Quantitative assessment method for high-altitude geological hazard chain risks adapted to cold and cold environments

Abstract

This invention discloses a quantitative assessment method for high-altitude geological hazard chain risks adapted to high-altitude and cold environments, relating to the field of geological hazard monitoring and prevention technology. The method includes the following steps: S100, establishing a joint irradiation and albedo observation layer, collecting the dynamic intensity of direct and reflected light, generating an exposure time-series fingerprint baseline, and determining the critical threshold of direct light intensity based on this baseline. This invention constructs an irradiation and albedo joint observation layer to extract light intensity features and generate an exposure time-series fingerprint baseline; it uses coherent phase decomposition and counterfactual playback to extract trigger point boundaries and reconstruct continuous trajectories; it predicts overexposure risk based on time-frequency scoring, and uses spectral gating and polarization rotation to regulate the exposure rhythm; and it implements inverse phase traction and phase conjugate intervention within the steady-state window to maintain boundary information and improve risk identification accuracy.

Description

Technical Field

[0001] This invention relates to the field of geological disaster monitoring and prevention technology, specifically to a quantitative assessment method for the risk of high-altitude geological disaster chains adapted to cold and high-cold environments. Background Technology

[0002] "Quantitative Risk Assessment of High-Altitude Geological Hazard Chains Adapted to Cold and Cold Environments" refers to the systematic modeling of the triggering, evolution, and coupling relationships of each link in a continuously evolving disaster chain involving multiple stages such as ice and rock collapses, landslides, debris flows, barrier lakes, and flood outbursts, combined with the extreme climatic conditions of high-altitude and cold regions. By acquiring key parameters through cold-resistant monitoring equipment, an assessment index system capable of quantifying the probability, intensity, and cascading effects of disasters is established. This enables interconnection and intelligent linkage monitoring among different types of disasters, thereby forming a quantitative risk identification and early warning capability for the entire disaster chain.

[0003] Existing technologies have the following shortcomings: Current technologies for monitoring high-altitude ice and rock avalanche trigger points typically rely on optical sensors to acquire information on surface cracks, avalanche boundaries, and morphological changes. However, in high-altitude and cold regions, sudden strong direct sunlight and snow reflection under clear skies can easily cause oversaturation of optical sensors, resulting in bright spots and overexposure in the images. This severely blurs the boundary identification results of ice and rock avalanche trigger points, leading to the masking or delayed identification of disaster chain initiation signals. Once this problem occurs, it directly affects the accuracy and timeliness of quantitative risk assessment, reducing the effectiveness of early warning and prevention of chain disasters.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a quantitative assessment method for the risk of high-altitude geological disaster chains that is adapted to cold and high-cold environments, so as to solve the problems in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a quantitative assessment method for high-altitude geological hazard chain risks adapted to high-altitude and cold environments, comprising the following steps:

[0007] S100 establishes a joint observation layer for irradiation and albedo, collects the dynamic intensity of direct and reflected light, generates an exposure time-series fingerprint baseline, and determines the critical threshold of direct light intensity based on the baseline, which is used as a reference for image processing and boundary recognition.

[0008] S200 performs coherent phase decomposition based on the exposure time fingerprint baseline, separates overexposed bright spot interference, retains trigger point boundary image information, and extracts phase kernel signals for image restoration;

[0009] S300 triggers a counterfactual playback chain based on the phase kernel signal, replaces overexposed image frames with low-exposure shadow sequences, restores the gradient continuity of the trigger point boundary, and generates a time-frequency consistency score for trajectory reconstruction.

[0010] S400, based on the high-value segment of the time-frequency consistency score, reconstructs the image frame sequence and uses dual mirror time-scale calibration to calibrate the cross-frame delay, thereby obtaining the continuously evolving trigger point boundary trajectory;

[0011] The S500 predicts the overexposure risk of the next time window based on the boundary trajectory of the trigger point, and links the spectral gating structure and polarization rotation structure to set the exposure rhythm and intensity threshold, forming a steady-state window reference.

[0012] The S600 performs time-reversal light field traction within a steady-state window, injects inverse phase micro-patterns, and combines a shadow shading array with a phase conjugate projection system to eliminate overexposure interference, maintain the continuity of trigger point boundaries, and achieve dynamic identification of disaster chain initiation signals and improve the accuracy of risk assessment.

[0013] Preferably, step S100 includes:

[0014] In high-altitude and cold regions, representative observation locations around the ice and rock avalanche trigger points are selected to collect dynamic intensity information of direct and reflected light.

[0015] The intensity of direct light is collected by deploying optical imaging devices with a wide dynamic response range, and the intensity of snow and ice reflection light at different slopes and heights is collected by deploying spectral detection devices with high spectral resolution.

[0016] The observation data of direct light and reflected light are synchronized in time and accumulated over a continuous solar cycle to generate an exposure time-series fingerprint baseline that integrates the characteristics of direct and reflected light interference.

[0017] Based on the key nodes and slope abrupt change segments of light intensity change in the exposure time-series fingerprint baseline, the critical threshold of direct light intensity is calculated, and the critical threshold is used for light intensity determination before image acquisition and for marking bright spot interference areas after image acquisition.

[0018] Preferably, step S200 includes:

[0019] The image frames are time-aligned with the exposure timing fingerprint baseline to identify bright spot interference areas and extract brightness transition boundaries.

[0020] In the bright spot interference region, spatial interference structure reconstruction is performed, and the light intensity response structure is reconstructed based on historical illumination differences to generate a phase interferogram;

[0021] The phase interferogram is used to divide the grid, filter the phase gradient coherent region, and extract the phase kernel signal with boundary extension capability;

[0022] The phase kernel signals are organized into a structured signal table according to the image frame number and time label, and bound to the exposure time sequence fingerprint baseline for subsequent image restoration to recover boundary structure information.

[0023] Preferably, the extraction of the phase kernel signal is based on the condition that the difference in phase gradient direction between adjacent grids is lower than a set threshold and shows a linear extension trend. The extracted phase kernel signal is accompanied by a time label, spatial coordinates, phase consistency score and illumination disturbance label.

