Mountain area multi-type interference identification method and system based on time sequence segmentation algorithm

Abstract

The invention provides a mountainous area multi-type interference identification method and system based on a time sequence segmentation algorithm, and the method comprises the steps: obtaining multispectral remote sensing image data covering a target region, carrying out the preprocessing and effective pixel screening, and forming a preprocessing image set; calculating a normalized vegetation index pixel by pixel and obtaining a continuous vegetation index time sequence data set through time sequence interpolation and reconstruction; constructing an interference sample true value point sequence based on verification sample remote sensing image data; inputting the vegetation index time sequence data set into a Bayesian time sequence decomposition algorithm, decomposing the vegetation index time sequence data set into a trend component, a season component and a mutation component, and extracting a mutation point sequence; and calculating a matching rate based on the truth value point sequence and the mutation point sequence of the interference sample so as to perform adaptive optimization and correction on parameters of the Bayesian time sequence decomposition algorithm until the matching rate reaches a preset threshold value, and outputting a corrected interference detection result, thereby realizing accurate and automatic identification of the interference information in the mountainous area. And the accuracy and repeatability of interference monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of geographic information technology, and in particular to a method and system for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm. Background Technology

[0002] Mountain disturbances refer to external forces in complex mountainous terrain, caused by global climate change and human activities, that significantly impact the structure, ecological processes, and functions of ecosystems, such as landslides, debris flows, pests and diseases, road construction, and agricultural and pastoral activities. Accurately identifying the timing, extent, and type of disturbances in mountainous areas helps analyze the response processes and recovery potential of mountain ecosystems to external disturbances, providing scientific support for ecological restoration and disaster prevention in mountainous regions.

[0003] Traditional interference data relies on disaster archives and ground surveys, which suffer from problems such as delayed data updates and limited spatial coverage, making it difficult to form a long-term continuous monitoring system. Multi-source satellites such as Landsat and Sentinel provide wide-coverage, long-term remote sensing observation data, reflecting changes in surface vegetation and abrupt disturbances in mountainous areas. However, factors such as clouds, fog, and shadows in mountainous areas can easily lead to noise and missing data, reducing data reliability. While existing time-series segmentation and abrupt change detection algorithms can extract some interference, they lack standardized procedures in mountainous applications, lack complete specifications from data preprocessing to abrupt change extraction, and lack consistency verification mechanisms for abrupt change detection results. Key algorithm parameters rely on empirical settings, making it difficult to meet the high-precision and engineering requirements for interference identification in mountainous areas. Summary of the Invention

[0004] This invention provides a method and system for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm, which enables accurate and automated identification of the time, range, and type of interference in mountainous areas, thereby improving the accuracy and repeatability of interference monitoring.

[0005] On the one hand, this invention provides a method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm, which includes: Acquire multispectral remote sensing image data covering the target area, and preprocess and filter effective pixels of the multispectral remote sensing image data to form a preprocessed image set; wherein the temporal duration of the multispectral remote sensing image data is longer than a preset duration; Based on the preprocessed image set, the normalized vegetation index is calculated pixel by pixel and a continuous vegetation index time-series dataset is obtained by temporal interpolation and reconstruction. A sequence of ground truth points for interference samples is constructed based on remote sensing image data of validation samples; the sequence of ground truth points for interference samples includes information on the time, spatial location, and type of interference. The vegetation index time series dataset is input into the Bayesian time series decomposition algorithm, which decomposes it into trend components, seasonal components and mutation components, and extracts the mutation point sequence as the preliminary interference identification result. The matching rate is calculated based on the ground truth sequence of the interference samples and the mutation sequence. The parameters of the Bayesian temporal decomposition algorithm are adaptively optimized and corrected according to the matching rate until the matching rate reaches a preset threshold. The corrected interference detection result is then output.

[0006] On the other hand, the present invention also provides a multi-type interference identification system for mountainous areas based on a time-series segmentation algorithm, which includes: The preprocessing module is used to acquire multispectral remote sensing image data covering the target area, and to preprocess and filter the effective pixels of the multispectral remote sensing image data to form a preprocessed image set; wherein the temporal duration of the multispectral remote sensing image data is longer than a preset duration. The reconstruction module is used to calculate the normalized vegetation index pixel by pixel based on the preprocessed image set and obtain a continuous vegetation index time-series dataset through temporal interpolation and reconstruction. The construction module is used to construct a sequence of ground truth points for interference samples based on the remote sensing image data of the validation samples; the sequence of ground truth points for interference samples includes information on the time, spatial location, and type of interference. The identification module is used to input the vegetation index time series dataset into the Bayesian time series decomposition algorithm, decompose it into trend components, seasonal components and mutation components, and extract the mutation point sequence as the preliminary interference identification result; The adaptive optimization module is used to calculate the matching rate based on the ground truth sequence of the interference samples and the mutation point sequence, and to adaptively optimize and correct the parameters of the Bayesian time series decomposition algorithm according to the matching rate until the matching rate reaches a preset threshold, and output the corrected interference detection result.

