Intelligent classification system for alpine grassland vegetation and bare land based on multi-band remote sensing image collaboration

Abstract

This invention discloses an intelligent classification system for alpine grassland vegetation and bare land based on multi-band remote sensing image collaboration, relating to the fields of remote sensing information processing and ecological environment monitoring technology. It includes a multi-source data acquisition module, an ice crystal suspected area generation module, a reflection anomaly identification module, an interference dynamic modeling module, and a dynamic classification correction module. The multi-source data acquisition module acquires remote sensing image data over multiple time periods and simultaneously records the location of shady slopes and frost signs, generating an observation data list for alpine regions. This invention achieves temporal correspondence between red and near-infrared bands through multi-time period, multi-band, and meteorological parameter acquisition, dynamically identifying and eliminating ice crystal interference. Combined with shady slope and frost feature screening and threshold temporal redistribution, along with backlighting patrols, shading patrols, and polarization adjustments, it achieves adaptive correction of classification results, improving the accuracy and stability of alpine grassland surface identification.

Description

Technical Field

[0001] This invention relates to the fields of remote sensing information processing and ecological environment monitoring technology, specifically to an intelligent classification system for alpine grassland vegetation and bare land based on multi-band remote sensing image collaboration. Background Technology

[0002] The intelligent classification system for alpine grassland vegetation and bare land, based on multi-band remote sensing image synergy, is an ecological monitoring technology system grounded in spectral difference analysis, computer vision recognition, and multi-temporal remote sensing data fusion. Its core idea is to utilize the differences in red and near-infrared reflectance characteristics between vegetation and bare land, combined with spatial feature extraction and morphological recognition mechanisms in computer vision, to intelligently identify and classify land cover types using key indicators such as the Normalized Difference Vegetation Index (NDVI). The system first acquires multi-source remote sensing images containing red and near-infrared bands, prioritizing images from June to August when cloud cover is low and vegetation growth is vigorous. Humidity and climate corrections are applied using data from the same period over the past 2-3 years to ensure temporal consistency. Subsequently, atmospheric correction and geometric alignment are performed on the images, red and near-infrared reflectance are calculated, texture and spatial structure features are extracted using computer vision algorithms, and the NDVI value is comprehensively calculated with a set threshold range to achieve intelligent differentiation between vegetation and bare land. The system can classify land cover based on regional characteristics using methods such as manual interpretation, supervised classification, unsupervised classification, or threshold segmentation. It also combines the computer vision comparison results of ground samples to correct the classification accuracy, thereby forming an automated land cover identification system covering alpine grassland ecological zones. This provides reliable data support for grassland degradation monitoring, ecological restoration assessment, and land use planning.

[0003] The existing technology has the following shortcomings: In existing technologies, when remote sensing systems are applied to high-altitude meadow areas, they are easily affected by interference from floating ice crystals. Due to the low temperatures and high humidity in these areas, suspended ice crystals or thin frost layers often form on the surface. These ice crystals produce abnormally high reflectivity signals in the near-infrared band, easily misidentified by the system as high-reflectivity vegetation areas. Especially during multi-band collaborative classification, the reflectivity characteristics of floating ice crystals are highly similar to the spectral response of healthy vegetation. During automatic classification, the system may incorrectly generate small patches of dense vegetation, leading to distorted land cover information. Such misidentification not only results in an inflated grassland cover rate but also interferes with degradation monitoring and ecological assessment results. Current technologies lack mechanisms for identifying spectral anomalies and dynamically correcting interference from high-altitude ice crystals.

[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 an intelligent classification system for alpine grassland vegetation and bare land based on multi-band remote sensing imagery, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, this invention provides the following technical solution: a multi-band remote sensing image-based intelligent classification system for alpine grassland vegetation and bare land, comprising a multi-source data acquisition module, an ice crystal suspected area generation module, a reflection anomaly identification module, an interference dynamic modeling module, and a dynamic classification correction module: The multi-source data acquisition module acquires remote sensing image data for multiple time periods, including red light band, near-infrared band, temperature, humidity, wind speed and solar altitude angle, and simultaneously records the location of shady slopes and signs of frost, generating a list of observation data for high-altitude cold regions for subsequent ice crystal interference identification. The ice crystal suspected area generation module, based on the list of observation data of high-altitude cold regions, filters the time and area where the light conditions are during dawn and dusk and fog, frost and shady slopes accumulate. Combining meteorological parameters and spatial location, it generates a set of suspected ice crystal areas, providing spatial constraints for subsequent reflection anomaly detection. The reflectance anomaly identification module takes the set of suspected ice crystal areas as the analysis scope, compares the red light reflectance and near-infrared reflectance of remote sensing images of the same geographical location over multiple time periods, extracts the surface pixels that show a sudden increase in near-infrared reflectance and a stable red light reflectance within a continuous observation period, and generates a list of high reflectance anomaly areas. The interference dynamic modeling module utilizes the near-infrared reflection change characteristics of each region in the list of high reflectance anomaly areas, combined with ground temperature difference and wind direction trajectory information, to identify the spatial drift pattern of the reflected signal, determine the geographical location and duration of the ice crystal interference area, and generate a data list of interference periods. The dynamic classification correction module adjusts the observation angle and observation time sequence of the remote sensing system based on the data list of interference periods, performs backlighting patrol, inserts short-term shading cruise process and reverse polarization adjustment process, and reallocates the temporal weights of multi-band classification thresholds to dynamically correct the land cover classification results, thereby improving the accuracy of alpine grassland vegetation and bare land classification.