[0024] Preferably, step S300 includes:

[0025] Based on the spatial location and time label of the phase kernel signal in the image sequence, image blocks under low exposure conditions in historical images are retrieved, cropped and brightness mapping adjustment is performed to generate image regions for replacement;

[0026] The brightness-adjusted image block is embedded into the overexposed area of ​​the current image frame, and boundary smoothing is performed using directional guidance and weighted fusion to reconstruct the gradient continuity of the trigger point boundary.

[0027] The stability of the replaced image frames is evaluated in terms of time and frequency dimensions, a comprehensive time-frequency consistency score is generated, and the score is recorded in the score table;

[0028] Based on the consistency scoring table, consecutive high-scoring image frames are selected, structural boundary points are extracted, and trajectory curves are constructed to complete the reconstruction and quality labeling of the trigger point boundary trajectory.

[0029] Preferably, step S400 includes:

[0030] Based on the time-frequency consistency score, consecutive high-scoring image frames were selected, and frames located in high-risk areas of illumination fluctuation were removed to determine stable frames for image sequence reconstruction.

[0031] In stable frames, image feature points with fixed spatial characteristics are selected, and the inter-frame time deviation is calibrated using a double mirror timescale method to unify image frames to the standard time axis.

[0032] The image frames are rearranged according to the corrected time order, the boundary contour segments of the trigger points are extracted, the structural coherence is determined, and the boundary trajectory sequence is constructed.

[0033] The boundary trajectory sequence is structured and recorded, including image frame number, timestamp, boundary coordinates and trajectory change characteristics, to form a dataset that supports time series modeling.

[0034] Preferably, each boundary point in the boundary trajectory sequence is assigned a time-frequency consistency score, a structural stability weight value, and a trajectory continuity evaluation score from the corresponding image frame to improve the reliability screening accuracy of the trajectory points in subsequent dynamic modeling.

[0035] Preferably, step S500 includes:

[0036] Extract continuous evolution segments of boundary trajectories on the time axis, calculate boundary change rate and image brightness parameters, and identify overexposure risk time segments with drastic boundary change trends and rapid brightness increase trends;

[0037] Based on the activation of the spectral gating adjustment structure and polarization rotation structure in the overexposure risk time segment, the incident light center wavelength, transmission direction and polarization angle are linked to control, forming an optical control channel that physically suppresses light input.

[0038] A multi-stage exposure control strategy is set within the predicted risk time range, with exposure time, aperture size and image gain set in segments, and parameters are dynamically adjusted according to image recognition rate and contrast fluctuations.

[0039] A steady-state window benchmark covering the entire risk time period is constructed, and the boundary recognition connectivity and structural integrity are verified in image frames to achieve closed-loop control of optical input adjustment.

[0040] Preferably, step S600 includes:

[0041] Construct a set of structural parameters for the interference light field, extract the spatial contour, brightness variation characteristics and frequency domain energy distribution of the overexposed area, and generate an inverse phase micro-pattern that has phase cancellation characteristics with the interference light field.

[0042] The inverse phase micro-pattern is injected into the light wave propagation path in real time through spatial light modulation, and phase interference superposition is performed in front of the incident light to reduce the energy of the interference wave on the imaging surface.

[0043] By introducing a shadow occlusion array and setting the transmittance of the absorption unit according to the boundary shape of the trigger point and the direction of illumination, the directional occlusion of local image saturation risk and the suppression of edge light intensity gradient are achieved.

[0044] By performing optical wave inverse reconstruction compensation on the boundary structure through phase conjugate projection, the connectivity and gradient stability of the restored region boundary are detected, and it is determined whether the image acquisition quality of the current frame meets the admission criteria.

[0045] Preferably, the injection position of the inverse phase micropattern is limited to the propagation plane before the incident light path of the overexposed area. The transmittance of the absorption unit in the shadow shading array is dynamically adjusted according to the shading angle and the incident light intensity. The phase conjugate projection only acts on the area with a boundary connectivity of less than 95%, and the reconstruction is based on the symmetrical phase information of the corresponding boundary in the historical image.

[0046] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0047] This invention constructs a joint irradiation and albedo observation layer to accurately acquire the dynamic intensity characteristics of direct and reflected light, generating an exposure time-series fingerprint baseline to provide a highly reliable reference for subsequent image overexposure identification and processing. Furthermore, it combines coherent phase decomposition and counterfactual playback techniques to remove interference signals at the physical fluctuation level, accurately extracting the structural core of the trigger point boundary. A time-frequency consistency score is then used to establish an image stability quantification model and reconstruct continuous boundary trajectories, significantly improving the coherence and completeness of trigger point identification. Simultaneously, spectral gating and polarization rotation techniques are introduced to construct an exposure rhythm control strategy, and time-reversed light field traction and phase conjugate intervention are implemented within the steady-state window to achieve active extinguishing of overexposed areas and physical compensation of boundary images. In summary, this invention constructs a full-chain linkage evaluation mechanism from information acquisition, image recognition, risk prediction to physical interference suppression, significantly improving the accuracy and real-time performance of chain geological disaster initiation signals in high-altitude and cold regions, and enhancing the applicability and accuracy of image data in disaster early warning and risk modeling. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0049] Figure 1 This is a flowchart of the method for quantitatively assessing the risk of high-altitude geological disaster chains adapted to cold and frigid environments, as described in this invention. Detailed Implementation

[0050] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0051] This invention provides, for example Figure 1 The method for quantitatively assessing the risk of high-altitude geological hazard chains adapted to cold and frigid environments, as shown, includes the following steps:

[0052] S100 establishes a joint observation layer for irradiation and albedo, collects dynamic intensity information of direct and reflected light in high-altitude and cold regions, generates an exposure time-series fingerprint baseline, and determines the critical threshold of direct light intensity based on the exposure time-series fingerprint baseline as a reference benchmark for subsequent image processing and boundary recognition.