[0007] The present invention provides a method and system for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm. Through multi-temporal remote sensing image preprocessing and effective pixel screening, pixel-by-pixel vegetation index time-series reconstruction and continuous processing, high-precision interference sample ground value point sequence construction, Bayesian time-series decomposition-driven interference event identification, and matching rate-driven parameter adaptive optimization and correction, the method achieves accurate and automated identification of the time, range, and type of interference in mountainous areas. It solves the problems of non-standard procedures, lack of verification, and parameter dependence on experience in traditional methods, and improves the accuracy and repeatability of interference monitoring. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0009] Figure 1 This is a flowchart illustrating the method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm provided in an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the amount of data affecting the Landsat multispectral data used in this invention; Figure 3 This is a schematic diagram of interference detection at a typical sample point in a certain study area; Figure 4 This is a spatial distribution map of the number of interference events occurring in the study area; Figure 5 This is a schematic diagram of the structure of a multi-type interference identification system for mountainous areas based on a time-series segmentation algorithm provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0011] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0012] Figure 1 This is a flowchart illustrating the method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm provided in an embodiment of the present invention.

[0013] like Figure 1 As shown, the method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm provided in this embodiment of the invention mainly includes the following steps: 101. Acquire multispectral remote sensing image data covering the target area, and preprocess and filter effective pixels of the multispectral remote sensing image data to form a preprocessed image set; Multispectral remote sensing image data can be image data with multiple bands acquired through satellite or airborne platforms, and its main function is to provide multi-dimensional information on the surface reflectance characteristics. For example, such data can be acquired through the Landsat series satellites. The temporal duration of the multispectral remote sensing image data is longer than a preset duration to obtain sufficient multispectral remote sensing image data. In this embodiment, Landsat multispectral data can be used as the data source. Figure 2 This is a schematic diagram illustrating the amount of influence data corresponding to the Landsat multispectral data used in this invention, as shown below. Figure 2 As shown, a total of 2749 Landsat images covering the target area from 2000 to 2019 can be obtained, providing a sufficient observational basis for long-term interference monitoring. Among them, Figure 2 The image shows images obtained by different sensor types (such as LCO8 / LE07 / LT05, etc.) in different years.

[0014] Preprocessing can employ radiometric correction, geometric correction, or other methods to eliminate systematic errors in images, while effective pixel selection can be achieved by setting observation quality thresholds or removing outliers, with the aim of ensuring data quality for subsequent analysis.

[0015] In a specific implementation process, this step can be achieved in the following way: The multispectral remote sensing image is sequentially subjected to radiometric calibration, geometric correction, and atmospheric correction to obtain a preprocessed image. A masking algorithm is then used to mask noise factors in the preprocessed image to obtain a masking result. For each pixel, based on the masking result, the number of valid observations in the entire time series is counted. The number of valid observations is compared with a preset temporal integrity threshold. Target pixels with a number of valid observations greater than or equal to the temporal integrity threshold are selected, and the valid observations in the target pixels are included in the preprocessed image set.

[0016] Radiometric calibration refers to correcting sensor response biases to ensure image data reflects the true surface radiation characteristics. This can be achieved using linear or nonlinear calibration models, aiming to eliminate systematic errors inherent in the sensor itself. Geometric calibration involves registering image data with geographic coordinates to maintain spatial consistency. This can be achieved using polynomial or projection transformation methods, ensuring comparability of images acquired at different times. Atmospheric calibration removes the effects of atmospheric scattering and absorption to ensure image data reflects the true surface reflectance. This can be achieved using MODTRAN or 6S models, aiming to improve the radiometric consistency of image data. Masking algorithms are techniques for identifying and masking unreliable pixels based on spectral features. This can be implemented using threshold segmentation or machine learning classification methods, aiming to eliminate interference factors such as clouds, snow, fog, and shadows. The effective observation count refers to the number of reliable observations a single pixel has passed quality checks over long-term observations. This can be achieved through statistical analysis or time series modeling methods, aiming to quantify data reliability. The temporal integrity threshold is the minimum number of observations required to ensure data continuity. It can be dynamically adjusted based on typical observation conditions in mountainous areas, aiming to ensure that selected pixels have sufficient continuous observation capabilities.

[0017] Specifically, firstly, by sequentially performing radiometric calibration, geometric correction, and atmospheric correction on multispectral remote sensing images, sensor response bias and atmospheric scattering effects were eliminated, ensuring consistency in radiometric and geometric dimensions across image data acquired at different times, laying the foundation for subsequent noise processing. Secondly, a masking algorithm was employed to identify and shield interfering factors such as clouds, snow, fog, and shadows, preventing noise from directly contaminating vegetation information and significantly improving the purity of the raw data. Building upon this, the number of valid observations for each pixel across all time series was statistically analyzed, objectively reflecting the severity of data loss in the complex mountainous environment. High-reliability pixels were then selected by comparing this number with a temporal integrity threshold. Finally, the selected pixels were incorporated into the preprocessed image set, ensuring the dataset's dual integrity in terms of temporal length and data quality, providing a stable and reliable data source for subsequent vegetation index calculations and mutation detection.

[0018] In this implementation, the target area is a mountainous region with complex terrain and variable climate. The proportion of effective observations of each pixel in the remote sensing images collected between 2000 and 2019 varies greatly. Therefore, sample points with a sample size of less than 226 (i.e., half the total number of time series points) were deleted to improve the robustness of the data analysis.

[0019] 102. Based on the preprocessed image set, the normalized vegetation index is calculated pixel by pixel and a continuous vegetation index time-series dataset is obtained through temporal interpolation and reconstruction. Specifically, the normalized vegetation index can be calculated pixel by pixel based on the preprocessed image set; and the normalized vegetation index can be interpolated and reconstructed using the moving median filtering method to obtain the vegetation index time series dataset.

