Tea garden identification method and system based on multi-source remote sensing data

Abstract

The invention discloses a tea garden identification method and system based on multi-source remote sensing data. The method comprises the following steps: S1, obtaining time sequence radar data and optical remote sensing data of a target area; s2, extracting key growth periods of the tea garden based on radar data, wherein the key growth periods comprise a growth season starting period and a growth season ending period; s3, calculating a vegetation index VI and a water body index WI by using the optical data, and screening a combination of the VI and the WI with the optimal distinction degree through a time-weighted dynamic time warping method; s4, determining a tea garden identification index according to the following method: omega1 is linear operation, omega2 is nonlinear operation, T represents a tea garden sample, G represents an evergreen vegetation sample, and DTW is a dynamic time warping method; and S5, realizing tea garden identification based on the tea garden identification index and the integrated learning model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural remote sensing image processing, and particularly relates to a tea garden recognition method and system based on multi-source remote sensing data. BACKGROUND

[0002] As an important economic crop and one of the three major beverages in the world, tea plays a significant role in promoting regional economic development, especially in developing countries. Accurate and efficient identification and monitoring of the spatial distribution of tea gardens is crucial for scientific guidance of land use planning, rational planting of tea gardens, disease prevention and control, yield estimation, and ecological environment protection.

[0003] Traditional tea garden area data acquisition relies on field investigation and agricultural census, which has significant defects such as time-consuming, subjective, lagging in updating, and lack of detailed spatial information. In contrast, satellite remote sensing technology has become the main means of land cover monitoring due to its wide monitoring range, strong temporal and spatial continuity, and relatively low cost, and has been widely applied in crop classification and identification. Existing research attempts to use remote sensing technology to identify tea gardens, but faces two major technical challenges:

[0004] Insufficient temporal integrity of optical data: In particular, in major tea-growing regions such as southern China, the frequent cloud cover caused by bad weather makes it difficult to obtain complete and high-quality time series data. Although time series interpolation methods can partially fill in the missing data, they are difficult to capture the unique, short-period (such as 5-8 days of picking period) phenological characteristics of tea gardens. Research using radar (SAR) data and optical data fusion to overcome cloud cover has made some progress, but its model is usually based on single-time multi-modal translation assumption, ignoring the seasonal and temporal evolution of ground object spectra, resulting in a lack of consistency and stability in the time series of the reconstruction results. In addition, existing methods often try to simulate up to 13 optical bands with limited radar channels (such as VV / VH), which is severely insufficient in information, making it difficult to accurately restore the spectral details and relative spectral shape of each band.

[0005] Spectral confusion between tea gardens and evergreen vegetation: Tea gardens and other evergreen vegetation (such as forests, shrubs) have highly similar spectral reflectance characteristics in the visible to near-infrared bands, due to similar chlorophyll content, water status, and canopy structure. This similarity is widespread in different climates, terrains, and management modes, and shows stability in multi-temporal observations, resulting in the lack of key discriminant features (such as red edge, near-infrared region). This seriously weakens the ability of classification algorithms to extract effective features and their generalization in complex environments, becoming a key bottleneck restricting the accurate identification of large-scale tea gardens. Although the introduction of hyperspectral data can enhance the capture of subtle physiological and biochemical differences, it has high acquisition costs, limited coverage, high data redundancy, and complex processing, making it difficult to meet the practical needs of large-scale, long-term tea garden monitoring.

[0006] Exponential-based methods have been concerned in crop identification due to their ability to enhance the spectral differences between target crops and other ground objects, and have shown good effectiveness and cost-efficiency in mapping other crops (e.g., soybean, rape, potato, winter wheat). However, so far, there has been no effective identification index designed for the unique and complex spectral-temporal characteristics of tea gardens, which can reliably distinguish tea gardens from other evergreen vegetation with similar spectra at a large scale.

[0007] In general, the main challenges faced by existing remote sensing technologies in the field of tea garden identification are: how to overcome the interference of clouds to obtain complete and high-quality time series data that can fully reflect the key phenological periods of tea gardens, especially the short-period characteristics; and how to extract features with high discriminability, strong robustness and good generalization ability from complex and variable spectral information to effectively solve the serious spectral confusion problem between tea gardens and evergreen vegetation; developing new technologies that can solve these two major problems at the same time is of urgent need and great significance for efficient, accurate and large-scale tea garden identification and dynamic monitoring. SUMMARY

[0008] In order to solve the technical problems of insufficient temporal integrity of optical data, serious spectral confusion between tea gardens and evergreen vegetation, and effective identification index designed for the unique and complex spectral-temporal characteristics of tea gardens in the prior art, the present application proposes a tea garden identification method and system based on multi-source remote sensing data, which significantly improves the stability of large-scale tea garden identification by establishing a "radar timing-optical classification" collaborative framework and a phenology-driven index.

[0009] To achieve this goal, the present application adopts the following technical solutions.

[0010] A tea garden identification method based on multi-source remote sensing data, the method comprising the following steps: S1: obtaining time series radar data and optical remote sensing data of the target area; S2: extracting key growth periods of tea gardens based on radar data, including the beginning and end of the growing season; S3: calculating vegetation index VI and water body index WI using optical data, and selecting the optimal VI and WI combination in terms of discriminability through time-weighted dynamic time warping method; S4: determining the tea garden identification index as follows: wherein Ω1 is a linear operation, Ω2 is a nonlinear operation, T represents tea garden samples, G represents evergreen vegetation samples, and DTW is a dynamic time warping method; S5: realizing tea garden identification based on the tea garden identification index and an ensemble learning model.

[0011] In addition, in the tea garden identification method based on multi-source remote sensing data, the key growth period of the tea garden is extracted based on radar data, which comprises: a) radiation calibration, geometric correction and filtering processing are performed on the radar data to construct a VV / VH polarization backscattering coefficient time series; b) for the VV / VH polarization backscattering coefficient time series, filtering smoothing is performed, and the calculation formula of filtering smoothing is: , wherein, is the value after filtering smoothing, is the original value, is the convolution coefficient, is the normalization coefficient, is the half-width of the filtering window, i is the offset relative to the center j of the sliding window, and the value range is -m to m; The beginning period SOS and the end period EOS of the growth season are determined based on the first derivative extreme point, and the method is as follows: , , wherein, the argmax function is used to find the maximum value point of the function in the given range, is the change amount of the backscattering coefficient, is the corresponding time interval.