[0053] In quantitative risk assessment of high-altitude geological hazard chains in cold and frigid regions, to address optical overexposure caused by strong direct sunlight and snow reflection, and thus improve the accuracy and stability of trigger point boundary identification, a dynamic light intensity acquisition and threshold determination method based on joint irradiance and albedo observations is proposed. This method is implemented through the following specific steps:

[0054] Within the target high-altitude and frigid observation area, several representative geographical locations near potential ice and rock avalanche trigger points were selected. High-sensitivity optical imaging equipment was deployed at observation sites directly opposite the trigger points to collect direct solar irradiance. Simultaneously, multiple high-precision spectroscopic detection devices were deployed at locations around the trigger points at different altitudes and with varying slopes to collect the intensity of reflected light from the ice and snow surface. The optical imaging equipment utilizes sensor arrays capable of continuous operation at temperatures as low as -35 degrees Celsius with a wide dynamic range to ensure accurate light intensity recording in both strong and weak light environments. The spectroscopic detection devices employ probes with angular resolution less than one degree and high spectral resolution, enabling them to separate the reflected light contributions of different wavelengths. In the actual deployment, slope aspect, ice and snow cover thickness, solar orbit projection, and wind speed and direction were analyzed individually. The equipment was fixed on unobstructed and structurally stable bases to ensure the continuity and comparability of long-term observation data. All imaging equipment and spectral detection devices are synchronized using high-precision atomic clocks, with sampling frequencies set at the millisecond level to ensure that light intensity records from all angles are perfectly aligned at the same moment, laying a precise time foundation for subsequent time-series analysis.

[0055] After the equipment completes simultaneous observations at multiple points, the direct and reflected light intensity data collected over at least thirty consecutive complete sunshine cycles will be systematically processed to construct a light intensity variation curve under real-world illumination conditions in high-altitude and cold regions. For each observation point, minute-by-minute light intensity data will be extracted throughout the day, and three characteristic phases—low-angle irradiance in the early morning, high-intensity stable irradiance at noon, and light intensity decay in the afternoon—will be marked. Within each phase, the light intensity increase rate, peak duration, and decay rate will be recorded. For reflected light data, corrections will be made based on the slope angle, snow reflectivity, and snow depth at each observation point. The contribution of reflected energy at each angle to the trigger point will be calculated point-by-point, and these contribution values ​​will be superimposed on the direct light curve along the time axis to obtain a complete combined irradiance and reflected light intensity curve. This curve accurately depicts the real-world illumination interference characteristics of high-altitude and cold regions at different time periods, and each time point will have corresponding geographical and meteorological parameter labels for subsequent traceability and calibration.

[0056] After obtaining the combined irradiance and albedo intensity curves, key inflection points are extracted based on continuous time series to construct an exposure time-series fingerprint. Specifically, the starting point of a sharp increase in intensity, the time of reaching the peak, the duration of the peak, the overlapping interval of enhanced albedo, and the ending point of a sharp decrease in intensity are identified one by one. These key nodes are marked according to time sequence and intensity amplitude, forming a characteristic fingerprint curve of intensity fluctuations within a day. The fingerprint curves of at least thirty sunshine cycles are overlaid and analyzed to remove outliers caused by occasional weather changes. The average time position, average intensity, and standard deviation of each key node are extracted, forming a highly reproducible exposure time-series fingerprint baseline in high-altitude and cold environments. This baseline not only records the typical patterns of intensity changes under specific meteorological conditions but also reflects the superposition intensity and temporal characteristics of direct light reflected from ice and snow, enabling subsequent image processing to detect and avoid high-risk exposure sections in advance. Unlike traditional methods that simply control intensity by fixing the upper limit of exposure, this baseline has dynamic adaptability in both time and intensity dimensions, automatically updating with changes in season and solar altitude angle.

[0057] Based on the constructed exposure time-series fingerprint baseline, time periods of abrupt slope changes in direct light intensity are extracted. Combined with the amplification ratio of reflected light intensity within these time periods, the critical exposure threshold of the imaging device under the current environment is calculated. This threshold is not a single fixed value, but a set of interval values ​​that evolve over time, covering the possible safe exposure range and overexposure risk range at every point in time throughout the day. By directly importing this threshold into subsequent image acquisition and boundary recognition steps, it is possible to determine whether the current time point is in a high-risk overexposure zone before image acquisition. If it is in a high-risk zone, the sensor exposure time, gain, and polarization angle are actively adjusted to ensure that the image frame to be acquired avoids the critical light intensity. If an image frame has already been acquired, possible bright spot interference areas are marked in the image data based on the fingerprint baseline, providing location and time indexes for subsequent coherent phase decomposition and bright spot removal.

[0058] S200 performs coherent phase decomposition on the basis of the exposure time fingerprint baseline, separates the bright spot interference signal in the oversaturated region, retains the boundary image information of the trigger point region, and extracts the corresponding phase kernel signal for subsequent image restoration processing.

[0059] After constructing the exposure-time fingerprint baseline, to remove bright spot interference caused by excessive illumination in the image and extract structural signals that can be used for image restoration, thereby ensuring the integrity and accuracy of trigger point boundary recognition, a phase decomposition processing method based on the coherence principle is proposed. This method includes the following steps:

[0060] The image frame acquired at the current time point is precisely time-aligned with the corresponding exposure time-series fingerprint baseline. Based on the critical thresholds for direct and reflected light marked in this baseline, regions in the image that may exhibit bright spot interference due to light intensity saturation are identified. Within the image frame, brightness values ​​are traversed pixel by pixel. When a brightness value exceeds the upper limit threshold marked in the exposure baseline at the corresponding time point, the pixel is marked as an overexposure candidate. Subsequently, in the two-dimensional image coordinate system, a brightness gradient detection region is constructed with each overexposure candidate point as its center, using 8 neighboring pixels as units, to determine if there are strong brightness jumps. If the jump gradient is greater than a set recognition threshold, the region is determined to be an overexposure interference area. To improve the accuracy of region localization, the original image is converted from the RGB color space to a brightness-hue-saturation space, and the bright spot morphology boundaries are extracted from the brightness component. This method can accurately identify high-brightness regions caused by direct light enhancement or snow reflection within a specific time period, providing a boundary framework for subsequent phase structure extraction.