[0020] In detail, after obtaining the preprocessed image set, the Normalized Difference Vegetation Index (NDVI) can be calculated for each pixel using the near-infrared and red reflectance data of the image. This index is a core indicator reflecting the vegetation growth status and changes in land cover. It can effectively distinguish between vegetated and non-vegetated areas and is sensitive to changes in vegetation after disturbance, providing key feature basis for disturbance identification.

[0021] Because remote sensing data from mountainous areas is easily affected by factors such as clouds and fog, the calculated NDVI time-series data may contain missing or outlier values. A moving median filtering method is used for time-series interpolation and reconstruction. This method sets a specific time window, takes the median NDVI value within the window as the current NDVI value, fills in missing values, and corrects outliers. Its advantage lies in not over-smoothing the time-series curve, effectively removing noise while preserving the original data trend and abrupt change characteristics, making the NDVI time-series data continuous and reliable, ultimately forming a vegetation index time-series dataset that meets the requirements for interference identification.

[0022] In this embodiment, the moving median filtering method is calculated as follows: .

[0023] Where median{} represents the median operation, and k is the window radius (set to 2). Let be the value of the normalized vegetation index at time t. This represents the median of NDVI within a sliding window centered at t and with radius k. Moving median filtering is a non-linear smoothing method used to handle noise and missing values ​​in time-series data, without significantly altering the original data curve characteristics.

[0024] 103. Construct a sequence of ground truth points for interference samples based on remote sensing image data of validation samples; In a specific implementation process, the verification sample remote sensing image data can be derived from high-resolution satellite imagery or UAV aerial imagery. The resolution requirement ensures accurate labeling of the spatial location of the interference event. This can be achieved using historical high-resolution remote sensing imagery from platforms such as WorldView, GF-2 / 6, and Google Earth, with the aim of providing a high-precision time-series comparison basis. Specifically, the construction of the ground truth sequence of the interference sample can be generated through manual annotation, automated classification, or other methods. Its core is to record the time, location, and type information of the interference event, thereby providing a benchmark for subsequent verification. The ground truth sequence of the interference sample includes information on the time, spatial location, and type of interference.

[0025] This step can be implemented as follows: Based on the verification sample remote sensing image data, identify areas where the change value of the target object is greater than the preset change value by using a time-series comparison method, and extract interference events; label the occurrence time, spatial location and interference type of each interference event to form a sample ground truth point sequence; merge the points in the sample ground truth point sequence where multiple interferences occur at the same location to construct an interference sample ground truth point sequence with unique spatial location and multi-attribute information.

[0026] In a specific implementation process, when multiple disturbances are detected at the same location and the interval between adjacent disturbances is less than the typical recovery cycle of the vegetation ecosystem, the multiple disturbance records are merged into a single composite disturbance event. This composite disturbance event is assigned a unique spatial location identifier, its occurrence time is marked as the occurrence time of the earliest disturbance, and its disturbance type is marked as the type of the strongest disturbance.

[0027] Specifically, time-series comparison refers to the technique of comparing and analyzing remote sensing images of the same geographical location at different time points to identify areas of significant change in a target object (such as vegetation or land surface). This can be achieved using pixel-level difference analysis or spectral feature change detection methods, aiming to accurately capture abrupt changes in land surface characteristics. Disturbance events refer to external actions caused by natural or human factors that significantly impact the structure, processes, and functions of an ecosystem. These can include landslides, debris flows, pests and diseases, road construction, etc., with the aim of comprehensively covering potential sources of disturbance. Merging processing refers to the technique of logically integrating multiple disturbance events occurring at the same spatial location. This can be achieved through time interval judgment and event attribute merging algorithms, aiming to eliminate redundant records and ensure the ecological rationality of the true value sequence. Composite disturbance events refer to the technical concept of merging multiple disturbance records within a short time interval into a single event. This can be achieved by setting time thresholds and disturbance intensity priority rules, aiming to accurately reflect the superimposed effects of disturbances and their comprehensive impact on the ecosystem.

[0028] In practical applications, a time-series comparison method was used to identify areas of significant vegetation or surface changes, extracting the locations of natural or man-made disturbance events such as pests and diseases, hurricanes, logging, road construction, building construction, and mining. For each disturbance event, the image change characteristics and spatial coordinate information before and after its occurrence were recorded. The time, spatial location, and type of disturbance were labeled for each extracted disturbance event, resulting in a total of 4387 disturbance event records. Considering that some pixels experienced multiple disturbances during the monitoring period, points experiencing two to three disturbances at the same location were merged, ultimately forming 3580 sample points with unique spatial locations and multi-attribute information, which served as the ground truth sample set for disturbance identification.

[0029] 104. Input the vegetation index time series dataset into the Bayesian time series decomposition algorithm to decompose it into trend components, seasonal components and mutation components, and extract the mutation point sequence as the preliminary interference identification result; In a specific implementation, the Bayesian time-series decomposition algorithm decomposes vegetation index time-series data to separate trend changes, seasonal fluctuations, and abrupt change signals, i.e., trend components, seasonal components, and abrupt change components. Furthermore, the extraction of abrupt change sequence can be achieved by setting a threshold for the magnitude of change, conducting statistical significance tests, or using other methods, with the aim of capturing structural changes that may correspond to disturbance events.