[0007] The preferred steps for generating the observation data list for high-altitude and cold regions are as follows: After determining the observation range in the alpine grassland and meadow areas, remote sensing image data with multiple time periods including red light band and near-infrared band were collected, and the collection time and solar altitude angle information were recorded. While collecting data in the red and near-infrared bands, temperature, humidity, wind speed and solar altitude angle parameters were recorded. The meteorological data and image time were then matched and numbered in chronological order. After completing the collection of red light band, near-infrared band and meteorological data, the terrain features were classified, and the location of shady slopes and signs of frost were identified by combining digital elevation model and solar altitude angle, and a set of shady slope and frost information was formed. The data on red light reflectance, near-infrared reflectance, temperature, humidity, wind speed, solar altitude angle, topographic aspect, slope, and frost characteristics are integrated to form a list of observational data for high-altitude cold regions with time labels and geographic coordinate indexes for ice crystal interference identification.

[0008] Preferably, when forming the observation data list for high-altitude cold regions, the information set of shady slopes and frost is matched with the reflectance data of red light band and near-infrared band. The correspondence between spectral information and meteorological conditions and terrain features is established through time tags and geographic coordinate indexes, so as to achieve spatiotemporal unification of multidimensional observation data and improve the accuracy of ice crystal interference identification.

[0009] Preferably, the steps for generating the suspected ice crystal region set are as follows: By jointly analyzing the solar altitude angle, geographic coordinates, and time labels of each time period in the observation data list of high-altitude and cold regions, the observation records with light conditions during the dawn and dusk phases were identified, and the extent of the surface area was obtained. By cross-matching temperature, humidity, and wind speed data with the time periods of dawn and dusk, the time and spatial locations that meet the conditions for fog and frost formation are identified, and a set of spatiotemporal data on fog and frost is formed. By combining the slope aspect, slope and solar altitude angle information in the high-altitude cold region observation data list, the distribution area of ​​shady slopes is identified, and the spatiotemporal data set of fog and frost is spatially superimposed with the distribution of shady slopes to form a set of shady slope cluster areas. By integrating the cluster of shady slope areas with data on morning and evening light periods and fog / frost conditions, and combining meteorological parameters with spatial location, a set of suspected ice crystal areas with time markers, spatial coordinates, meteorological parameters, and topographic feature labels is generated.

[0010] Preferably, when generating a set of suspected ice crystal areas, surface units in the shady slope aggregation area that simultaneously possess low solar altitude angle, high humidity, low temperature and low wind speed are spatially aggregated, and continuous screening is performed based on time tags, so that the formed suspected ice crystal areas remain continuous in space and consistent in time, thereby ensuring that the set of suspected ice crystal areas has high spatiotemporal consistency.

[0011] Preferably, the steps for generating the list of high reflectivity anomaly areas are as follows: Spatial location of geographic coordinates is performed within the suspected ice crystal area set, and remote sensing image data of multiple time periods corresponding to these coordinates are called on the time axis. The red light band and near-infrared band images are arranged in the order of acquisition time and time identifiers are set. By comparing the red light reflectance and near-infrared reflectance of the same geographical location over multiple time periods, surface pixels with a sudden increase in near-infrared reflectance and a stable red light reflectance are extracted within a continuous observation period. Spatial clustering and temporal correlation are performed on surface pixels with a sudden increase in near-infrared reflectivity and stable red reflectivity to generate a set of anomalous reflectivity units with spatiotemporal attributes. Information on anomalous reflectance unit sets is summarized, including geographic coordinate range, time series number, reflectance variation range and meteorological parameter information, to generate a list of high reflectance anomaly areas.

[0012] Preferably, when generating the list of high reflectivity anomaly areas, the geographical coordinate range of each anomaly area is matched with the time series number, and the temperature, humidity, wind speed and solar altitude angle information at the time of observation are matched together, so that the list of high reflectivity anomaly areas has both spatiotemporal distribution characteristics and environmental parameter correlation information, which can be used for the identification and tracking of ice crystal interference areas.

[0013] Preferably, the steps for generating the data list for the interference period are as follows: Near-infrared reflectance change characteristics were extracted and time-series organized for each region in the list of high reflectance anomalies. Near-infrared reflectance records from different time periods were arranged in chronological order and a reflectance change sequence was established. By matching ground temperature difference data with reflectance change sequences in time and space, we can analyze the correlation trend between temperature difference and near-infrared reflectance changes and identify temperature reversal and sudden drop in temperature. By introducing wind direction trajectory information, the wind direction trajectory is aligned with the spatial center position of the high reflectance anomaly area, and the continuous displacement characteristics of the near-infrared high reflectance area along the wind direction trajectory are identified. By summarizing the spatial displacement path, time duration interval, and meteorological parameters, a list of interference period data is generated, which includes the geographical location range, start time, end time, reflection change amplitude, temperature difference status, and wind direction.

[0014] Preferably, during the process of generating the data list of interference periods, ice crystal interference areas with the same temperature difference gradient and consistent wind direction trajectory are classified, and the reflection change amplitude, duration and spatial displacement direction of each type of interference area are grouped and organized to form a set of interference types with temporal continuity and spatial consistency, which is used for targeted correction processing in the subsequent dynamic classification correction stage.

[0015] Preferably, the following steps are taken to dynamically correct the land cover classification results by adjusting the remote sensing observation angle and time sequence based on the data list of the interference period, performing backlighting patrol, conducting short-term shading patrol and reverse polarization adjustment, and reallocating the temporal weights of the multi-band classification thresholds: Based on the data list of interference periods, the start time, end time, geographical location range, wind direction and trajectory, and temperature difference status of the interference are read, and the observation time sequence is rearranged and the observation window is determined. By combining wind direction trajectory and solar altitude angle information, the relationship between illumination direction is analyzed and a backlighting inspection is carried out. Under backlighting conditions, red light band and near-infrared band images are collected and reflection characteristics are recorded. For areas with reflection drift during the interference period, a short-term shading cruise process and a reverse polarization adjustment process are performed to reduce ice crystal reflection interference by adjusting the observation angle and polarization angle. Based on the interference intensity, duration, and illumination characteristics recorded in the interference period data list, the temporal weights of the multi-band classification thresholds are redistributed to dynamically correct the land cover classification results.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention establishes a temporal correspondence between the red and near-infrared bands based on the joint acquisition of data from multiple time periods, meteorological parameters, and spectral bands. This allows for the identification and elimination of ice crystal reflection anomalies in three dimensions: time, space, and spectrum. By dynamically screening the characteristics of shaded slopes and frost under dawn and dusk lighting conditions, it enables the early identification and real-time monitoring of areas affected by ice crystal interference. This ensures that the classification data maintains temporal continuity and spectral stability in complex, high-altitude environments, thereby improving the accuracy and reliability of vegetation and bare soil identification.