[0061] Within the identified bright spot regions, a spatial interferometric structure reconstruction operation is performed to extract potential spatial structure phase information from image areas covered by high-saturation light. Specifically, for each bright spot region, the pixel intensity fluctuation values ​​of its adjacent areas are extracted. Using this region as a window, image frames of the same region acquired under different lighting conditions throughout history are retrieved, and image overlay and spatial matching are performed. The light intensity response structure of the region under normal exposure conditions is reconstructed using the brightness shift caused by differences in lighting over multiple time periods. These light intensity structures are converted into complex wavefront representations, and the phase interferogram of the region is reconstructed based on the light wave propagation path. Then, by integrating the phase difference at the same location in multiple image frames, high-frequency perturbation terms from direct or reflected paths are removed, retaining low-frequency phase responses with consistent direction and stable fluctuations, thereby obtaining the spatial continuity structure of the original scene beneath the bright spot region. Through this coherent interferometry method, even in visually completely distorted bright spot regions, deep structural information can be extracted using response differences from historical lighting changes, providing data support for image restoration.

[0062] Based on the generated phase interferogram, phase change data within the overexposed area is classified and filtered to extract phase kernel signals with boundary extension capabilities. This process uses the consistency of phase distribution within the region as a criterion. Specifically, the bright spot region is divided into several 10-pixel × 10-pixel small grids. The phase gradient direction is calculated in each grid and compared with the direction changes of its neighboring grids. When the gradient direction difference between multiple neighboring grids is below a preset threshold and exhibits a clear linear extension trend, it is determined that a phase-continuous path suitable for structural recovery exists within the region. Subsequently, phase kernel extraction segments are constructed according to gradient direction consistency. Starting from each core point, the search extends to both sides until a sudden change in phase direction or a break in the brightness boundary occurs. All paths satisfying phase consistency and extension are marked as phase kernel signals. Each phase kernel signal is accompanied by a time label, spatial coordinates, phase consistency score, and illumination perturbation identifier to form a structured data index required for subsequent retrieval.

[0063] All extracted phase kernel signals are uniformly organized into a structured phase signal table and stored sequentially according to image frame number. In the corresponding index table for each frame, the specific location, extension direction, continuous length, and consistency level of the extracted phase kernel signals for all bright spot regions in that frame are recorded. Simultaneously, to enhance their adaptability in subsequent image restoration stages, these phase kernel signals are time-stamped with the exposure time-series fingerprint baseline, ensuring each phase kernel has a traceable illumination environment background. When proceeding to the counterfactual image reconstruction or trigger point boundary trajectory construction stage, these signal tables can be used to quickly locate the original boundary signals preserved in overexposed areas. Guided by these signals, gradient completion, structural extension, or pixel interpolation can be performed, improving the accuracy and structural stability of image restoration.

[0064] S300 triggers a counterfactual playback chain based on the extracted phase kernel signal, replaces overexposed image frames with shadow sequence images recorded under low exposure conditions, reconstructs the gradient continuity of the trigger point boundary, and generates a time-frequency consistency score that reflects the stability of the image for subsequent boundary trajectory reconstruction.

[0065] To recover the structural information of overexposed areas in images caused by strong light exposure in high-altitude and cold environments, and to ensure the stable identification of trigger point boundaries in a time series, a counterfactual playback chain construction method based on phase kernel signal driving is proposed. This method uses historical low-exposure images as the data source and improves the completeness and accuracy of boundary identification through image replacement, continuous reconstruction, and scoring evaluation. Specifically, it includes the following steps:

[0066] Based on the extracted phase kernel signals, the frame number, corresponding spatial coordinate range, and structural extension direction of each phase kernel in the image sequence are located. Using the acquisition time of the image frame containing the phase kernel as a reference, at least five historical images captured under different exposure conditions are retrieved both forward and backward, with a focus on selecting image frames recorded under low-exposure conditions at the same spatial location. During the selection process, image frames with lower solar incidence angles, weaker reflected light intensity, and overall brightness levels less than 40% of the current frame are prioritized as candidates. Subsequently, image blocks completely consistent with the current overexposed area are cropped from these candidate image frames, and their boundary pixel values, gradient directions, and brightness ranges are recorded. To eliminate brightness differences caused by the imaging device capturing images at different times, brightness mapping adjustments are performed on the cropped image blocks, adjusting their average brightness value to match the brightness of the unexposed area in the current image frame. A contrast transition within a 5-pixel range is applied to high-contrast edges to ensure consistency in the overall style of subsequent replacement areas.

[0067] The brightness-adjusted image block is embedded into the corresponding overexposed area in the current image frame. To achieve seamless replacement, the boundary orientation of the area to be replaced in the original image must first be analyzed, and the optimal insertion path is determined based on the directional information recorded by the phase kernel signal. During the insertion process, a pixel-by-pixel registration strategy is adopted, starting from the upper left corner of the image block, aligning with the spatial coordinates of the corresponding pixels in the current frame, and filling the image block line by line to the entire overexposed area. A 10-pixel-wide linear buffer band is set in the transition area of ​​the replacement edge. By calculating the brightness difference between the boundary pixels of the original image and the boundary pixels of the replacement image block, weighted fusion is performed to generate a transition brightness distribution. For areas with significant changes in structural features, such as crack edges or snow-covered boundaries, a spatial interpolation method guided by the main direction of the phase kernel is used for structural extension, ensuring the continuity of the boundary lines inside the replacement block at the connection points and preventing discontinuities, misalignments, or morphological distortions. Finally, a new image frame is formed that has no obvious visual replacement traces and is structurally continuous and discernible.

[0068] For the replaced image frames, their stability in the image sequence is evaluated from two dimensions: time and frequency. In the time dimension, along the principal axis of the region where the phase kernel is located, the boundary gradient information of the current frame and its four adjacent frames (the previous two and the next two frames) at the same spatial location is extracted, and the gradient change rate of this region between frames is calculated. If this change rate remains within 5%, the region is considered to have temporal continuity. In the frequency dimension, the brightness sequence of this region in five images is subjected to a Fast Fourier Transform to extract the principal components of the spectrum. The concentration of brightness energy in the low-frequency band is evaluated. If the energy proportion of the principal components exceeds 85%, the region is judged to have good frequency domain stability. The temporal continuity score and the frequency stability score are combined with a weight ratio of 6:4 to form the consistency score of the image frame in the time-frequency dimension. All scoring results are recorded in the consistency score table indexed by the image frame number, providing a quantitative basis for trajectory extraction and quality screening. This scoring method differs from traditional image quality judgment based on single-frame edge sharpness by introducing two orthogonal dimensions: time series and frequency response, thus strengthening the systematic evaluation capability of image stability.