[0030] This step can be achieved as follows: extract the nested mutation points from the trend component and the seasonal component to form a candidate mutation point set; combine the mutation intensity, occurrence probability and confidence interval corresponding to the nested mutation points to perform a preliminary screening of the candidate mutation point set, and obtain the mutation point sequence as the preliminary interference identification result.

[0031] In a specific implementation, the candidate mutation point set refers to the initial set formed after capturing potential mutation locations from the trend component and the seasonal component. In practical applications, the trend component is used to reflect long-term ecological evolution characteristics, while the seasonal component is used to capture periodic vegetation fluctuation information. The combination of the two can comprehensively cover two key interference signals: long-term trend mutations and seasonal anomalies. The formation of the candidate mutation point set can be achieved by using a time-series decomposition algorithm to extract nested mutation points, the purpose of which is to avoid the omission of interference events caused by single-component analysis.

[0032] Specifically, mutation intensity refers to a quantitative indicator of the magnitude of change in a mutation point, used to filter out weak mutation points with changes below a threshold, thereby eliminating short-term fluctuations or noise interference. Occurrence probability is an indicator assessing the likelihood of a mutation event occurring, used to eliminate low-confidence candidate points and ensure the reliability of the screening results. Confidence interval refers to the range of mutation intensity values ​​calculated using statistical methods, used to verify whether the change in a mutation point is statistically significant. The introduction of these indicators can be achieved through techniques such as joint decision domain construction and statistical significance testing, with the aim of improving the rigor and scientific nature of mutation point screening.

[0033] In a specific implementation, the formula for the Bayesian temporal decomposition algorithm is as follows: .

[0034] Where T represents the trend component and S represents the seasonal component. and These are the abrupt change points embedded in the trend component and the seasonal component, respectively, including their number and location, where ε is a variable with a mean of 0 and a variance of 0. The Gaussian random error term, i.e. , This represents the time series value of the vegetation index at time t.

[0035] Specifically, Figure 3 This is a schematic diagram of interference detection at typical sample points in a certain study area. Figure 3 As can be seen from part a), the left image (August 2004) shows that the area was still mainly natural grassland, while the right image (July 2015) clearly shows newly added buildings and road facilities, indicating that a significant human disturbance process occurred at this sampling point between the two time points.

[0036] Figure 3 b) Figure 3 f) shows the results of the multi-component decomposition and mutation probability analysis of this sample point, clearly presenting five key curves and their corresponding relationships: Normalized Difference Vegetation Index (NDVI) time-series value curve (denoted as NDVI in the figure): As the original data curve, it comprehensively reflects the actual state of vegetation growth. Figure 3 As can be seen in b), the NDVI time series showed a significant step drop around June 9, 2007, suddenly shifting from a relatively stable high value to a low value range. This abrupt change reflects the rapid degradation of vegetation cover, is a direct representation of the occurrence of the disturbance event, and is also the basis for subsequent component decomposition analysis. Seasonal component curve (denoted as "season" in the figure): A periodic fluctuation curve after stripping away the long-term trend, reflecting the natural growth pattern of vegetation with seasonal changes. From Figure 3c) It can be seen that the seasonal components maintain a stable annual cycle oscillation characteristic as a whole, and no obvious disruption of the cycle structure occurs before and after the mutation.

[0037] The probability curve of seasonal component mutation points (denoted as Pr(spr) in the figure): used to quantify the confidence level of the existence of seasonal mutations at each time point. From Figure 3 As can be seen from d), the probability of seasonal component mutation remains at a low level throughout the entire time series, indicating that no significant seasonal structural mutations have occurred at this sampling point, and the interference is mainly reflected in the long-term trend. Trend component curve (denoted as trend in the figure): from Figure 3 As can be seen in e), the trend component underwent a significant abrupt change around June 9, 2007, dropping sharply from a previously relatively stable high NDVI state to a low value range, and then entered a slow recovery phase. This indicates that this time point corresponds to a strong exogenous disturbance process, which has had a substantial impact on the long-term evolution trajectory of vegetation. The probability curve of trend-based component mutation points (denoted as Pr(tpr) in the figure): quantifies the confidence level of trend-based mutations at each time point. From Figure 3 As can be seen from f), the probability of the trend component mutation shows a significant peak at the corresponding position on June 9, 2007, while the probability values ​​are lower in other time periods, further verifying that the sample point only experienced a significant trend mutation at that moment.

[0038] based on Figure 3 a) On-site image comparison and Figure 3 The multidimensional quantitative results from b) to 3f) clearly show that this sample point has only one significant trend abrupt change (June 9, 2007). This abrupt change is simultaneously manifested as a step drop in the original NDVI sequence, a structural transition in the trend component, and a significant peak in the trend abrupt change probability. After screening candidate abrupt change points by combining the mutation intensity, occurrence probability, and confidence interval, June 9, 2007, was finally determined as the only valid interference mutation time for this sample point. This result provides a reliable benchmark for subsequent ground truth matching of interference samples and optimization of algorithm parameters.

[0039] 105. Calculate the matching rate based on the ground truth sequence of the interference samples and the mutation point sequence, and adaptively optimize and correct the parameters of the Bayesian time series decomposition algorithm according to the matching rate until the matching rate reaches a preset threshold, and output the corrected interference detection result.