[0012] In addition, in the tea garden identification method based on multi-source remote sensing data, the VI and WI combination with the optimal discrimination degree is screened by the time-weighted dynamic time warping method, which comprises: Constructing the tea garden time sequence and the weighted distance matrix of the evergreen vegetation time sequence is: wherein, denotes the weighted distance between points and . The elements of the cumulative distance matrix are calculated: , The cumulative distance matrix is constructed, wherein denotes the minimum cumulative distance from the starting point of the sequence to the current position . The time-weighted dynamic time warping distance is output as: .

[0013] In addition, in the tea garden identification method based on multi-source remote sensing data, the vegetation index includes at least one of a normalized difference vegetation index NDVI, an enhanced vegetation index EVI, and a ratio vegetation index RVI, and the water body index includes at least one of a land water index LSWI, a normalized difference water index NDWI, and a modified normalized difference water index mNDWI.

[0014] In addition, in the tea garden identification method based on multi-source remote sensing data, the evergreen vegetation pre-extraction step is further included. a) For each of the normalized difference vegetation index NDVI, the enhanced vegetation index EVI, the ratio vegetation index RVI, the land water index LSWI, the normalized difference water index NDWI, and the modified normalized difference water index mNDWI, four time series statistical features, i.e., annual mean, maximum value, minimum value, and annual amplitude, are extracted, and a 24-dimensional multi-temporal feature space is constructed; b) Random forest, support vector machine, and extreme gradient boosting methods are combined to classify ground objects in the target area, and an evergreen vegetation area is extracted; c) For the classification result, a morphological processing method, including opening and closing operations and region connectivity analysis, is used to remove isolated misclassified pixels and enhance the spatial continuity of the result.

[0015] In addition, in the tea garden identification method based on multi-source remote sensing data, the tea garden identification is realized by using an ensemble learning model, which includes: The probability outputs of three classifiers, i.e., random forest, support vector machine, and extreme gradient boosting, are integrated. , wherein, is the final classification probability, is the probability prediction of the i-th classifier, and M is the number of classifiers; The feature space is combined with the terrain parameters of elevation, slope, and aspect.

[0016] In addition, in the tea garden identification method based on multi-source remote sensing data, the radar data is Sentinel-1 C-band synthetic aperture radar SAR data, and the optical data is Sentinel-2 multispectral data.

[0017] In addition, the present application further includes a tea garden identification system based on multi-source remote sensing data, which includes a data acquisition module, a phenology analysis module, an index construction module, a classification module, and an output module, wherein, The data acquisition module is used to acquire time-series radar data and optical remote sensing data of the target area; The phenology analysis module is used to extract the key growth period of the tea garden based on the radar data; The index construction module is configured to calculate vegetation indices VI and water indices WI using optical data, and screen a combination of VI and WI with the best discrimination degree through a time-weighted dynamic time warping method; and determine a tea garden identification index according to the following method: wherein Omega 1 is a linear operation, Omega 2 is a nonlinear operation, T represents a tea garden sample, G represents an evergreen vegetation sample, and DTW is a dynamic time warping method; The classification module is configured to realize tea garden identification based on the tea garden identification index and an ensemble learning model, The output module is configured to generate a tea garden spatial distribution map according to the tea garden identification result.

[0018] In addition, in the tea garden identification system based on multi-source remote sensing data, the phenology analysis module extracts key growth periods of tea gardens based on radar data, including: a) performing radiation calibration, geometric correction and filtering processing on the radar data to construct a VV / VH polarization backscattering coefficient time series; b) performing filtering smoothing on the VV / VH polarization backscattering coefficient time series, and the calculation formula of the filtering smoothing is: , wherein, is the value after filtering smoothing, is the original value, is a convolution coefficient, is a normalization coefficient, is a filter window half-width, and i is an offset relative to the center j of the sliding window, and the value range is -m to m; The first derivative extreme point is used to determine the start of growth season (SOS) and the end of growth season (EOS), and the method is as follows: , , wherein, the argmax function is used to find the maximum value point of the function in the given range, is the change amount of the backscattering coefficient, is the corresponding time interval.

[0019] In addition, in the tea garden identification system based on multi-source remote sensing data, the index construction module is configured to calculate vegetation indices VI and water indices WI using optical data, and screen a combination of VI and WI with the best discrimination degree through a time-weighted dynamic time warping method, including: Constructing a tea garden time sequence and a weighted distance matrix of an evergreen vegetation time sequence is: wherein, represents a point and Weighted distance between them; Calculate the elements of the cumulative distance matrix: , Construct the cumulative distance matrix ,in This represents the distance from the start of the sequence to the current position. The minimum cumulative distance; The output time-weighted dynamic time warp distance is: .

[0020] The technical effects of this invention include the following.

[0021] The Time-Weighted Dynamic Time Warping (TWDTW) method not only considers the similarity of time series shapes but also incorporates time dimension matching constraints into the calculation, making the technical solution of this invention more sensitive to seasonal changes. This characteristic is particularly important when distinguishing vegetation types with similar spectral characteristics but different phenological rhythms (such as tea gardens and other evergreen vegetation). TWDTW more accurately quantifies the distance between time series by finding the optimal alignment path between two time series while penalizing excessive distortion on the time axis. First, time series of multiple vegetation indices and water indices need to be calculated. Then, the index time series of tea garden samples and other evergreen vegetation are extracted separately to ensure the integrity and comparability of the time series data. The time series distance between tea garden and other evergreen vegetation samples is calculated using TWDTW, and the vegetation index (VI) and water index (WI) with the largest TWDTW distance are selected through statistical analysis. These indices will be used for subsequent TPRI index construction.