[0017] This invention enables remote sensing observations to adapt by dynamically modeling high-reflectivity anomaly regions and temporally redistributing threshold weights. Through dynamic correction methods such as backlighting patrols, short-term shading patrols, and reverse polarization adjustments, it effectively reduces the interference of ice crystal reflections on the near-infrared band, allowing land cover classification results to automatically optimize with environmental changes. This improves the spatiotemporal consistency of land cover identification and provides stable data support for alpine grassland ecological monitoring and degradation assessment. Attached Figure Description

[0018] 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.

[0019] Figure 1 This is a schematic diagram of the modules of the intelligent classification system for alpine grassland vegetation and bare land based on multi-band remote sensing image collaboration of the present invention. Detailed Implementation

[0020] 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.

[0021] This invention provides, for example Figure 1 The multi-band remote sensing image-based intelligent classification system for alpine grassland vegetation and bare land, as shown, includes a multi-source data acquisition module, a module for generating suspected ice crystal areas, a module for identifying reflection anomalies, a module for dynamic interference modeling, and a module for dynamic classification correction. The multi-source data acquisition module acquires remote sensing image data for multiple time periods, including red light band, near-infrared band, temperature, humidity, wind speed and solar altitude angle, and simultaneously records the location of shady slopes and signs of frost, generating a list of observation data for high-altitude cold regions for subsequent ice crystal interference identification. To create a data inventory of high-altitude cold regions for ice crystal interference identification, red light band, near-infrared band, and meteorological parameters were simultaneously collected and processed. Through refined observation of remote sensing images over multiple time periods, a high-altitude cold region observation data system with temporal continuity and environmental relevance was established. The specific implementation steps are as follows: After determining the observation range in the target alpine grassland and meadow area, remote sensing image data sources covering the red light and near-infrared bands were selected, and images were acquired over multiple time periods. To ensure the temporal continuity and climatic representativeness of the observation results, red light band reflectance information and near-infrared band reflectance information were acquired separately for different time periods such as morning, noon, and dusk, and the acquisition time was recorded in correspondence with the solar altitude angle. During the acquisition process, the spatial resolution, illumination angle, and visible surface range of the image data were uniformly configured to ensure that the red light and near-infrared band data remained consistent in space and time. Red light band data is used to reflect the absorption characteristics of visible light by surface vegetation, while near-infrared band data is used to characterize the reflection characteristics of near-infrared radiation by surface structures. By establishing multi-time-period red light and near-infrared data sets within the same geographical area, basic reflectance feature data for ice crystal interference identification is obtained.

[0022] After acquiring red and near-infrared band data, meteorological conditions for the same time period were simultaneously observed and recorded, including temperature, humidity, wind speed, and solar altitude angle. Temperature parameters characterize the thermal energy state of the surface in high-altitude and cold regions; humidity parameters reflect the water vapor content in the air; wind speed parameters determine the formation and migration conditions of ice crystals or frost on the surface; and solar altitude angle parameters calculate the impact of the incident angle of sunlight on reflectance characteristics. Simultaneously with each remote sensing image acquisition, meteorological data was linked to the image time, forming data sets with time stamps and climatic characteristics. Each data set was numbered and organized according to the acquisition time sequence to ensure that changes in reflectance characteristics could be tracked based on the time series during subsequent ice crystal interference identification. Through the coupling of meteorological and band data, the spectral response of high-altitude and cold regions under different climatic conditions was quantified, providing data support for analyzing the formation mechanism of ice crystal interference.

[0023] After simultaneously recording red light, near-infrared, and meteorological data, the topographic features of the alpine grassland area were classified and labeled, with a focus on identifying shady slopes and frost-prone areas. Shady slopes typically have lower light intensity and more stable temperature and humidity conditions, making them high-risk locations for ice crystal and thin frost formation. Therefore, during topographic classification, digital elevation models and solar altitude angle data were combined to identify surface aspect and slope. By calculating the relative illuminance and slope direction of each surface unit, shady slopes were separated from the overall topography, and a correspondence table between topographic features and light conditions was established. Subsequently, frost signs were identified in measured or auxiliary imagery, including surface reflection anomalies, low-temperature retention areas, and water vapor condensation zones. This information was combined with topographic labels to form a set of shady slope and frost information with spatial location, time labels, and climate parameters. In this process, shady slope and frost signs correspond one-to-one with the red light and near-infrared data collected in the previous step, ensuring that each observation point simultaneously possesses spectral, meteorological, and topographic features.

[0024] After fusing multidimensional data from the red band, near-infrared band, meteorological parameters, and topographic features, observation information from all time periods is integrated according to the acquisition sequence to generate a high-altitude cold region observation data list. This list includes red reflectance, near-infrared reflectance, temperature, humidity, wind speed, solar altitude angle, topographic aspect, slope, and frost characteristics for each observation period. To maintain data consistency, each record corresponds to a unique timestamp and geographic coordinate index, and is matched with its corresponding meteorological conditions and topographic information. Once the high-altitude cold region observation data list is formed, it will serve as the input basis in the subsequent ice crystal interference identification step, used to analyze reflectance changes in multi-temporal remote sensing images.