[0069] Based on the time-frequency consistency scoring results, consecutive image frames with consistency scores higher than a preset threshold are selected from the image sequence as boundary trajectory extraction targets. Within these image frames, spatial regions corresponding to the phase kernel signals are extracted. Using a pixel boundary tracing method, boundary inflection points in the image are connected point-by-point along the structural extension direction to construct a complete spatial boundary trajectory curve. To improve the smoothness and anti-interference capability of the trajectory, third-order spline interpolation is introduced during the connection process to perform fitting processing on the trajectory curve, ensuring that the curve has continuity of the first derivative and finiteness of the second derivative at the connection points. Simultaneously, the time-frequency consistency score of the image frame containing each trajectory point is superimposed as a quality weight label for that point, used for automatic selection of trajectory credibility in the subsequent dynamic modeling stage. If the time label of the image frame containing the trajectory point falls within the time interval marked as "high-risk segment" in the exposure time-series fingerprint baseline, the point is separately marked as a "structurally suspicious node," indicating that its accuracy should be verified in subsequent analysis. Through the above methods, a boundary trajectory reconstruction process consisting of multi-frame image fusion reconstruction, scoring evaluation, and structural fitting was constructed, which ensured the continuity and stability of the disaster trigger point boundary under the background of light interference, and provided a reliable data foundation for subsequent disaster chain evolution analysis and risk prediction.

[0070] S400 performs image frame sequence reconstruction based on the high-value segment of time-frequency consistency score, uses a dual mirror time-stamping method to calibrate the cross-frame time difference, rearranges the image sequence, and obtains the continuously evolving trigger point boundary trajectory as time series support data for the trigger point evolution process.

[0071] To achieve continuous identification and dynamic characterization of the temporal changes of trigger point boundaries, a sequence reconstruction method based on image frames with high time-frequency consistency scores is proposed. This method calibrates the inter-frame time difference using dual-mirror timescales, rearranges the image sequence, and generates boundary trajectory data suitable for temporal modeling. Specifically, it includes the following steps:

[0072] Based on the calculated image frame time-frequency consistency score table, consecutive frame segments with scores greater than 90 points were selected from the complete image sequence as the starting data for reconstruction. During the selection process, the consistency score values ​​of each frame were compared one by one using the image frame number as an index. High-scoring segments with at least five consecutive frames and a score fluctuation of no more than 5 points between adjacent frames were identified. The starting frame number, ending frame number, inter-frame time interval, consistency mean, and maximum brightness gradient difference of the segment were extracted as five-dimensional feature descriptions. Subsequently, these candidate frame segments were cross-compared with the previously established exposure time-series fingerprint baseline, and frame segments located in high-risk areas of illumination fluctuation were deleted. This ensured that the finally selected frame segments were under conditions of structural stability, balanced brightness, and no interfering background, providing quality assurance for subsequent image frame time reconstruction.

[0073] To eliminate inter-frame time differences in image acquisition at high altitudes, such as time errors caused by equipment clock drift, mechanical shutter response delay, and optical trigger misalignment, a dual-mirror time-stamp calibration strategy is introduced to synchronize the time of all selected image frames. Specifically, within the trigger point boundary region of each frame, at least thirty image feature points with fixed spatial locations, stable morphology, and clear contours are selected, such as the tips of ice and rock fissures, the bends of landslide edges, or the shadow boundaries of snow surface protrusions. Using these feature points, the horizontal and vertical pixel displacements are calculated between every two frames to derive the true inter-frame time offset. Then, using the current image sequence as the forward time axis, a reverse mirror time axis is constructed, and symmetrical fitting is performed on the trajectories of all feature points on this reverse time axis. By comparing the convergence of the forward and reverse trajectories on each frame, the minimum time difference point is found, and this point is used as a reference point for time axis realignment. Finally, the timestamps of all frames are adjusted to unify the image frames onto a standard time axis centered on this reference point, achieving fine correction of acquisition time differences.

[0074] After completing inter-frame time synchronization, all image frames are rearranged according to the corrected timestamps to establish an image frame sequence with strong continuity, small time difference, and stable structure. To verify the structural consistency of the images in the rearranged sequence, contour segments of the trigger point boundaries are extracted from all frames, and their shape parameters (including contour length, average curvature, and contour direction angle) are calculated and compared frame-by-frame with corresponding regions in adjacent frames. If the contour direction difference between consecutive frames is less than 5 degrees, the contour length change rate is less than 10%, and the curvature fluctuation value is less than 0.1, then the image frame segment is determined to meet the structural stability standard. Based on the above judgment, image sequence segments that meet the structural consistency requirements are extracted from all rearranged sequences, and based on these, the spatial coordinates of the trigger point boundary lines are extracted sequentially along the image frame sequence direction. All boundary coordinates form a boundary trajectory sequence in chronological order. This sequence describes the morphological evolution process of the trigger points with pixel-level spatial distribution and millisecond-level temporal precision, becoming a high-precision dynamic representation of the disaster evolution behavior.

[0075] The reconstructed boundary trajectory sequence is further structured to construct a standard dataset to support temporal modeling of trigger points. This dataset stores data items such as image frame number, correction timestamp, boundary coordinates, trajectory direction, boundary change rate, and deformation rate of adjacent points for each boundary trajectory point in a multi-dimensional structure. To enhance the analytical capabilities of this data in subsequent early warning simulations, each trajectory point is appended with the time-frequency consistency score of the image frame in which the point is located, the image illumination level label, the structural stability weight value, and the trajectory continuity evaluation score. Based on the overall trend of the trajectory, abrupt change points with a slope exceeding a set threshold are identified and highlighted as potential precursor events for the activation of disaster chains.

[0076] The S500 predicts the risk of image overexposure in the next time window based on the boundary trajectory data of the trigger point, and controls the spectral gating structure and polarization rotation structure in conjunction. Based on the prediction results, it sets the exposure rhythm and intensity threshold strategy to form a steady-state window reference for optical input control.