[0040] In a specific implementation process, parameter adaptive optimization and correction can be achieved through grid search, genetic algorithm or other optimization strategies, with the aim of dynamically adjusting algorithm parameters to improve detection accuracy.

[0041] Specifically, a quantitative verification mechanism is established by calculating the time deviation between detected mutation points and sample ground truth points, and combining this with a preset time tolerance to determine the matching status. When the matching rate does not reach a preset threshold, key parameters of the Bayesian time series decomposition algorithm are dynamically adjusted, including the range of seasonal mutation points and the range of trend mutation points, until the matching rate meets the requirements. This forms a closed-loop feedback mechanism, enabling the algorithm to automatically calibrate detection sensitivity according to different mountain ecological types, and ultimately output verified and corrected interference detection results, ensuring that the recognition accuracy meets the needs of practical applications. Figure 4 This is a spatial distribution map of the number of disturbances occurring in the study area. For example... Figure 4 As shown in the figure, the values ​​(0-20) represent the cumulative number of interference events occurring at the corresponding latitude and longitude locations during the monitoring period. The higher the value, the more frequent the interference. In this way, the spatial pattern of interference can be presented intuitively, providing a direct reference for the delineation of key areas for interference prevention and control in mountainous areas and for ecological restoration planning. Figure 4 The horizontal axis represents longitude, and the vertical axis represents latitude.

[0042] In detail, when calculating the matching rate based on the interference sample ground truth point sequence and the mutation point sequence, it can be achieved in the following way: calculate the time deviation between each pair of detected mutation points and sample ground truth points; preset the time tolerance, and if the time deviation is less than or equal to the time tolerance, it is determined to be a matching point; count the number of all matching points and the number of sample ground truth points, and calculate the matching rate.

[0043] Specifically, time bias refers to the difference in time between the detected mutation point and the true value point of the sample, which can be achieved by directly calculating the absolute difference between their timestamps. In practical applications, time bias can be calculated in various ways, such as direct difference calculation based on timestamps or difference calculation based on time series indexes. The purpose is to quantify the time difference between the algorithm's detection result and the actual interference event, thereby providing a basis for subsequent matching judgment.

[0044] The time tolerance can be defined as the allowable range of time error, which can be set based on the typical recovery cycle of vegetation ecosystems or the update characteristics of remote sensing data. In practical applications, the time tolerance can be set using a fixed threshold method, a dynamic adjustment method, or a method based on statistical distribution. The purpose is to adapt to the temporal dynamic characteristics of disturbance events in mountainous areas and avoid reasonable deviations being misjudged as mismatches due to strict time point matching.

[0045] Furthermore, the matching rate refers to the ratio of the number of matching points to the number of true value points in the sample. It can be calculated by statistically analyzing the number of matching points and combining this with the total number of true value points in the sample. In practical applications, the matching rate can be calculated using a simple proportional method, or a weighted statistical method can be introduced to improve the calculation accuracy. The purpose is to provide quantifiable feedback indicators for parameter optimization, ensuring the objectivity and representativeness of the optimization process.

[0046] In detail, the first step is to obtain the sequence of mutation points identified by the Bayesian time series decomposition algorithm. With the sequence of sample truth points ,in This represents the time for detecting the i-th mutation point. This represents the time of the true value point of the j-th sample.

[0047] The second step is to calculate the time deviation between each pair of detected mutation points and the true value points. The calculation formula is as follows: .

[0048] Step 3: Set a time tolerance of six months. If the time interval is less than or equal to 0.5 years, the detected mutation point is considered to match the true value point. If the time difference is greater than 0.5 years, it is considered a mismatch. In this embodiment, the deviation between a mutation point identified in a sample pixel and the actual sample ground truth point is 20 days, which is within the allowable time range. The identified trend mutation point is judged to have no interference due to its low intensity, which is consistent with the field verification results.

[0049] Step 4: Analyze the matching results of all detected mutation points and ground truth points, and calculate the matching rate R. The calculation formula is as follows: .

[0050] in, Indicates the number of matching points, This indicates the number of truth points.

[0051] Furthermore, adaptively optimizing and correcting the parameters of the Bayesian time series decomposition algorithm based on the matching rate until the matching rate reaches a preset threshold may include: when the matching rate is lower than the preset threshold, adjusting the key parameters of the Bayesian time series decomposition algorithm; the key parameters include at least one of the following: the range of seasonal mutation points, the range of trend mutation points, the minimum length of seasonal segments, the minimum length of trend segments, and the mutation intensity threshold; after updating the parameters, repeating the Bayesian time series decomposition and matching rate calculation until the matching rate reaches the preset threshold.

[0052] The range of seasonal mutation points can be a constrained interval for the number of mutation points that may exist in the seasonal component. It can be achieved through dynamic programming or heuristic search algorithms, with the aim of improving the robustness of seasonal feature extraction.

[0053] The range of trend change points refers to the range within which the number of possible change points in a trend component is limited. This can be achieved using Bayesian inference or maximum likelihood estimation methods, with the aim of enhancing the accuracy of trend change detection.

[0054] The minimum length of seasonal segments refers to a parameter that constrains the minimum duration of each segment in a seasonal component. It can be achieved through a sliding window or piecewise regression technique, with the aim of avoiding noise interference caused by over-segmentation.