[0025] The ice crystal suspected area generation module, based on the list of observation data of high-altitude cold regions, filters the time and area where the light conditions are during dawn and dusk and fog, frost and shady slopes accumulate. Combining meteorological parameters and spatial location, it generates a set of suspected ice crystal areas, providing spatial constraints for subsequent reflection anomaly detection. Based on the observation data list of high-altitude and cold regions, the time and areas where ice crystals are likely to occur are determined. A comprehensive screening and spatial correlation analysis is conducted on illumination conditions, meteorological characteristics, and topographic distribution to generate a set of suspected ice crystal areas with high spatiotemporal consistency, thus providing spatial constraints for subsequent reflection anomaly detection. The specific implementation steps are as follows: A joint analysis of solar altitude angle, geographic coordinates, and time labels for each time period in the high-altitude cold region observation data list was conducted to identify observation records where illumination conditions were during the dawn and dusk phases. To ensure the correlation between illumination conditions and surface reflection response, time periods with solar altitude angles within the low angle of incidence were extracted, and the observation intervals before and after sunrise and sunset were identified by combining time labels. In this process, the influence of illumination conditions on surface reflection characteristics was analyzed by comparing the reflection ratios of the red and near-infrared bands at different angles of incidence, ensuring that the selected dawn and dusk periods fully represent the radiation conditions under low solar altitude angles. Subsequently, these time periods were matched with geographic coordinate information to obtain the surface area range under dawn and dusk illumination conditions, providing a spatiotemporal basis for the subsequent extraction of humidity and frost characteristics. Through this process, the temporal and spatial dimensions of the high-altitude cold region observation data list can be correlated, providing a clear illumination background for subsequent screening work.

[0026] After determining the twilight and daylight periods, temperature, humidity, and wind speed data within the same time interval are comprehensively screened to identify the time periods and spatial locations where conditions for fog and frost formation are present. Through time-series analysis of temperature data, periods with temperatures near or below freezing are extracted; through the spatial distribution of humidity data, areas with consistently high humidity are identified; and wind speed information is combined to determine local airflow patterns, identifying stable stagnant areas or low-wind-speed clusters where fog and frost may form. By cross-matching the time-series information of temperature, humidity, and wind speed with the previously obtained twilight and daylight periods, the times and areas where meteorological conditions under twilight and daylight conditions favor fog and frost formation can be located. At this stage, meteorological characteristics are mapped one-to-one with geographic coordinates, associating each eligible time period with a specific spatial location, forming a preliminary spatiotemporal data set of fog and frost with meteorological characteristics, providing input for the subsequent clustering analysis of shady slope areas.

[0027] After obtaining a spatiotemporal dataset containing conditions conducive to frost formation, the spatial distribution of shady slope areas is identified by combining topographical information. Based on the slope aspect, gradient, and solar altitude angle information recorded in the high-altitude cold region observation data list, the relative illuminance distribution of different surface units is calculated, and surface units with low illuminance values ​​are identified as shady slope areas. Shady slope areas are more prone to frost and ice crystal accumulation due to shorter daylight hours, gentler temperature changes, and higher humidity accumulation. Therefore, during shady slope identification, the frost and fog conditions identified in the previous step are spatially overlaid with the shady slope distribution to filter out times and regions that simultaneously meet the requirements of twilight, frost and fog conditions, and shady slope characteristics. This process allows for spatial focusing on characteristic areas prone to ice crystal reflection interference and temporal consistency with meteorological conditions. Through this spatiotemporal overlay filtering, a set of shady slope accumulation areas with multi-dimensional characteristic constraints is formed, providing a clear topographical and climatic background for the subsequent formation of suspected ice crystal areas.

[0028] After the aggregator of shady slope clusters is formed, its spatial distribution characteristics are integrated, and a set of suspected ice crystal areas is generated by combining meteorological parameters and spatial location. This step unifies the shady slope clusters with data on dawn / dusk illumination time periods and fog / frost conditions, aggregating surface units with low solar altitude angles, high humidity, low temperatures, and low wind speeds within the same time window into continuous spatial regions. In this way, the set of suspected ice crystal areas not only includes meteorological conditions reflecting the ice crystal formation environment but also shady slope characteristics representing topographic influences, and is correlated with sunlight incidence conditions. Each suspected ice crystal area is generated with a corresponding time stamp, spatial coordinates, meteorological parameters, and topographic feature labels, which can serve as the spatial constraint basis for the subsequent reflection anomaly identification stage. After the generation of the suspected ice crystal area set, the time and location of possible ice crystal interference in high-altitude and cold regions are effectively limited, providing a high-precision spatiotemporal reference framework for subsequent reflection anomaly detection and dynamic correction.

[0029] The reflectance anomaly identification module takes the set of suspected ice crystal areas as the analysis scope, compares the red light reflectance and near-infrared reflectance of remote sensing images of the same geographical location over multiple time periods, extracts the surface pixels that show a sudden increase in near-infrared reflectance and a stable red light reflectance within a continuous observation period, and generates a list of high reflectance anomaly areas. Within the spatial range of suspected ice crystal areas, surface regions affected by ice crystal reflection interference are identified. Through comparative analysis of red and near-infrared reflectance from multi-time-period remote sensing images, continuous observation feature comparisons are performed on surface pixels at the same geographical location to generate a list of high-reflectance anomaly areas containing these anomalous reflectance regions. The specific implementation steps are as follows: After identifying the suspected ice crystal areas, the geographic coordinates within this area are spatially located, and multi-time-segment remote sensing image data corresponding to these coordinates are retrieved on the timeline. To ensure the continuity of reflectance variation analysis, the red and near-infrared image data are arranged chronologically according to their acquisition time, and a time identifier is assigned to each observation point. This ensures that each surface pixel has reflectance records for multiple time periods within the time series. During this process, the boundary coordinates of the suspected ice crystal areas are used as a limiting condition for the analysis range, ensuring that reflectance comparisons only occur within the spatial range where ice crystal interference may exist. Each surface pixel is uniquely identified within this region to maintain spatial correspondence in subsequent comparative analysis. Red reflectance characterizes the surface's absorption of visible light, while near-infrared reflectance characterizes the radiative response of surface structure and moisture content; both together constitute the main part of the spectral response curve, providing fundamental spectral information for subsequent anomaly identification.