[0022] The rationality of the TWDTW distance maximization principle is reflected on two levels: From a phenological perspective, the TWDTW distance can accurately measure the temporal differences between tea gardens and other evergreen vegetation throughout their growth cycle. A larger TWDTW distance means that the constructed tea garden identification index can better capture the unique growth patterns of tea gardens. From the perspective of index construction objectives, since the standard for selecting the optimal VI and WI is the maximum TWDTW distance, their combination (i.e., TPRI) should produce an even larger TWDTW distance. Otherwise, it indicates that this combination weakens rather than enhances the distinguishing ability of the original index, violating the original intention of index construction. Therefore, by verifying whether the TWDTW distance of the tea garden identification index exceeds that of all individual indices, the consistency of the evaluation criteria is ensured, and the combination process is guaranteed to truly play an optimization role.

[0023] The present application systematically explores and evaluates the distinguishing ability of different index combinations, and finally selects the index combination that can best distinguish tea garden and evergreen forest, and thereby constructs the TPRI index for identifying tea garden. The selection of one index from VI and one index from WI, instead of selecting the two largest D_index from all indexes, is based on the complementarity of the two types of indexes in reflecting the characteristics of ground objects: VI mainly reflects the physiological characteristics of vegetation such as biomass and photosynthesis, while WI focuses on the water content of vegetation and soil.

[0024] The construction of tea garden identification index TPRI is based on the following core ideas: first, the complementarity of the selected VI and WI indexes is utilized to comprehensively consider the growth conditions and water characteristics of vegetation; second, the unique spectral and phenological characteristics of tea garden are considered to highlight its differences from other evergreen vegetation; third, the distinguishing ability in the key growth period is strengthened to improve the accuracy of tea garden remote sensing classification.

[0025] In the rough classification stage, the research objects mainly involve construction land, water body, cultivated land and evergreen vegetation, which have significant spectral differences between these categories, so the performance of each classifier is less different in the preliminary identification. Therefore, the use of three classifiers and the selection of the best performance can ensure the accuracy of the preliminary classification results, thereby providing reliable prior information for subsequent fine classification. In the fine classification stage, the present application uses a multi-model ensemble strategy to integrate the probability prediction results of multiple classifiers through a soft voting mechanism to fully utilize the complementary advantages of different methods to reduce the uncertainty of a single algorithm, thereby significantly improving the overall classification accuracy.

[0026] In addition, in the fine classification stage, an ensemble learning method based on soft voting mechanism is used. First, on the basis of rough classification, the evergreen vegetation category is further subdivided into tea garden and other evergreen vegetation. This method uses the time series characteristics of tea garden separation index in the key phenological period, combined with terrain parameters (elevation, slope and aspect) to construct a feature space, which integrates three machine learning algorithms. Soft voting mechanism is an advanced ensemble strategy, unlike traditional hard voting (minority submits to majority), it fully utilizes the probability prediction information of each classifier. For each sample, each classifier will output a probability distribution, and the soft voting mechanism obtains the final classification result by weighted average of these probability distributions, so it has higher accuracy.

[0027] For the salt and pepper noise and fine pixels that may appear after classification, the application applies morphological processing technology, including opening and closing operation and region connectivity analysis, to remove isolated misclassified pixels and enhance the spatial continuity and visual expression effect of the results. In the post-processing stage, special processing is also carried out for the class imbalance problem to ensure effective identification of small area evergreen vegetation patches. The final evergreen vegetation distribution map not only has high classification accuracy, but also maintains good spatial integrity, laying a solid foundation for subsequent fine distinction of tea gardens and other evergreen vegetation.

[0028] The "radar timing, optical classification" collaborative framework proposed by the application realizes the functional complementarity of multi-source remote sensing data in theory. Unlike traditional simple data superposition or feature level fusion, the application performs task division in the time and space dimensions. Radar data assumes the function of defining the phenological period in the time dimension, and optical data is responsible for the class distinction task in the spatial dimension. The core advantage of this design is to fully exert the inherent characteristics of different data sources, that is, the all-weather observation capability of SAR data ensures the accurate capture of the key growth period, and the rich spectral information of optical data provides a basis for fine classification. This division of labor mode avoids the common problems of information redundancy and noise accumulation in multi-source data fusion. The application of radar data in the phenological analysis stage essentially utilizes its sensitivity to vegetation structure changes to construct a time constraint condition, providing an optimal time window for subsequent feature extraction of optical data. The introduction of this time sequence constraint significantly improves the discrimination of optical index in tea garden identification, verifying the effectiveness of multi-source data collaboration.

[0029] In addition, the application intentionally selects an ensemble learning model based on traditional machine learning rather than using a deep learning method, which is because the latter has strong feature extraction capability, but has large calculation amount and high resource consumption, making it difficult to support efficient classification and real-time monitoring of large-scale tea gardens. In actual application scenarios, especially in areas with limited computing resources, an algorithm with high efficiency and low resource consumption is more valuable. Therefore, the application effectively balances the advancement of the classification method and the feasibility of implementation under limited infrastructure conditions.

[0030] This invention, by constructing the TPRI index and establishing a multi-source data collaborative method for tea garden identification, not only achieves innovation in technical methods but also has significant practical value in the field of agricultural and forestry remote sensing applications. This method provides a new approach to solving the technical challenge of precise identification of economic crops and has positive significance for promoting the in-depth application of remote sensing technology in agriculture and forestry. From an industrial application perspective, accurate tea garden distribution information plays a crucial supporting role in tea industry planning, market analysis, and insurance assessment. This method can provide reliable basic data for relevant government departments and enterprises, supporting industrial policy formulation and business decisions. Furthermore, the crop identification approach based on phenological characteristics explored in this invention provides a new methodological framework for crop remote sensing monitoring, possessing strong theoretical innovation and methodological promotion value. From an engineering application perspective, future development should focus on the standardized deployment and industrial application of the algorithm. Exploring the direct deployment of the entire algorithm process to a cloud computing platform, utilizing distributed computing resources to process large-scale remote sensing data, and solving the problem of computing resource limitations is also important. Cloud platform deployment not only overcomes the limitations of single-machine computing power but also enables standardized algorithm services, providing unified tea garden identification capabilities for different users and supporting dynamic monitoring of tea gardens at regional, national, and even global scales. Simultaneously, this invention helps establish a comprehensive technical standard system and operational mechanism, including standardized data preprocessing procedures, parameter setting specifications, and accuracy evaluation standards. It also facilitates the development of user-friendly software tools and data service interfaces, promoting the widespread application and industrialization of the technology. Attached Figure Description

[0031] Figure 1 This is a flowchart illustrating a tea garden identification method based on multi-source remote sensing data according to a specific embodiment of the present invention.