[0055] The minimum length of a trend segment can be a parameter that limits the minimum duration of each segment in a trend component. It can be implemented using piecewise linear fitting or dynamic time warping algorithms, with the aim of ensuring the significance of trend changes.

[0056] The mutation intensity threshold is the lower limit value of the intensity used to screen mutation points. It can be achieved through statistical distribution analysis or empirical threshold setting methods. Its purpose is to eliminate weak change points to improve detection accuracy.

[0057] In this embodiment, the key parameters cover at least one of the following: the range of seasonal mutation points, the range of trend mutation points, the minimum length of seasonal segments, the minimum length of trend segments, and the mutation intensity threshold. These parameters collectively determine the detection sensitivity and stability of mutation points in the time series decomposition. By selectively adjusting these parameters, the mutation characteristics of vegetation index time series under complex mountainous terrain can be accurately adapted. After the parameters are updated, Bayesian time series decomposition and matching rate calculation are repeatedly executed to form an iterative optimization process, enabling the algorithm to continuously correct the parameter combination based on real-time feedback until the matching rate reaches the preset threshold. This adaptive correction mechanism based on the matching rate not only improves the accuracy of mutation point detection but also enhances the method's adaptability to different mountainous environments, shifting the interference identification process from a static empirical mode to a dynamic feedback mode.

[0058] Furthermore, adjusting the key parameters of the Bayesian time series decomposition algorithm can include: A hierarchical adaptive parameter tuning mechanism is adopted to adjust the key parameters of the Bayesian time series decomposition algorithm; The hierarchical adaptive parameter tuning mechanism includes: (1) Select representative samples in the target area for parameter sensitivity analysis to determine the initial parameter range; Specifically, within the target area, a stratified random sampling method can be used to select a spatially representative set of sample points from the sequence of ground truth points of the interference samples based on at least one of the following factors: land type, elevation gradient, vegetation cover, and historical interference frequency. Based on the vegetation index time series data of the sample point set, the sensitivity analysis of the key parameters of the Bayesian time series decomposition algorithm is carried out using grid search and cross-validation methods to determine the initial parameter range. The sensitivity analysis process includes: The values ​​of each key parameter are changed sequentially, the change in the matching rate of the sample point set after each change is recorded, and the sensitivity of each key parameter to the matching rate is calculated. Key parameters with sensitivity greater than the preset sensitivity are selected as parameters to be tuned, and the initial parameter range of each parameter to be tuned is determined by combining the physical meaning of the parameter with the empirical range.

[0059] In detail, a stratified random sampling method can be used to select a set of sample points within the target area in the mountainous region. When sampling, one or more factors such as land type (e.g., forest, grassland, cultivated land), altitude gradient (high altitude, medium altitude, low altitude), vegetation cover (high cover, medium cover, low cover) and historical interference frequency (high frequency, medium frequency, low frequency) should be comprehensively considered to ensure that the selected sample points can uniformly cover the different environmental characteristics and interference backgrounds of the target area, have sufficient spatial representativeness, and avoid inaccurate sensitivity analysis results caused by sample bias.

[0060] Based on time-series vegetation index data from a selected set of sample points, parameter sensitivity analysis was conducted using a grid search method and cross-validation. The grid search method divided the possible values ​​of each key parameter into several grid nodes, traversing all node combinations. Cross-validation divided the sample data into training and validation sets, evaluating the performance of different parameter combinations through multiple training and validation iterations. This combination of methods allows for a comprehensive and efficient analysis of the impact of each key parameter.

[0061] During the sensitivity analysis, the values ​​of each key parameter are changed sequentially while keeping other parameters constant. The change in the matching rate of the sample point set after each parameter change is recorded. By calculating the ratio of the change in matching rate to the change in parameter, the sensitivity of each key parameter to the matching rate is obtained. A preset sensitivity threshold is set, and key parameters with sensitivity higher than the threshold are selected as parameters to be optimized (these parameters have a significant impact on the identification results and are the focus of parameter adjustment). Combining the physical meaning of these parameters to be optimized (such as the mutation intensity threshold reflecting the minimum intensity of interference) and industry experience range, unreasonable value ranges are eliminated, and the initial parameter range of each parameter to be optimized is finally determined.

[0062] (2) Based on the initial parameter range, perform zoning optimization according to differences in land type or ecological type; Specifically, considering the significant differences in land types (such as forest land, cultivated land, and construction land) or ecological types (such as forest ecosystems and grassland ecosystems) across different mountainous areas, their disturbance characteristics (such as disturbance frequency, disturbance intensity, and recovery speed) also vary. Using globally uniform parameters is insufficient to adapt to all areas. Therefore, based on the initial parameter range determined in the first step, the target area is divided into multiple sub-regions according to land type or ecological type. For each sub-region, key parameters are adjusted based on its disturbance characteristics and data features, resulting in regional optimization so that the parameters for each sub-region can be adapted to its specific environment and disturbance characteristics.

[0063] (3) Perform local fine-tuning on pixels whose consistency test error after partition optimization is greater than the preset error.