[0030] After matching the time series with spatial coordinates, a comparative analysis of red and near-infrared reflectance over multiple time periods at the same geographical location was conducted to extract the reflectance variation trend within continuous observation periods. During the analysis, red reflectance was used as a stable reference band. By comparing the magnitude of red reflectance changes in adjacent observation periods, the stability of the surface vegetation status was determined. Simultaneously, near-infrared reflectance was used as a sensitive band to detect its magnitude and direction of change within continuous observation periods. When a sudden increase in near-infrared reflectance was observed in adjacent observation periods, while red reflectance remained stable, this pattern was marked as a potential ice crystal reflection interference feature. By establishing a comparative relationship across continuous observation periods in the time dimension, instantaneous reflection enhancement caused by ice crystal formation or drift can be captured, preventing long-term vegetation changes from being misidentified as anomalous reflection. In this case, the temporal comparison of red and near-infrared reflectance is not only used to identify anomalies but also to eliminate non-ice crystal reflection changes caused by variations in illumination angle or humidity fluctuations, thereby improving the accuracy of anomaly identification.

[0031] After identifying surface pixels exhibiting a sudden increase in near-infrared reflectance and stable red reflectance, these pixels are spatially clustered and temporally correlated. Pixels showing similar reflectance characteristics within consecutive observation periods are spatially merged to form preliminary high-reflectance distribution zones. To ensure spatial continuity, surface units exhibiting a sudden increase in near-infrared reflectance in two or more consecutive observation periods are merged to form temporally continuous and spatially adjacent regional segments. For discontinuously distributed anomalous pixels, they are regrouped according to their spatial adjacency, allowing reflectance characteristics within the same region to form closed or semi-closed structures in space, thus reflecting the actual range of ice crystal reflectance distribution. During this process, the center coordinates, area, near-infrared reflectance variation amplitude, and red reflectance stability parameters of each anomalous region are recorded, generating a set of anomalous reflectance units with spatiotemporal attributes. This spatial organization process not only reflects the spatial clustering pattern of ice crystal reflectance but also provides a continuous temporal span and spatial boundary for subsequent dynamic modeling stages.

[0032] After spatial clustering, high reflectance anomaly areas are uniformly integrated and their information summarized to generate a directory of high reflectance anomaly areas. This directory includes the geographic coordinates of each anomaly area, the observation period number in which the anomaly first appeared in the time series, the variation range of near-infrared reflectance, the stable range of red reflectance, and its positional relationship within the set of suspected ice crystal areas. To maintain spatiotemporal data consistency, a correspondence is established between each anomaly area in the directory and its corresponding suspected ice crystal area, along with relevant meteorological parameters, including temperature, humidity, wind speed, and solar altitude angle at the time of observation. In this way, the directory of high reflectance anomaly areas not only records the spatiotemporal distribution of anomalous reflectance but also reflects its correlation with environmental conditions in its data structure. The formation of the directory makes ice crystal interference areas traceable in both time and space, providing a complete input basis for subsequent dynamic modeling of interference and correction of classification results. This directory can reflect the frequency of occurrence, duration, and spatial migration trend of ice crystal interference, providing a direct reference for identifying the drift patterns of ice crystal reflectance.

[0033] The interference dynamic modeling module utilizes the near-infrared reflection change characteristics of each region in the list of high reflectance anomaly areas, combined with ground temperature difference and wind direction trajectory information, to identify the spatial drift pattern of the reflected signal, determine the geographical location and duration of the ice crystal interference area, and generate a data list of interference periods. By utilizing the near-infrared reflectance variation characteristics of each region recorded in the high reflectance anomaly area directory, combined with ground temperature difference and wind direction trajectory information, the drift patterns of reflected signals in the spatial and temporal dimensions were further identified. This allowed for the determination of the specific geographical location and duration of ice crystal interference areas, resulting in a data list of interference periods. An orderly spatiotemporal correlation analysis and hierarchical organization of the dynamic behavior of high reflectance anomaly areas were then conducted. The specific implementation steps are as follows: Near-infrared reflectance variation characteristics of each region in the list of high reflectance anomalies were extracted and temporally processed. Near-infrared reflectance records for each anomaly region at different time periods were arranged into a time series, and the magnitude of reflectance change and its corresponding geographic coordinates at each time point were identified. To ensure the spatiotemporal continuity of reflectance characteristics, regions with consistent near-infrared reflectance variation trends in adjacent time periods were considered as the same reflectance evolution unit, thus constructing a near-infrared reflectance variation sequence for each high reflectance anomaly region. This reflectance variation sequence can intuitively reflect the formation, enhancement, and dissipation of ice crystal interference over time. Simultaneously, the red reflectance state at each time point was synchronously correlated to ensure that the independence of near-infrared changes stems from ice crystal interference rather than changes in surface vegetation growth. Through this process, a multi-dimensional data framework incorporating time, space, and reflectance intensity was formed, providing a foundation for subsequent coupled analysis of surface meteorological factors.