[0077] To achieve predictive suppression of image overexposure risk and dynamic stable control of optical input, a prediction and linkage control method based on trigger point boundary trajectory data is proposed. This method integrates trajectory evolution trends, illumination variation characteristics, and physical intervention methods to construct a steady-state window benchmark, thereby improving boundary recognition clarity and temporal image stability. Specifically, it includes the following steps:

[0078] Based on the constructed trigger point boundary trajectory data, continuous trajectory evolution segments are extracted on the time axis, and the boundary position change rate, spatial displacement amplitude, and boundary curvature change rate at each time node are calculated frame by frame. For each frame, a multi-parameter set is formed, including boundary displacement velocity, boundary change direction angle, and local contour abrupt change rate. This set is then cross-compared with the average brightness, the distribution of the highest brightness pixel location, and the maximum brightness gradient of image frames from the same time period to identify time segments that simultaneously exhibit both a "drastic boundary change trend" and a "rapid increase in brightness trend." If the above parameters show an accelerating trend for three consecutive frames within the same segment, it is identified as a potential overexposure risk segment. This judgment criterion does not rely on a single brightness value but integrates the coupled characteristics of image structure perturbation and illumination changes, thereby improving the prediction accuracy of impending image quality instability.

[0079] After identifying the overexposure risk zone, the spectral gating and polarization rotation structures for input light control are activated based on the start and end frame numbers of that zone on the time axis. In the spectral control stage, a controllable dielectric filter is selected, with its incident light center wavelength precisely adjusted by a micro-voltage adjustment unit. When the predicted light intensity continues to increase, the cutoff band is actively narrowed towards shorter wavelengths (e.g., the 400nm to 500nm band), thereby reducing the incidence of high-energy short-wavelength light components. Based on this, the spectral filter is superimposed with a liquid crystal phase retardation film. By controlling the alignment direction and periodic spacing of the liquid crystal molecules, selective transmission of light energy distribution at different incident angles is achieved. Simultaneously, for polarization control, a double-layer polarization rotation film is used. The first layer has a fixed direction for reference, while the second layer's polarization angle is driven by an electronic control unit, allowing arbitrary rotation within the range of 0 to 180 degrees. When the trigger point is located in the direction of the rising sun or in an area with enhanced snow and ice reflection, the polarization angle is automatically rotated to a direction perpendicular to the reflected polarization direction, suppressing high-reflection interference. When the two work together, they form a light input control channel that can simultaneously physically suppress spectral energy and polarization direction.

[0080] After completing the light input path adjustment preparation, the exposure control strategy is divided into segments based on the number of frames and duration of the predicted risk time window. This window is further divided into multiple sub-segments, each no shorter than two frames, with their respective exposure time, aperture size, and image gain parameters set. In practice, the initial stage of the risk window is set as the "fast response segment," employing a fast exposure mode with an exposure time 30% shorter than the normal duration to avoid high-energy light pulse input. The middle segment is set as the "transition adjustment segment," where the exposure time gradually increases while maintaining constant gain. The final segment is set as the "stable maintenance segment," restoring the normal exposure time but compressing the amount of light input through a small aperture to maintain controllable image brightness. Five buffer frames are set before and after each sub-segment to adjust parameters frame by frame, preventing false structure misjudgments caused by sudden changes in image brightness. All parameter settings are based on real-time feedback of image boundary recognition rate, image contrast fluctuation rate, and phase kernel extraction success rate. If the recognition rate drops by more than 10%, the exposure time of the current sub-segment is immediately reduced or the aperture is decreased, thus forming an exposure control process that dynamically adapts to recognition accuracy.

[0081] After completing spectral adjustment, polarization control, and exposure strategy settings, an optical input steady-state window benchmark covering the entire overexposure risk time window is constructed. This steady-state window not only has clear start and end frame markers on the time axis but also has fixed physical control parameters in the spatial dimension, such as the filter center band, polarization angle direction, exposure time period, and aperture light-passing area. This window provides feedforward protection in actual image acquisition, ensuring that the light signal entering the image sensor in each frame has been "pre-regulated," significantly reducing the probability of sudden bright spots. Simultaneously, within the image frames of the window, indicators such as brightness peak, trigger point boundary recognition connectivity, and structural extension direction integrity are recorded to verify the effectiveness of the steady-state window. If the boundary connectivity remains above 95% for three consecutive frames and the structural integrity score remains stable within the set range, the steady-state window is considered "effective." If boundary breaks, brightness saturation values ​​exceeding the threshold, or image structure drift occur, the current adjustment parameters are immediately closed, and the strategy is reset from the previous step. Through this verifiable, feedback-enabled, and adjustable closed-loop control process, intelligent steady-state regulation of the light input at the image acquisition end is achieved, fundamentally improving the stability of image quality and ensuring the spatiotemporal continuity of the trigger point boundary structure in the disaster chain identification task.

[0082] S600 performs a time-reversal light field traction process within a steady-state window, injects inverse phase micro-patterns corresponding to the interference light field, and jointly controls the shadow shading array and phase conjugate projection system to eliminate interference in overexposed areas of the image, maintain the continuity of the boundary image information of the trigger point, and realize the dynamic identification of high-level geological disaster chain initiation signals and improve the accuracy of risk quantitative assessment.

[0083] After establishing a steady-state window through spectral control, polarization adjustment, and exposure rhythm settings, a time-reversal light field traction process is implemented to completely suppress local image overexposure caused by high-energy light wave residue and ensure image continuity and clarity of the trigger point boundary structure. This involves precisely injecting inverse-phase micro-patterns and using a shadow shading array and phase conjugate projection path to dynamically reduce residual interference light on the image acquisition surface. The specific implementation steps include the following:

[0084] Based on image sequences acquired in real-time within a steady-state window, residual high-intensity interference regions in the light field are identified, and their optical features are inverted and reconstructed. Specifically, the brightness distribution around the trigger point in three consecutive frames is modeled, overexposed pixel clusters with brightness higher than 225 (8-bit grayscale) are extracted, and their spatial expansion rate, contrast edge diffusion range, and center wavelength frequency distribution are calculated. Spatially, the edge morphology of the overexposed pixel clusters is fitted with a high-resolution contour, and their morphological change trajectory is tracked on the time axis to quantify their center offset rate. In the frequency domain, the light wave morphology of the corresponding region in the image is subjected to Fourier transform to identify the main interference light energy concentration frequency band and dominant direction. These data form a set of structural parameters for the interference light field, including propagation path, amplitude variation, frequency energy concentration value, and typical interference direction, providing a reference for inverse phase interferometry design.