[0064] After the zonal optimization is completed, a consistency check is performed on the pixels in each sub-region, and the error between the recognition result of each pixel and the sample ground truth point is calculated. For pixels whose consistency check error is greater than the preset error, it indicates that the environmental characteristics or interference of the pixel are relatively special, and the unified parameters of the zonal still cannot meet its recognition accuracy requirements. Local fine-tuning needs to be performed on these individual pixels to further optimize the parameters and ensure that the recognition result of each pixel can achieve a high accuracy.

[0065] In this embodiment, the hierarchical adaptive parameter tuning mechanism fully combines the heterogeneous characteristics of mountainous land or ecological types. It achieves parameter adaptation to regional characteristics through partitioned tuning, and then solves the problem of recognition accuracy of special pixels through local fine-tuning, thereby improving the overall accuracy of interference recognition and the robustness of the algorithm. At the same time, the hierarchical strategy avoids blind tuning of the entire region, reduces the computational load of parameter tuning, improves tuning efficiency, and makes the algorithm more practical in large-scale mountainous remote sensing data processing.

[0066] Based on the same general inventive concept, this invention also protects a mountain multi-type interference identification system based on a time-series segmentation algorithm. The mountain multi-type interference identification system based on a time-series segmentation algorithm provided by this invention will be described below. The mountain multi-type interference identification system based on a time-series segmentation algorithm described below can be referred to in correspondence with the mountain multi-type interference identification method based on a time-series segmentation algorithm described above.

[0067] Figure 5 This is a schematic diagram of the structure of a multi-type interference identification system for mountainous areas based on a time-series segmentation algorithm provided in an embodiment of the present invention, as shown below. Figure 5 As shown, the mountain multi-type interference identification system based on time-series segmentation algorithm in this embodiment includes a preprocessing module 51, a reconstruction module 52, a construction module 53, an identification module 54, and an adaptive optimization module 55.

[0068] The preprocessing module 51 is used to acquire multispectral remote sensing image data covering the target area, and to preprocess and filter effective pixels of the multispectral remote sensing image data to form a preprocessed image set; wherein the temporal duration of the multispectral remote sensing image data is greater than a preset duration. Reconstruction module 52 is used to calculate the normalized vegetation index pixel by pixel based on the preprocessed image set and obtain a continuous vegetation index time series dataset through temporal interpolation and reconstruction. Construction module 53 is used to construct a sequence of ground truth points for interference samples based on the remote sensing image data of the verification samples; the sequence of ground truth points for interference samples includes information on the time of interference occurrence, spatial location, and type of interference; The identification module 54 is used to input the vegetation index time series dataset into the Bayesian time series decomposition algorithm to decompose it into trend components, seasonal components and mutation components, and extract the mutation point sequence as the preliminary interference identification result. The adaptive optimization module 55 is used to calculate the matching rate based on the ground truth sequence of the interference samples and the mutation point sequence, and to adaptively optimize and correct the parameters of the Bayesian time series decomposition algorithm according to the matching rate until the matching rate reaches a preset threshold, and output the corrected interference detection result.

[0069] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device may include: a processor 610, a communication interface 620, a memory 630, and a communication bus 640. The processor 610, communication interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute a method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm.

[0070] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0071] It should be noted that all relevant information that may be involved in the various embodiments of the present invention is processed in strict accordance with the requirements of laws and regulations, following the principles of legality, legitimacy, and necessity, based on the reasonable purpose of the business scenario, and is information that users actively provide or generate during the use of the product / service, as well as information obtained with user authorization.

[0072] The information processed by this invention may vary depending on the specific product / service scenario and should be based on the specific scenario in which the user uses the product / service. This may involve user account information, device information, or other related information. This invention will treat the relevant information and its processing with the utmost diligence.

[0073] This invention places great emphasis on the security of relevant information and has adopted reasonable and feasible security protection measures that comply with industry standards to protect user information and prevent unauthorized access, public disclosure, use, modification, damage or loss of relevant information.

[0074] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0075] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying multiple types of interference in mountainous areas based on a temporal segmentation algorithm, characterized in that, include: Acquire multispectral remote sensing image data covering the target area, and preprocess and filter effective pixels of the multispectral remote sensing image data to form a preprocessed image set; wherein the temporal duration of the multispectral remote sensing image data is longer than a preset duration; Based on the preprocessed image set, the normalized vegetation index is calculated pixel by pixel and a continuous vegetation index time-series dataset is obtained by temporal interpolation and reconstruction. A sequence of ground truth points for interference samples is constructed based on remote sensing image data of validation samples; the sequence of ground truth points for interference samples includes information on the time, spatial location, and type of interference. The vegetation index time series dataset is input into the Bayesian time series decomposition algorithm, which decomposes it into trend components, seasonal components and mutation components, and extracts the mutation point sequence as the preliminary interference identification result. The matching rate is calculated based on the ground truth sequence of the interference samples and the mutation sequence. The parameters of the Bayesian temporal decomposition algorithm are adaptively optimized and corrected according to the matching rate until the matching rate reaches a preset threshold. The corrected interference detection result is then output.

2. The method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm according to claim 1, characterized in that, The multispectral remote sensing image data is preprocessed and effective pixel is filtered to form a preprocessed image set, including: The multispectral remote sensing images are sequentially subjected to radiometric calibration, geometric correction and atmospheric correction to obtain the preprocessed images; A masking algorithm is used to mask the noise factors in the preprocessed image to obtain the masking result. For each pixel, based on the masking results, the number of valid observations in the entire time series is counted. The number of valid observations is compared with a preset temporal integrity threshold; Target pixels with a number of valid observations greater than or equal to the temporal integrity threshold are selected, and the valid observations in the target pixels are included in the preprocessed image set.