[0034] After obtaining the near-infrared reflectance variation characteristics of the high reflectance anomaly region, the ground temperature difference data and reflectance variation sequences were spatiotemporally correlated to analyze the correlation trend between temperature changes and reflectance changes. The temperature difference data were derived from ground temperature records for each time period in the high-altitude cold region observation data list. By calculating the differences between temperature values ​​of adjacent time periods, the surface temperature difference distribution for different time periods was obtained. Since the formation of ice crystal disturbance is closely related to the difference between the cold and warm layers of the ground, the temperature difference data can reflect the influence of surface thermal changes on near-infrared reflectance. In this stage, the temperature difference data was mapped to the geographic coordinate range corresponding to the high reflectance anomaly region and matched with the near-infrared reflectance variation sequence on the time axis. Through this matching method, the characteristics of temperature changes before and after a sudden increase in near-infrared reflectance can be identified, such as a sudden drop in temperature or a temperature reversal, thereby inferring the possible time periods of ice crystal formation or drift. Simultaneously, the spatial gradient of the temperature difference was compared with the spatial distribution of reflectance changes to analyze the movement trend of ice crystal disturbance on the surface, providing a thermodynamic basis for the subsequent introduction of wind direction trajectories.

[0035] After establishing the correspondence between temperature difference and reflectance variation characteristics, wind direction trajectory information is introduced to overlay analysis of the spatial distribution of high reflectance anomaly areas with wind direction changes, in order to identify the spatial drift pattern of reflectance signals. Wind direction trajectory information comes from wind speed and direction records in the high-altitude cold region observation data list. By sorting wind direction changes over consecutive time periods, a wind direction movement trajectory line can be formed. During this process, the wind direction trajectory is aligned with the spatial center position of the high reflectance anomaly area to identify the spatial shift direction of the near-infrared high reflectance area at different time periods. If the high reflectance anomaly area shows continuous displacement along the wind direction trajectory in the time series, it can be determined that the reflectance change in this area is affected by ice crystal drift. Simultaneously, the cross-relationship between the wind direction trajectory and the temperature difference gradient is analyzed to determine the correspondence between the direction of cold and moist air flow and the migration path of ice crystal distribution. By overlaying the reflectance variation trend, temperature difference variation characteristics, and wind direction trajectory in three-dimensional space, the drift pattern of ice crystal interference in time and space can be identified, clarifying its dynamic process of formation, movement, and dissipation.

[0036] After identifying the spatial drift patterns of the reflected signals, the spatial displacement paths, time durations, and related meteorological parameters of each high-reflectance anomaly area were compiled to generate a data list of interference periods. This list includes information such as the geographical location, start time, end time, reflection variation amplitude, temperature difference, wind direction, and wind speed distribution of each ice crystal interference area. To ensure data integrity, each interference period is associated with a corresponding set of high-reflectance anomalies and suspected ice crystal areas, making the interference period data traceable both spatially and temporally. During the data aggregation process, interference areas with similar temperature gradients and wind trajectories were categorized so that targeted corrections could be made based on the interference type in the subsequent dynamic correction phase. The generated data list of interference periods not only records the spatiotemporal range of ice crystal interference but also reflects the environmental conditions and drift direction of the interference, providing a crucial time window and spatial constraint basis for subsequent dynamic correction.

[0037] The dynamic classification correction module adjusts the observation angle and observation time sequence of the remote sensing system based on the data list of interference periods, performs backlighting patrol, inserts short-term shading cruise process and reverse polarization adjustment process, reallocates the temporal weight of multi-band classification thresholds, and dynamically corrects the land cover classification results, thereby improving the accuracy of alpine grassland vegetation and bare land classification. Based on the data list of interference periods, the remote sensing observation process was optimized in terms of timing and angle. Dynamic redistribution of multi-band classification thresholds was achieved through methods such as backlighting patrols, short-term shading patrols, and reverse polarization adjustments. This continuously corrected the land cover classification results, ensuring the stability of the identification accuracy of alpine grassland vegetation and bare land in both time and space. Systematic dynamic adjustments to the observation strategy, illumination response, and band weight allocation are required. The specific implementation steps are as follows: After the interference period data list is generated, the start and end times, geographical locations, wind direction and trajectory, and temperature differences recorded within are read one by one to determine the observation window corresponding to each interference period. During this process, the interference periods are compared with the planned remote sensing image acquisition sequence to identify time intervals and spatial areas at risk of ice crystal interference. To avoid acquiring images affected by ice crystal reflection during the interference period, the observation time sequence is first rearranged, postponing observation tasks within the interference period to after the interference dissipates or advancing them to the time window before the interference forms. This adjustment allows for dynamic adaptability in the multi-band image acquisition sequence within the same observation cycle, ensuring the stability of spectral reflectance characteristics. Simultaneously, by combining wind direction and trajectory information with solar altitude angle information, the solar incidence direction and reflection path characteristics of each observation area at different time periods are determined, providing a basis for subsequent adjustments to the observation angle. In this way, the adjustment of observation time and angle no longer relies on a fixed plan but is dynamically controlled based on the time markers in the interference period data list, enabling the observation process to proactively avoid high-risk periods of ice crystal reflection interference.

[0038] After optimizing the observation time sequence, the illumination direction relationship of each interference area was analyzed, and a backlighting inspection process was executed. Backlighting inspection refers to supplementing observations of the same area under conditions where the illumination direction is opposite to the observation direction, in order to capture the reflection differences of ice crystal interference under different incident angles. During implementation, the observation path in the backlighting direction was determined by calculating the solar altitude angle and azimuth angle information in relation to the surface coordinates. For areas within the interference period, red and near-infrared band images were acquired under backlighting conditions to reduce the reflection enhancement effect caused by normal incidence. In this stage, the observation results from the backlighting inspection were recorded together with the observation results from the normal direction and paired on the time axis, allowing the reflection characteristics of the same surface pixel under the two illumination conditions to form a complementary data structure. In this way, the low reflection characteristics under backlighting conditions can be used to correct the high reflection anomalies under normal illumination, making the overall reflection curve more stable in the time dimension, thereby reducing the impact of ice crystal interference on classification accuracy.

