8, characterized in that, The step of implementing frame-by-frame light pulse duty cycle adjustment based on the corrected control parameters to compensate for the perceptual drift of the human eye to the fusion frequency of flickering lights includes: The target duty cycle reference D_ref is calculated based on the ambient light intensity L_env; The actual duty cycle change rate is adjusted according to the human eye adaptation function H(τ) = 1 e^{τ / τ_c}, where τ is a time variable and τ_c represents the adaptation time constant; Introducing a duty cycle buffer factor b enables a smooth transition in regulation; The final output duty cycle D_final = D_ref × H(τ) + b × DP_robust ensures dynamic stability.

Images