3. The method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm according to claim 1, characterized in that, Normalized vegetation indices are calculated pixel-by-pixel, and a continuous temporal dataset of vegetation indices is obtained through temporal interpolation and reconstruction, including: Based on the preprocessed image set, the normalized vegetation index is calculated pixel by pixel; The normalized vegetation index was interpolated and reconstructed using the moving median filtering method to obtain the time series dataset of the vegetation index.

4. The method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm according to claim 1, characterized in that, A sequence of ground truth points for interference samples was constructed based on the validation sample remote sensing image data, including: Based on the validation sample remote sensing image data, the method of comparing the time series before and after is used to identify areas where the change value of the target object is greater than the preset change value, and to extract interference events; Each interference event is labeled with its occurrence time, spatial location, and interference type to form a sequence of sample ground truth points; Points that are disturbed multiple times at the same location in the sample ground truth point sequence are merged to construct a disturbed sample ground truth point sequence with unique spatial location and multi-attribute information.

5. The method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm according to claim 1, characterized in that, The mutation point sequence was extracted as a preliminary result of interference identification, including: Extract the nested mutation points from the trend component and seasonal component to form a candidate mutation point set; By combining the mutation intensity, occurrence probability and confidence interval corresponding to the nested mutation points, the candidate mutation point set is initially screened to obtain the mutation point sequence as the preliminary interference identification result.

6. The method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm according to claim 1, characterized in that, The matching rate is calculated based on the ground truth sequence of the interference samples and the mutation sequence, including: Calculate the time deviation between each pair of detected mutation points and the sample true value point; If the time deviation is less than or equal to the preset time tolerance, it is determined to be a matching point; Count the total number of matching points and the total number of true points in the sample, and calculate the matching rate.

7. The method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm according to claim 1, characterized in that, The parameters of the Bayesian time series decomposition algorithm are adaptively optimized and corrected based on the matching rate until the matching rate reaches a preset threshold, including: When the matching rate is lower than a preset threshold, the key parameters of the Bayesian time series decomposition algorithm are adjusted; the key parameters include at least one of the following: the range of seasonal mutation points, the range of trend mutation points, the minimum length of seasonal segments, the minimum length of trend segments, and the mutation intensity threshold. After updating the key parameters, repeat the Bayesian temporal decomposition and matching rate calculation until the matching rate reaches the preset threshold.

8. The method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm according to claim 7, characterized in that, Adjusting key parameters of the Bayesian time series decomposition algorithm, including: A hierarchical adaptive parameter tuning mechanism is adopted to adjust the key parameters of the Bayesian time series decomposition algorithm; The hierarchical adaptive parameter tuning mechanism includes: Representative samples were selected from the target area for parameter sensitivity analysis to determine the initial parameter range; Based on the initial parameter range, zoning optimization is performed according to differences in land type or ecological type; For pixels whose consistency test error after partition optimization is greater than the preset error, perform local fine-tuning.

9. The method for identifying multiple types of interference in mountainous areas based on a time-series segmentation algorithm according to claim 8, characterized in that, Representative samples were selected from the target region for parameter sensitivity analysis to determine the initial parameter range, including: Within the target area, a stratified random sampling method is used to select a spatially representative set of sample points from the sequence of ground truth points of the interference samples based on at least one of the following factors: land type, elevation gradient, vegetation cover, and historical interference frequency. Based on the vegetation index time series data of the sample point set, the sensitivity analysis of the key parameters of the Bayesian time series decomposition algorithm is carried out using grid search and cross-validation methods to determine the initial parameter range. The sensitivity analysis process includes: The values ​​of each key parameter are changed sequentially, the change in the matching rate of the sample point set after each change is recorded, and the sensitivity of each key parameter to the matching rate is calculated. Key parameters with sensitivity greater than the preset sensitivity are selected as parameters to be tuned, and the initial parameter range of each parameter to be tuned is determined by combining the physical meaning of the parameter with the empirical range.

10. A multi-type interference identification system for mountainous areas based on a time-series segmentation algorithm, characterized in that, include: The preprocessing module is used to acquire multispectral remote sensing image data covering the target area, and to preprocess and filter the effective pixels of the multispectral remote sensing image data to form a preprocessed image set; wherein the temporal duration of the multispectral remote sensing image data is longer than a preset duration. The reconstruction module is used to calculate the normalized vegetation index pixel by pixel based on the preprocessed image set and obtain a continuous vegetation index time-series dataset through temporal interpolation and reconstruction. The construction module is used to construct a sequence of ground truth points for interference samples based on the remote sensing image data of the validation samples; the sequence of ground truth points for interference samples includes information on the time, spatial location, and type of interference. The identification module is used to input the vegetation index time series dataset into the Bayesian time series decomposition algorithm, decompose it into trend components, seasonal components and mutation components, and extract the mutation point sequence as the preliminary interference identification result; The adaptive optimization module is used to calculate the matching rate based on the ground truth sequence of the interference samples and the mutation point sequence, and to adaptively optimize and correct the parameters of the Bayesian time series decomposition algorithm according to the matching rate until the matching rate reaches a preset threshold, and output the corrected interference detection result.

Images