[0039] After the backlighting survey is completed, a short-term shading cruise and a reverse polarization adjustment process are performed in areas with strong reflection drift during the interference period to further reduce the impact of ice crystals on the reflection of red and near-infrared wavelengths. The short-term shading cruise involves adjusting the observation angle within a short time interval to partially shade the observation path, thereby reducing the energy reflected from the ice crystal layer into the sensor's field of view. This process is executed based on the reflection drift direction and duration recorded in the interference period data list, selecting appropriate shading duration and angle variation ranges to ensure the short-term cruise is completed under stable lighting conditions. The subsequent reverse polarization adjustment process suppresses the high reflection response in the near-infrared band caused by ice crystal surface reflection by changing the relationship between the observation direction and the polarization angle of the incident light. The polarization adjustment process uses the wind trajectory during the interference period as a reference direction, observing away from the wind distribution direction to weaken the specular reflection component of the ice crystal surface. The combination of short-term shading and reverse polarization physically reduces the reflection characteristics of ice crystal interference in images acquired during the interference period, thus providing more stable spectral input conditions for subsequent redistribution of multi-band classification thresholds.

[0040] After dynamically adjusting the observation angle and illumination direction, the multi-band classification thresholds are reassigned temporally to dynamically correct the land cover classification results. Based on the interference intensity, duration, and illumination characteristics of each time period in the interference period data list, the weights of the red and near-infrared bands in the classification calculation are assigned temporally, biasing the weights towards stable bands during the interference period and gradually restoring the original weight ratios after the interference dissipates. This method allows the classification model to rely more on red band information during ice crystal interference, thus avoiding the impact of abnormal near-infrared reflection on the classification results. Simultaneously, supplementary data obtained from backlighting surveys and reverse polarization observations are incorporated into the weight adjustment criteria, giving them a time compensation attribute in the threshold allocation, enabling a smooth transition of the classification thresholds over time. Finally, in the classification output stage, the land cover results for the affected areas are dynamically corrected, ensuring that areas misclassified as having high reflectivity vegetation due to ice crystal reflection are corrected to match the actual land surface condition. This process not only achieves temporal consistency of classification results, but also enables adaptive adjustment of classification weights and lighting environment, thereby effectively improving the recognition accuracy and spatial stability of alpine grassland vegetation and bare land.

[0041] This invention establishes a temporal correspondence between the red and near-infrared bands based on the joint acquisition of data from multiple time periods, meteorological parameters, and spectral bands. This allows for the identification and elimination of ice crystal reflection anomalies in three dimensions: time, space, and spectrum. By dynamically screening the characteristics of shaded slopes and frost under dawn and dusk lighting conditions, it enables the early identification and real-time monitoring of areas affected by ice crystal interference. This ensures that the classification data maintains temporal continuity and spectral stability in complex, high-altitude environments, thereby improving the accuracy and reliability of vegetation and bare soil identification.

[0042] This invention enables remote sensing observations to adapt by dynamically modeling high-reflectivity anomaly regions and temporally redistributing threshold weights. Through dynamic correction methods such as backlighting patrols, short-term shading patrols, and reverse polarization adjustments, it effectively reduces the interference of ice crystal reflections on the near-infrared band, allowing land cover classification results to automatically optimize with environmental changes. This improves the spatiotemporal consistency of land cover identification and provides stable data support for alpine grassland ecological monitoring and degradation assessment.

[0043] 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 multi-band remote sensing image collaborative alpine grassland vegetation and bare land intelligent classification system, characterized in that, It includes a multi-source data acquisition module, an ice crystal suspected area generation module, a reflection anomaly identification module, an interference dynamic modeling module, and a dynamic classification correction module: The multi-source data acquisition module acquires remote sensing image data over multiple time periods and simultaneously records the location of shady slopes and signs of frost, generating a list of observation data for high-altitude and cold regions. The ice crystal suspected area generation module, based on the list of observation data of high-altitude cold regions, filters the time and area where the light conditions are during dawn and dusk and fog and frost and shady slopes accumulate, and generates a set of suspected ice crystal areas by combining meteorological parameters and spatial location. The reflectance anomaly identification module takes the set of suspected ice crystal areas as the analysis scope, compares the red light reflectance and near-infrared reflectance of remote sensing images of the same geographical location over multiple time periods, extracts the surface pixels that show a sudden increase in near-infrared reflectance and a stable red light reflectance within a continuous observation period, and generates a list of high reflectance anomaly areas. The interference dynamic modeling module utilizes the near-infrared reflection change characteristics of each region in the list of high reflectance anomaly areas, combined with ground temperature difference and wind direction trajectory information, to identify the spatial drift pattern of the reflected signal, determine the geographical location and duration of the ice crystal interference area, and generate a data list of interference periods. The dynamic classification correction module adjusts the observation angle and observation time sequence of the remote sensing system based on the data list of interference periods, performs backlighting patrol, inserts short-term shading cruise process and reverse polarization adjustment process, and reallocates the temporal weights of multi-band classification thresholds to dynamically correct the land cover classification results. 2.The alpine meadow vegetation and bare land intelligent classification system of multi-band remote sensing image synergy according to claim 1, wherein, The steps for generating a list of observation data for high-altitude and cold regions are as follows: After determining the observation range in the alpine grassland and meadow areas, remote sensing image data with multiple time periods including red light band and near-infrared band were collected, and the collection time and solar altitude angle information were recorded. While collecting data in the red and near-infrared bands, temperature, humidity, wind speed and solar altitude angle parameters were recorded. The meteorological data and image time were then matched and numbered in chronological order. After completing the collection of red light band, near-infrared band and meteorological data, the terrain features were classified, and the location of shady slopes and signs of frost were identified by combining digital elevation model and solar altitude angle, and a set of shady slope and frost information was formed. The data on red light reflectance, near-infrared reflectance, temperature, humidity, wind speed, solar altitude angle, topographic aspect, slope, and frost characteristics are integrated to form a list of observational data for high-altitude cold regions with time labels and geographic coordinate indexes for ice crystal interference identification. 3.The alpine meadow vegetation and bare land intelligent classification system of multi-band remote sensing image synergy according to claim 2, characterized in that, When creating the observation data list for high-altitude and cold regions, the information set on shady slopes and frost is matched with the reflectance data of the red and near-infrared bands. The correspondence between spectral information and meteorological conditions and terrain features is established through time tags and geographic coordinate indexes.