[0032] Figure 2 This is a schematic diagram of a tea garden identification method based on multi-source remote sensing data according to a specific embodiment of the present invention.

[0033] Figure 3 This is a detailed flowchart of a tea garden identification method based on multi-source remote sensing data according to a specific embodiment of the present invention. Detailed Implementation

[0034] The present invention will now be described in detail with reference to the accompanying drawings.

[0035] The following detailed exemplary embodiments are disclosed. However, the specific structural and functional details disclosed herein are merely for the purpose of describing exemplary embodiments.

[0036] However, it should be understood that the present invention is not limited to the specific exemplary embodiments disclosed, but covers all modifications, equivalents, and substitutions falling within the scope of this disclosure. Throughout the description of the drawings, the same reference numerals denote the same elements.

[0037] Referring to the structure, proportion, size, etc. shown in the drawings attached to the present specification, which are only used to cooperate with the content disclosed in the present specification for understanding and reading by those skilled in the art, and are not used to limit the conditions that the present application can be implemented, so they do not have technical significance. Any modification of structure, change of proportion relationship or adjustment of size, which does not affect the effect that the present application can produce and the purpose that the present application can achieve, should still fall within the scope of the technology disclosed by the present application. At the same time, the positional limitation terms used in the present specification are only for the convenience of clear description, and are not used to limit the scope of the present application. The change or adjustment of the relative relationship, without substantially changing the technical content, is also considered as the scope of the present application.

[0038] It should also be understood that the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. In addition, it should be understood that when a component or unit is referred to as being "connected" or "coupled" to another component or unit, it can be directly connected or coupled to the other component or unit, or there can be an intermediate component or unit. In addition, other words used to describe the relationship between components or units should be understood in the same way (for example, "between" versus "directly between", "adjacent" versus "directly adjacent", etc.).

[0039] Figure 1 A flowchart of a tea garden identification method based on multi-source remote sensing data according to an embodiment of the present application is shown in Figure 1. As shown in the figure, the embodiment of the present application includes a tea garden identification method based on multi-source remote sensing data, which includes the following steps: S1: Obtain time-series radar data and optical remote sensing data of the target area; S2: Extract the key growth period of the tea garden based on the radar data, including the beginning and end of the growth season; S3: Calculate the vegetation index VI and water body index WI using optical data, and select the optimal VI and WI combination through time-weighted dynamic time warping method; S4: Determine the tea garden identification index as follows: Where TRPI is the tea garden identification index, Ω1 is a linear operation, Ω2 is a nonlinear operation, T represents the tea garden sample, G represents the evergreen vegetation sample, and DTW is the dynamic time warping method; S5: Realize tea garden identification based on the tea garden identification index and the ensemble learning model.

[0040] The application innovatively determines a novel tea garden identification index TPRI, which combines a vegetation index and a water body index to accurately draw a tea garden map. The fundamental reason why TPRI can accurately identify tea gardens is that it accurately captures the essential characteristics of tea gardens as artificial management ecosystems. Unlike natural evergreen forests, tea gardens undergo frequent human disturbance during the growth cycle, especially periodic picking activities, which significantly change the canopy structure and water conditions of the vegetation, forming a unique spectral response pattern. Traditional single vegetation indices can only reflect changes in biomass and cannot fully characterize this composite feature. TPRI combines vegetation indices with water body indices, reflecting not only changes in biomass but also changes in water conditions closely related to tea garden management, achieving comprehensive expression of the multi-dimensional characteristics of tea garden ecosystems. More importantly, the application uses an index optimization mechanism based on time-weighted dynamic time warping distance to quantify the temporal differences between tea gardens and other evergreen vegetation throughout the growth cycle, ensuring that the selected index combination has the strongest class discrimination ability. This construction strategy based on maximizing phenological temporal differences enables TPRI to stably identify the unique growth rhythm characteristics of tea gardens under different geographical and climatic conditions, thereby achieving high-precision identification of tea gardens in complex evergreen vegetation backgrounds.

[0041] The establishment of TPRI first investigates the key growth stages of tea gardens and then designs appropriate methods to combine vegetation indices and water body indices to maximize the information differences between tea gardens and other evergreen vegetation. The workflow of the specific embodiments of the application consists of two parts: (1) derivation of key growth periods, which includes phenological analysis of tea gardens and other land cover types, and extraction of evergreen vegetation. (2) TPRI determination, application in tea garden mapping, and verification of tea garden mapping results.

[0042] In addition, in the tea garden identification method based on multi-source remote sensing data of the specific embodiments of the application, the extraction of the key growth period of tea gardens based on radar data includes: a) Radiometric calibration, geometric correction and filtering processing are performed on the radar data to construct a VV / VH polarization backscattering coefficient time series; b) For the VV / VH polarization backscattering coefficient time series, filtering smoothing is performed, and the calculation formula of filtering smoothing is: , wherein, is the value after filtering smoothing, is the original value, i.e., the value of the VV / VH polarization backscattering coefficient time series, is the convolution coefficient, is the normalization coefficient, is the half width of the filter window, i is the offset relative to the center j of the sliding window, and i ranges from -m to m; The method realizes data smoothing through local polynomial fitting, and has the advantages of better retaining signal detail characteristics while reducing noise compared with traditional mean filtering or median filtering.

[0043] In the specific embodiments of the application, a dynamic threshold can also be calculated based on the smoothed scattering coefficient time series for screening effective observation data and eliminating outliers: Threshold = min + 0.4 * (max-min) Wherein, is the threshold, and min and max are the minimum and maximum values of the time series respectively. The dynamic threshold method has better adaptability than the fixed threshold, and can automatically adjust according to the growth characteristics of tea gardens in different regions. , The first derivative reflects the change rate of the growth stage, and the extreme point corresponds to the key turning point of the vegetation growth state.