[0085] Based on the set of structural parameters of the interfering light field, an inverse phase micro-pattern with complete phase cancellation characteristics is constructed in reverse. This pattern is then projected onto a predetermined interference region in the light field propagation path to generate direction-selective phase pulling. The generation process of this inverse phase pattern consists of three steps: First, a spatiotemporal propagation model of the interfering light field is established, and the propagation path, phase delay, and incident angle of each frequency band wavefront are compositely mapped. Second, a set of inverse phase mapping maps is generated based on the propagation model, and each pixel precisely encodes the phase information propagating in the reverse direction. Finally, a spatial light control array is used to map the above pattern onto the corresponding position of the physical interference region in real time, and an inverse phase superposition interference operation is performed on the propagation plane before the incident light wave reaches the imaging surface. Through this active injection method, the original interfering wavefront is canceled by the inverse phase wave during propagation, achieving a reduction in light wave energy at a specific location in space, thereby suppressing the formation of overexposure bands on the imaging surface.

[0086] While reducing inverse phase light interference, a shadow shading array is introduced to supplement the optical path adjustment for asymmetric brightness interference caused by complex terrain reflections or incident angle fluctuations near the trigger point boundary. The shadow shading array consists of densely packed miniature variable intensity absorption units arranged in the physical space in front of the image imaging path, with each unit corresponding to a trigger point area on the image sensor's imaging surface. In implementation, an equidistant shading geometric model is first constructed based on the shape of the trigger point boundary. By calculating the angle, distance, and incident light direction difference between each shading unit and the imaging surface, the shading position and absorbance are determined. Subsequently, the units in the shading array are driven to deflect and adjust their transmittance accordingly, achieving targeted shading of high-risk areas. The shading method employs a gradual edge compression strategy, maintaining high absorption in the core shading area while transitioning to low transmittance at the edges. This significantly reduces the risk of local image saturation without disrupting the continuity of the image boundary, ensuring a stable boundary structure.

[0087] To recover boundary information that may still be damaged due to residual interference, a phase conjugate projection process is introduced based on the imaging plane image signal to dynamically reconstruct and compensate for the boundary structure of the trigger point region. Phase conjugate projection acquires the complete phase pattern of the boundary contour from the preceding image and performs alignment fitting with the missing boundary areas in the current frame's imaging information, automatically identifying the spatial deviation and morphological differences between the distorted region and the historical boundary image. Based on this, a transparent phased-array micro-light projection array is used to project the symmetrical phase pattern of the historical boundary image back onto the target location. Through real-time superposition and light wave reconstruction, the boundary contour in the current frame is reconstructed. The projection area is limited to regions in the current frame image with a boundary recognition rate below 90%, and the continuity and physical rationality of the restored boundary lines are confirmed through a bidirectional gradient check, ensuring that the completed image is structurally seamless and maintains the continuity of the trigger point's contour evolution. After this projection intervention, a final boundary recognizability test is performed on the current frame image. If the boundary connectivity rate reaches 95% or higher and the gradient fluctuation rate is less than 3%, it is judged as a high-quality acquisition frame and is eligible for subsequent risk quantitative analysis calculations.

[0088] This invention constructs a joint irradiation and albedo observation layer to accurately acquire the dynamic intensity characteristics of direct and reflected light, generating an exposure time-series fingerprint baseline to provide a highly reliable reference for subsequent image overexposure identification and processing. Furthermore, it combines coherent phase decomposition and counterfactual playback techniques to remove interference signals at the physical fluctuation level, accurately extracting the structural core of the trigger point boundary. A time-frequency consistency score is then used to establish an image stability quantification model and reconstruct continuous boundary trajectories, significantly improving the coherence and completeness of trigger point identification. Simultaneously, spectral gating and polarization rotation techniques are introduced to construct an exposure rhythm control strategy, and time-reversed light field traction and phase conjugate intervention are implemented within the steady-state window to achieve active extinguishing of overexposed areas and physical compensation of boundary images. In summary, this invention constructs a full-chain linkage evaluation mechanism from information acquisition, image recognition, risk prediction to physical interference suppression, significantly improving the accuracy and real-time performance of chain geological disaster initiation signals in high-altitude and cold regions, and enhancing the applicability and accuracy of image data in disaster early warning and risk modeling.

[0089] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A high-altitude geological disaster chain risk quantitative evaluation method suitable for high-altitude and high-cold environments, characterized in that, The method comprises the following steps: S100, a joint observation layer of irradiation and reflection is established, dynamic intensities of direct light and reflected light are collected, an exposure timing fingerprint baseline is generated, and a critical threshold of direct light intensity is determined according to the baseline; S200, coherent phase decomposition is performed based on the exposure timing fingerprint baseline, overexposure bright spot interference is separated, image information of a trigger point boundary is retained, and a phase core signal is extracted for image repair; S300, an anti-factuality playback chain is triggered according to the phase core signal, overexposure image frames are replaced by low-exposure shadow sequences, gradient continuity of the trigger point boundary is restored, and a time-frequency consistency score for trajectory reconstruction is generated; S400, image frame sequences are reconstructed based on high-value sections of the time-frequency consistency score, a double-mirror time scale is used to calibrate cross-frame time delay, and a continuously evolving trigger point boundary trajectory is obtained; S500, overexposure risks in a next time window are predicted according to the trigger point boundary trajectory, a spectral gating structure and a polarization rotating structure are linked, an exposure rhythm and an intensity threshold are set, and a steady-state window reference is formed; S600, time reversal light field traction is performed in the steady-state window, an inverse phase micro-pattern is injected, a shadow shading array and a phase conjugate projection system are combined, overexposure interference is eliminated, and continuity of the trigger point boundary is maintained.