4. The multi-band remote sensing image collaborative alpine meadow vegetation and bare land intelligent classification system according to claim 2, characterized in that, The steps for generating a collection of suspected ice crystal regions are as follows: By jointly analyzing the solar altitude angle, geographic coordinates, and time labels of each time period in the observation data list of high-altitude and cold regions, the observation records with light conditions during the dawn and dusk phases were identified, and the extent of the surface area was obtained. By cross-matching temperature, humidity, and wind speed data with the time periods of dawn and dusk, the time and spatial locations that meet the conditions for fog and frost formation are identified, and a set of spatiotemporal data on fog and frost is formed. By combining the slope aspect, slope and solar altitude angle information in the high-altitude cold region observation data list, the distribution area of ​​shady slopes is identified, and the spatiotemporal data set of fog and frost is spatially superimposed with the distribution of shady slopes to form a set of shady slope cluster areas. By integrating the clusters of ice crystals on shady slopes with data on morning and evening light periods and fog / frost conditions, and combining meteorological parameters with spatial location, a cluster of suspected ice crystal areas is generated.

5. The multi-band remote sensing image collaborative alpine meadow vegetation and bare land intelligent classification system according to claim 4, characterized in that, When generating a cluster of suspected ice crystal areas, surface units in the shady slope cluster area that simultaneously possess low solar altitude angle, high humidity, low temperature and low wind speed are spatially aggregated, and continuous screening is performed based on time tags to ensure that the formed suspected ice crystal areas remain continuous in space and consistent in time.

6. The multi-band remote sensing image collaborative alpine meadow vegetation and bare land intelligent classification system according to claim 4, characterized in that, The steps for generating the list of high reflectivity anomaly areas are as follows: Spatial location of geographic coordinates is performed within the suspected ice crystal area set, and remote sensing image data of multiple time periods corresponding to these coordinates are called on the time axis. The red light band and near-infrared band images are arranged in the order of acquisition time and time identifiers are set. By comparing the red light reflectance and near-infrared reflectance of the same geographical location over multiple time periods, surface pixels with a sudden increase in near-infrared reflectance and a stable red light reflectance are extracted within a continuous observation period. Spatial clustering and temporal correlation are performed on surface pixels with a sudden increase in near-infrared reflectivity and stable red reflectivity to generate a set of anomalous reflectivity units with spatiotemporal attributes. Information on anomalous reflectance unit sets is summarized, including geographic coordinate range, time series number, reflectance variation range and meteorological parameter information, to generate a list of high reflectance anomaly areas.

7. The system according to claim 6, wherein the system is characterized by: When generating the list of high reflectivity anomaly areas, the geographical coordinate range of each anomaly area is matched with the time series number, and the temperature, humidity, wind speed and solar altitude angle information at the time of observation are matched together, so that the list of high reflectivity anomaly areas has both spatiotemporal distribution characteristics and environmental parameter correlation information.

8. The multi-band remote sensing image collaborative alpine meadow vegetation and bare land intelligent classification system according to claim 6, characterized in that, The steps for generating the data list for the interference period are as follows: Near-infrared reflectance change characteristics were extracted and time-series organized for each region in the list of high reflectance anomalies. Near-infrared reflectance records from different time periods were arranged in chronological order and a reflectance change sequence was established. By matching ground temperature difference data with reflectance change sequences in time and space, we can analyze the correlation trend between temperature difference and near-infrared reflectance changes and identify temperature reversal and sudden drop in temperature. By introducing wind direction trajectory information, the wind direction trajectory is aligned with the spatial center position of the high reflectance anomaly area, and the continuous displacement characteristics of the near-infrared high reflectance area along the wind direction trajectory are identified. The spatial displacement path, time duration interval, and meteorological parameters are summarized to generate a list of interference period data.

9. The multi-band remote sensing image collaborative alpine meadow vegetation and bare land intelligent classification system according to claim 8, characterized in that, In the process of generating the data list of interference periods, ice crystal interference areas with the same temperature difference gradient and consistent wind direction trajectory are classified, and the reflection change amplitude, duration and spatial displacement direction of each type of interference area are grouped and organized to form a set of interference types with temporal continuity and spatial consistency.

10. The multi-band remote sensing image collaborative alpine meadow vegetation and bare land intelligent classification system according to claim 8, characterized in that, Based on the data list of interference periods, the remote sensing observation angle and time sequence are adjusted, a backlighting patrol is performed, a short-term shading cruise and reverse polarization adjustment are conducted, and the temporal weights of the multi-band classification thresholds are reallocated. The following are the steps for dynamically correcting the land cover classification results: Based on the data list of interference periods, the start time, end time, geographical location range, wind direction and trajectory, and temperature difference status of the interference are read, and the observation time sequence is rearranged and the observation window is determined. By combining wind direction trajectory and solar altitude angle information, the relationship between illumination direction is analyzed and a backlighting inspection is carried out. Under backlighting conditions, red light band and near-infrared band images are collected and reflection characteristics are recorded. For areas with reflection drift during the interference period, a short-term shading cruise process and a reverse polarization adjustment process are performed to reduce ice crystal reflection interference by adjusting the observation angle and polarization angle. Based on the interference intensity, duration, and illumination characteristics recorded in the interference period data list, the temporal weights of the multi-band classification thresholds are redistributed to dynamically correct the land cover classification results.

Images