[0044] The start of growth season (SOS) and the end of growth season (EOS) are determined based on the extreme point of the first derivative, as follows: , , Wherein, the argmax function is used to find the maximum point of the function in the given range, is the change amount of the backscattering coefficient, is the corresponding time interval.

[0045] Accurate extraction of key growth periods of tea garden is the key prerequisite for realizing accurate identification of tea garden. Considering the limitation that optical remote sensing data is easily affected by weather conditions in areas with much cloud and rain, in the specific embodiment of the present application, a method for identifying key phenological periods of tea garden based on SAR radar data is proposed. SAR data has all-weather and all-time observation capability, and its backscattering coefficient is sensitive to the change of vegetation structure, which can effectively capture the structural feature change of tea garden in key stages such as growth and picking. Tea garden is an artificially cultivated economic crop, and its growth cycle has obvious seasonal characteristics, mainly including the start of season (SOS) and the end of season (EOS). SOS usually corresponds to the spring tea bud germination and rapid growth stage, and EOS corresponds to the slow growth and dormancy stage in autumn.

[0046] In the specific embodiment of the present application, first, the Sentinel-1 SAR data is radiometrically calibrated, geometrically corrected and filtered, and the backscattering coefficient time series of VV and VH polarization is constructed. VV polarization mainly reflects the vertical structure characteristics of vegetation, and VH polarization is more sensitive to vegetation volume scattering, and the combination of the two polarization modes can comprehensively describe the structure and growth state change of tea garden. Then, Savitzky-Golay (SG) filtering algorithm is used to smooth the backscattering coefficient time series to eliminate noise and highlight the trend of phenological change. SG filtering realizes data smoothing through local polynomial fitting, and its mathematical expression is:

[0047] wherein, is the smoothed value, is the original data, is the convolution coefficient, is the normalization coefficient, is the window half-width, i is the offset relative to the center j of the sliding window, and the value range is -m to m.

[0048] The dynamic threshold is calculated based on the smoothed time series, which is used to filter the effective observation data and eliminate outliers: , wherein, is the threshold, and The minimum and maximum values of the time series, respectively. The threshold position corresponding to the coefficient of 0.4 can effectively distinguish the normal growth state and the abnormal state of the tea garden: the observation value below the threshold value usually corresponds to signal attenuation caused by cloudy and rainy weather, terrain shadow effect or data quality problem, while the observation value above the threshold value represents the real growth state of the tea garden. The coefficient is selected through comparative analysis in the specific embodiment of the application, which can more efficiently distinguish the state of the tea garden. The dynamic threshold method has better adaptability than the fixed threshold, and can automatically adjust according to the growth characteristics of the tea garden in different regions. The first derivative of the phenology curve is analyzed to identify the key time nodes as follows: , wherein, is the backscattering coefficient at time t, is the first derivative, which reflects the change rate of the growth stage, and the extreme point corresponds to the key turning point of the vegetation growth state.

[0049] The start of the growth season (SOS) is defined as the maximum point of the first derivative in the period from the beginning of the year to the middle of the year, .

[0050] The end of the growth season (EOS) is defined as the minimum point of the first derivative in the period from the middle of the year to the end of the year,

[0051] , wherein, is a function for finding the maximum point of the function in the given range, is the change amount of the backscattering coefficient, is the corresponding time interval.

[0052] In addition, in the tea garden identification method based on multi-source remote sensing data in the specific embodiment of the application, the VI and WI combination with the best discrimination degree is screened out by the time-weighted dynamic time warping method, which includes: Constructing the time sequence of the tea garden and the time sequence of the evergreen vegetation The weighted distance matrix is: , wherein denotes the weighted distance between points and ; wherein, can be given by prior knowledge or an experience function, which is used to measure the weighting degree of the time deviation (or the growth period difference, etc.) on the matching error at the i-th and j-th observation time.

[0053] The elements of the cumulative distance matrix are calculated as follows: , Constructing cumulative distance matrix wherein denotes the minimum cumulative distance from the start of the sequence to the current position ; The output time-weighted dynamic time warping distance is: .

[0054] The TWDTW distance is smaller, indicating that the time series of the tea garden and other evergreen vegetation are more similar; on the contrary, the distance is larger, indicating that the two types of samples have large differences in the time series track.

[0055] In addition, in the tea garden recognition method based on multi-source remote sensing data, the vegetation index includes at least one of normalized difference vegetation index NDVI, enhanced vegetation index EVI, and ratio vegetation index RVI, and the water body index includes at least one of land water index LSWI, normalized water index NDWI, and modified normalized difference water index mNDWI.

[0056] In addition, the tea garden recognition method based on multi-source remote sensing data further comprises an evergreen vegetation pre-extraction step: a) For each of the normalized difference vegetation index NDVI, the enhanced vegetation index EVI, the ratio vegetation index RVI, the land water index LSWI, the normalized water index NDWI, and the modified normalized difference water index mNDWI, four time series statistical features are extracted: annual mean, maximum value, minimum value, and annual amplitude, and a 24-dimensional multi-temporal feature space is constructed. This time series compression strategy condenses high-dimensional time series into low-dimensional statistics, retains key phenological information, greatly reduces data dimension and computational complexity, and avoids the problem of dimension disaster; b) Combined with random forest, support vector machine and extreme gradient boosting method, the ground object classification of the target area is carried out, and the evergreen vegetation area is extracted; c) For the classification result, a morphological processing method including opening and closing operation and region connectivity analysis is used to remove isolated misclassified pixels and enhance the spatial continuity of the result.

[0057] In the embodiment of the present application, six key vegetation indices are calculated from the time series data: normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), ratio vegetation index (RVI), land water index (LSWI), normalized difference water index (NDWI) and modified normalized difference water index (mNDWI). For each index, four time series statistical features are extracted: annual mean, maximum, minimum and annual amplitude (maximum-minimum), and a 24-dimensional multi-temporal feature space is constructed. This combination of multi-index and multi-feature design can fully capture the differences in phenology and spectral characteristics between evergreen vegetation and other ground object categories.