2. The high-altitude cold-adapted high geological disaster chain risk quantitative evaluation method according to claim 1, characterized in that, Step S100 comprises: Representative observation positions around an ice-rock avalanche trigger point in a high-cold and high-altitude area are selected, dynamic intensity information of direct light and reflected light is collected; An optical imaging device with a wide dynamic response range is arranged to collect direct light intensity, and a spectral detection device with high spectral resolution is arranged to collect ice and snow reflection light intensity at different slopes and different heights; The observation data of direct light and reflected light are time-synchronized and continuously accumulated in a continuous sunshine period to generate an exposure timing fingerprint baseline that fuses direct and reflected interference characteristics; According to key nodes and slope mutation sections of light intensity changes in the exposure timing fingerprint baseline, a critical threshold of direct light intensity is calculated, and the critical threshold is used for light intensity judgment before image collection and bright spot interference area marking after image collection.

3. The high-altitude cold-adapted high geological disaster chain risk quantitative evaluation method according to claim 1, characterized in that, Step S200 comprises: Image frames are time-aligned with the exposure timing fingerprint baseline, bright spot interference areas are identified, and brightness jump boundaries are extracted; Spatial interference structure reconstruction is performed in the bright spot interference area, light intensity response structure is reconstructed based on historical light differences, and a phase interference map is generated; The phase interference map is divided into grids, phase gradient coherent areas are screened, and phase core signals with boundary extension capability are extracted; The phase core signals are arranged into a structured signal table according to image frame numbers and time labels, and are bound with the exposure timing fingerprint baseline for restoring boundary structure information in subsequent image repair.

4. The high-altitude cold-adapted high geological disaster chain risk quantitative evaluation method according to claim 3, characterized in that, The extraction of the phase core signal is subject to the condition that the direction difference of the phase gradient between adjacent grids is lower than a set threshold and shows a linear extension trend, and the extracted phase core signal is attached with a time label, a spatial coordinate, a phase consistency score, and a light disturbance identifier.

5. The high-altitude cold-adapted high geological disaster chain risk quantitative evaluation method according to claim 1, characterized in that, Step S300 comprises: According to the spatial position and time label of the phase core signal in the image sequence, image blocks under low-exposure conditions in historical images are searched, brightness mapping adjustment is performed, and image areas for replacement are generated; The image block after brightness adjustment is embedded in the overexposed area of the current image frame, and boundary smoothing is performed by using direction guidance and weighted fusion to reconstruct the gradient continuity of the trigger point boundary; The replaced image frame is evaluated in time and frequency dimensions for stability, a comprehensive time-frequency consistency score is generated, and recorded in the score table; According to the consistency score table, continuous high-score image frames are selected, structure boundary points are extracted, and trajectory curves are constructed to complete the reconstruction and quality marking of the trigger point boundary trajectory.

6. The high-altitude cold-adapted high geological disaster chain risk quantitative evaluation method according to claim 1, characterized in that, Step S400 includes: According to the time-frequency consistency score, continuous high-score image frame segments are selected, frame segments located in the high-risk area of light fluctuation are removed, and stable frame segments for image sequence reconstruction are determined; In the stable frame segment, image feature points with fixed spatial features are selected, the time deviation between frames is calibrated using a double-mirror time tag method, and the image frames are unified to a standard time axis; According to the corrected time sequence, the image frames are rearranged, the trigger point boundary contour line segment is extracted, the structural continuity is judged, and the boundary trajectory sequence is constructed; The boundary trajectory sequence is structured, the image frame number, timestamp, boundary coordinates and trajectory change characteristics are recorded, and a data set supporting time sequence modeling is formed.

7. The high-altitude cold-adapted high-geological-disaster-chain risk quantitative evaluation method according to claim 6, characterized in that, Each boundary point in the boundary trajectory sequence is attached with the time-frequency consistency score, the structural stability weight value and the trajectory continuity evaluation score of the corresponding image frame, which is used to improve the credibility screening accuracy of the trajectory point in subsequent dynamic modeling.

8. The high-altitude cold-adapted high geological disaster chain risk quantitative evaluation method according to claim 1, characterized in that, Step S500 includes: Extract the continuous evolution section of the boundary trajectory on the time axis, calculate the boundary change rate and image brightness parameter, and identify the overexposure risk time section with the trend of boundary dramatic change and the trend of rapid increase of image brightness; Based on the overexposure risk time section, activate the spectral gating adjustment structure and the polarization rotation structure, and link the control of the central wavelength, transmission direction and polarization angle of the incident light to form an optical control channel for physically suppressing light input; In the prediction risk time section, set a multi-stage exposure control strategy, segment the exposure time, aperture size and image gain, and dynamically adjust the parameters according to the image recognition rate and contrast fluctuation; A steady-state window benchmark covering the entire risk time section is constructed, and the boundary recognition connectivity and structural integrity are verified in the image frame to realize closed-loop control of optical input adjustment.

9. The high-altitude cold-adapted high geological disaster chain risk quantitative evaluation method according to claim 1, characterized in that, Step S600 includes: Construct the interference light field structure parameter set, extract the spatial profile, brightness change characteristics and frequency domain energy distribution of the overexposed area, and reversely generate the inverse phase micro-pattern with phase cancellation characteristics with the interference light field; The inverse phase micro-pattern is injected into the light wave propagation path in real time through spatial light control, and the phase interference is superimposed in front of the incident plane to reduce the interference wave energy on the imaging plane; Introduce a shadow shading array, set the light transmittance of the absorption unit according to the trigger point boundary shape and light direction, and realize directional shielding and edge light intensity gradual suppression of local image saturation risk; Through phase conjugate projection, the boundary structure is reconstructed and compensated by light waves, the boundary connectivity and gradient stability of the recovery area are detected, and it is judged whether the current frame image acquisition quality meets the admission standard.

10. The high-altitude cold-adapted high-geological-disaster-chain risk quantitative evaluation method according to claim 9, characterized in that, The injection position of the reverse phase micro-pattern is defined in a propagation plane in front of the light wave incident path of the overexposed area, the light transmittance of the absorbing unit in the shadow mask array is dynamically adjusted according to the shielding angle and the incident light intensity, the phase conjugate projection only acts on the area with a boundary connectivity rate lower than 95%, and the reconstruction is based on the symmetric phase information of the corresponding boundary in the historical image.

Images