[0058] Subsequently, using these feature space data, combined with three machine learning classifiers-random forest (RF), support vector machine (SVM) and extreme gradient boosting (XGBoost), the ground object classification of the study area is carried out. To ensure the reliability of the classification results, stratified random sampling strategy is used to generate training samples, and independent validation samples are set to evaluate the classification effect. Through comprehensive evaluation indexes such as overall accuracy (OA), F1 score, Kappa coefficient and detailed confusion matrix analysis, the performance of the three classifiers is systematically compared, and the best classification result is selected. Among them, random forest is based on decision tree ensemble, is good at processing high-dimensional features and nonlinear relationships, and has strong robustness to noise and outliers; support vector machine realizes complex nonlinear classification boundary through kernel function mapping, and still maintains good generalization ability under small sample condition; extreme gradient boosting optimizes loss function through gradient boosting framework, and can automatically select features and handle class imbalance problem.

[0059] In one embodiment of the present application, in view of the salt and pepper noise and fine pixels that may occur after classification, morphological processing techniques including opening and closing operations and region connectivity analysis are applied in the embodiment of the present application to remove isolated misclassified pixels and enhance the spatial continuity and visual expression effect of the results. In the post-processing stage, special processing is also carried out for the class imbalance problem to ensure effective identification of small-area evergreen vegetation patches. The final evergreen vegetation distribution map not only has high classification accuracy, but also maintains good spatial integrity, laying a foundation for subsequent fine distinction of tea garden and other evergreen vegetation.

[0060] In addition, in the tea garden recognition method based on multi-source remote sensing data of the present application, tea garden recognition is realized by using an ensemble learning model, which includes: Integrating the probability outputs of the three classifiers of random forest, support vector machine and extreme gradient boosting: , Among them, is the final classification probability, Pi is the probability prediction of the ith classifier, M is the number of classifiers; The feature space is constructed in combination with the terrain parameters of elevation, slope, and aspect.

[0061] In addition, in the tea garden identification method based on multi-source remote sensing data, the radar data is Sentinel-1 C-band synthetic aperture radar (SAR) data, and the optical data is Sentinel-2 multispectral data.

[0062] For example, in a specific embodiment, the present application takes Sentinel-1 radar data and Sentinel-2 optical data as the basic data source. For Sentinel-2 optical data, it is necessary to use the quality assessment band (Quality Assessment Band) to perform volume screening and cloud removal processing. In order to ensure the spatial consistency of multi-temporal data, accurate geometric correction can be performed first, and the multi-period data can be constructed into a time series data set in chronological order. For Sentinel-1 SAR data, it is necessary to perform radiometric calibration to convert the original DN value (DN value, Digital Number is the original brightness value of each pixel in remote sensing image) to backscattering coefficient, and to perform terrain correction to eliminate the influence of terrain undulation on radar signal. In order to improve the data quality, it is also necessary to use appropriate filtering method to reduce the speckle noise of SAR image.

[0063] In order to identify the most discriminative TPRI index, it is necessary to calculate the time series similarity of different remote sensing indexes in the key growth period. The specific embodiment of the present application first calculates the time series of three vegetation indexes (EVI, RVI, NDVI) and three water body indexes (NDWI, mNDWI, LSWI), and standardizes each index, so as to avoid the influence of the order of magnitude difference between different indexes on the TWDTW distance comparison. Then, the standardized index time series of tea garden samples and other evergreen vegetation are extracted respectively, and the present application uses time-weighted dynamic time warping algorithm (Time-Weighted Dynamic Time Warping, TWDTW) to quantify the time sequence difference between the two types of samples. This method is an improvement of the traditional DTW method, which can better handle the seasonal changes of agricultural phenology characteristics by introducing a time weight function. This method can effectively identify vegetation types with similar spectral characteristics but different phenology rhythms, and through statistical analysis, the vegetation index (VI) and water body index (WI) with the largest TWDTW distance are selected, so as to ensure that the selected index combination has the best discriminative ability in tea garden identification, and lay a foundation for subsequent TPRI index construction.

[0064] The tea plantations in the study area are classified using Sentinel-2 optical data with cloud cover <10%. In one embodiment of the present application, the first 10 bands of Sentinel-2 optical data (bands 2-8, 8A and bands 11-12) are used because they are designed for vegetation monitoring. In order to preserve the detailed spatial and spectral information provided by the 10m bands, the original bands with a spatial resolution of 20m are resampled to a spatial resolution of 10m. Therefore, a total of 10 bands of S2 optical data with a spatial resolution of 10m are used for tea plantation identification.

[0065] Figure 2 A structural schematic diagram of a tea plantation identification method based on multi-source remote sensing data according to an embodiment of the present application. As shown in the figure, the present embodiment also includes a tea plantation identification system based on multi-source remote sensing data, which includes a data acquisition module, a phenology analysis module, an index construction module, a classification module, and an output module, wherein, The data acquisition module is used to acquire the time-series radar data and optical remote sensing data of the target area. The phenology analysis module is used to extract the key growth period of tea plantations based on radar data. The index construction module is used to calculate vegetation index VI and water index WI using optical data, and to screen the optimal VI and WI combination with the highest discrimination degree through a time-weighted dynamic time warping method; and to determine the tea plantation identification index as follows: wherein Ω1 is a linear operation, Ω2 is a nonlinear operation, T represents tea samples, G represents evergreen vegetation samples, and DTW is a dynamic time warping method. The classification module realizes tea plantation identification based on the tea plantation identification index and an ensemble learning model. The output module is used to generate a tea plantation spatial distribution map according to the tea plantation identification result.

[0066] The present application uses the strategy of selecting the index with the largest TWDTW distance from the VI and WI indexes to construct TPRI, which fully utilizes the complementary advantages of vegetation index reflecting biomass characteristics and water index reflecting water conditions. Through linear and nonlinear combination methods, the discrimination ability of different index combinations is evaluated, and the combination method that best distinguishes tea plantations and evergreen forests is finally selected to construct the TPRI index.

[0067] After the construction, the application verifies the effectiveness by comparing the TWDTW distance of TPRI and each single index in the key growth period. The verification logic is that if the TWDTW distance of TPRI exceeds all single indexes, it proves that the combination process indeed enhances the discrimination ability of the original index, and achieves the expected goal of index construction; otherwise, it means that the combination method needs to be further optimized. This verification method ensures the consistency of evaluation criteria and the reliability of combination effect.

[0068] In addition, in the tea garden recognition system based on multi-source remote sensing data of the application, the phenology analysis module extracts the key growth period of tea garden based on radar data, including: a) Radiometric calibration, geometric correction and filtering processing are performed on the radar data to construct the VV / VH polarization backscattering coefficient time series; b) For the VV / VH polarization backscattering coefficient time series, filtering smoothing is performed, and the calculation formula of filtering smoothing is: , Wherein, is the value after filtering smoothing, is the original value, i.e. the VV / VH polarization backscattering coefficient time series, is the convolution coefficient, is the normalization coefficient, is the filter window half-width, i is the offset relative to the center j of the sliding window, and the value range is-m to m; The beginning of the growth season SOS and the end of the season EOS are determined based on the first derivative extreme point, and the method is as follows: , , Wherein, the argmax function is used to find the maximum value point of the function in the given range, is the change amount of backscattering coefficient, is the corresponding time interval. Wherein the beginning of the growth season SOS is in the first half of the year (0~183 days), and the end of the season EOS is in the second half of the year (184~365 days), it should be noted that here the day is not forced to synchronize with the date of natural year.

[0069] In addition, in the tea garden recognition system based on multi-source remote sensing data of the application, the index construction module is used to calculate the vegetation index VI and water index WI by using optical data, and to screen the optimal VI and WI combination by time-weighted dynamic time warping method, including: Construct tea garden time series The weighted distance matrix of evergreen vegetation time series is: ,​ representing the weighted distance between points and ; The elements of the cumulative distance matrix are computed as: , The cumulative distance matrix is constructed as where represents the minimum cumulative distance from the sequence start to the current position ; The time-weighted dynamic time warping distance is output as: .

[0070] Figure 3 is a detailed flowchart of a tea plantation identification method based on multi-source remote sensing data according to an embodiment of the present application. The following will describe the embodiment of the present application in more detail. Figure 3

[0071] As shown in the figure, the embodiment of the present application includes preprocessing of remote sensing data. In the embodiment of the present application, first, the Sentinel-1 SAR data is subjected to data filtering, radiation correction, geometric correction and filtering processing. On the other hand, the Sentinel-2 (S2) MSI images with cloud cover <10% are used to classify the tea plantations in the region to be analyzed. The S2 images can be collected, for example, from Google Earth Engine (GEE), covering all the regions to be analyzed. In the embodiment of the present application, the first 10 bands (bands 2-8, 8A and bands 11-12) are used because they are designed for vegetation monitoring. In order to preserve the detailed spatial and spectral information provided by the 10m bands, the original bands with a spatial resolution of 20m are resampled to a spatial resolution of 10m. That is, a total of 10 S2 bands with a spatial resolution of 10m are used for tea plantation mapping.

[0072] For S2 images, data filtering, cloud and shadow removal, geometric correction, resampling operations are performed in combination with their data characteristics, and vegetation indices and water body indices are calculated based thereon.

[0073] Specifically, for Sentinel-1 SAR data, the backscattering coefficient time series of VV and VH polarization are constructed. VV polarization mainly reflects the vertical structure characteristics of vegetation, and VH polarization is more sensitive to vegetation volume scattering, and the combination of the two polarization modes can comprehensively describe the structure and growth state changes of tea plantations. Subsequently, the SG filtering algorithm is used to smooth the backscattering coefficient time series to eliminate noise and highlight the trend of phenological change.

[0074] ​After the SG filtering operation, a dynamic threshold calculation can be further performed, and the basic mode is to calculate a dynamic threshold based on the smoothed time series, which is used to screen valid observation data and eliminate abnormal values. Based on the calculation result of the dynamic prediction, the start of growth season (SOS) and the end of growth season (EOS) can be determined.

[0075] Next, the core part in the specific embodiment of the present application, i.e., the construction of the tea garden recognition index, is performed, and remote sensing information indexes are extracted from the Sentinel-2 (S2) MSI image. Specifically, six key vegetation indexes or water body indexes, i.e., NDVI, EVI, RVI, LSWI, NDWI and mNDWI, are calculated from the time series data. For each index, four time series statistical features, i.e., annual mean, maximum value, minimum value and annual amplitude (maximum value-minimum value), are extracted, and a 24-dimensional multi-temporal feature space is constructed. In the construction of the multi-temporal feature space, statistical features of these indexes can be extracted, such as annual mean, maximum value, minimum value and annual amplitude.

[0076] In the specific embodiment of the present application, the time series of the vegetation indexes and the water body indexes are standardized, so as to avoid the influence of the order of magnitude difference between different indexes on the TWDTW distance comparison. Then, the standardized index time series of the tea garden samples and other evergreen vegetation are extracted respectively, and the TWDTW is used to quantify the time series difference between the two types of samples. Through statistical analysis, the vegetation index (VI) and the water body index (WI) with the largest TWDTW distance are screened out, so as to ensure that the selected index combination has the optimal discrimination ability in tea garden recognition.

[0077] Then, the parameter space is exhausted, and different combination sequences are constructed by using the linear combination or nonlinear combination of VI and WI, and the tea garden recognition index is obtained on this basis.

[0078] Further, in the specific embodiment of the present application, the time series curve of the tea garden recognition index can also be used to classify the ground objects in the analyzed area by combining the RF, SVM and XGBoost methods. The soft voting mechanism is used to obtain the tea garden spatial classification result. In order to ensure the reliability of the distribution result, the stratified random sampling strategy is used to generate the training samples, and the independent verification samples are set to evaluate the distribution effect. The performance of the three classifiers is compared systematically through comprehensive evaluation indexes such as overall accuracy (OA), F1 score, Kappa coefficient and detailed confusion matrix analysis, and the optimal classification result is selected.

[0079] Therefore, the present application has the following technical effects.

[0080] The time-weighted dynamic time warping method TWDTW not only considers the similarity of the shape of the time series, but also takes into account the matching constraint of the time dimension in the calculation, so that the technical scheme of the application is more sensitive to seasonal changes. This feature is particularly important in distinguishing vegetation types with similar spectral characteristics but different phenological rhythms (such as tea gardens and other evergreen vegetation). The time-weighted dynamic time warping method TWDTW more accurately quantifies the distance between time series by finding the optimal alignment path between two time series while penalizing excessive distortion on the time axis. First, the time series of various vegetation indices and water indices need to be calculated. Then, the index time series of tea garden samples and other evergreen vegetation are extracted respectively, ensuring the integrity and comparability of the time series data. The time series distance between tea garden and other evergreen vegetation samples is calculated by the time-weighted dynamic time warping method TWDTW, and the vegetation index (VI) and water index (WI) with the largest TWDTW distance are selected through statistical analysis, which will be used for subsequent TPRI index construction.

[0081] The rationality of the TWDTW distance maximization principle is reflected in two aspects: from the perspective of phenology, TWDTW distance can accurately measure the timing difference between tea gardens and other evergreen vegetation throughout the growth cycle, and a larger TWDTW distance means that the constructed tea garden recognition index can better capture the unique growth pattern characteristics of tea gardens; from the perspective of index construction, since the standard for selecting the optimal VI and WI is to maximize the TWDTW distance, their combination (i.e., TPRI) should produce a larger TWDTW distance, otherwise it means that this combination weakens rather than enhances the discrimination ability of the original index, which goes against the original intention of index construction. Therefore, by verifying whether the TWDTW distance of the tea garden recognition index exceeds all single indices, both the consistency of the evaluation standard and the true optimization effect of the combination process are ensured.

[0082] The construction of the TPRI index determined in the specific embodiment of the application is based on the TWDTW distance maximization principle, which has deep theoretical significance. The TWDTW algorithm can more accurately quantify the timing difference between land cover types with similar spectral characteristics but different phenological rhythms by introducing time weights. In the specific application of tea garden recognition, this method successfully captures the unique timing response pattern of tea gardens during the picking period, which is due to the periodic disturbance of artificial management activities on the canopy structure of the vegetation.

[0083] In terms of technical operability, the key steps involved in the construction process of the determined TPRI index, such as TWDTW calculation and SG filtering, have mature open-source implementations, which facilitate technical promotion and replication. In contrast, deep learning methods often require complex network architecture design, hyperparameter tuning, and large-scale GPU cluster support. Therefore, another advantage of the specific embodiments of the present invention is strong interpretability, with each processing step and parameter setting having a clear biophysical meaning, facilitating understanding and adjustment, and improving adaptability in different application scenarios.

[0084] From an ecological perspective, tea gardens, as artificially managed agroforestry systems, have their phenological characteristics influenced by both natural environmental regulation and human management measures. Frequent picking activities result in tea gardens exhibiting distinct canopy dynamic changes during the growing season, which is different from natural evergreen vegetation. This difference is the biological basis for the TPRI index to effectively distinguish tea gardens from other evergreen vegetation. By combining vegetation indices with water body indices, TPRI not only reflects the biomass changes of tea gardens but also captures the changes in water conditions closely related to tea garden management, achieving comprehensive expression of the multi-dimensional characteristics of tea garden ecosystems.

[0085] The present invention systematically explores and evaluates the discrimination ability of different index combinations, ultimately selecting the index combination that best distinguishes tea gardens from evergreen forests, and thereby constructing the TPRI index specifically for identifying tea gardens. The selection of one index from VI and WI respectively, rather than only selecting the two largest D_index from all indices, is based on the complementarity of these two types of indices in reflecting ground features: VI mainly reflects the biomass, photosynthesis, and other physiological characteristics of vegetation, while WI focuses on the water content of vegetation and soil.

[0086] The construction of the tea garden identification index TPRI is based on the following core ideas: first, utilize the complementarity of the selected VI and WI indices to consider vegetation growth conditions and water characteristics comprehensively; second, consider the unique spectral and phenological characteristics of tea gardens to highlight their differences from other evergreen vegetation; third, strengthen the discrimination ability in key growth periods to improve the accuracy of tea garden remote sensing classification.

[0087] In the coarse classification stage, the research objects mainly involve construction land, water body, cultivated land and evergreen vegetation, etc. These categories have significant spectral differences, so the performance of each classifier is less different when initially identified. Therefore, by using three classifiers and selecting the best-performing results, the accuracy of the preliminary classification results can be ensured, thereby providing reliable prior information for subsequent fine classification. In the fine classification stage, the invention addresses the identification challenges caused by the high spectral similarity between tea gardens and other evergreen vegetation. A multi-model ensemble strategy is used to integrate the probability prediction results of multiple classifiers through a soft voting mechanism. This strategy fully utilizes the complementary advantages of different method models to reduce the uncertainty of a single algorithm, thereby significantly improving the overall classification accuracy.

[0088] In addition, in the fine classification stage, an ensemble learning method based on a soft voting mechanism is used. First, based on the coarse classification, the evergreen vegetation category is further subdivided into tea gardens and other evergreen vegetation. This method uses the tea garden separation index in the time series characteristics of the key phenological period, combined with terrain parameters (elevation, slope and aspect) to construct a feature space, integrating three machine learning algorithms. The soft voting mechanism is an advanced ensemble strategy, unlike the traditional hard voting (minority submits to majority), which fully utilizes the probability prediction information of each classifier. For each sample, each classifier outputs a probability distribution, and the soft voting mechanism obtains the final classification result by weighted averaging these probability distributions, thus achieving higher accuracy.

[0089] For the possible salt and pepper noise and fine pixels after classification, the invention applies morphological processing techniques, including opening and closing operations and region connectivity analysis, to remove isolated misclassified pixels and enhance the spatial continuity and visual expression effect of the results. In the post-processing stage, special processing is performed for the class imbalance problem to ensure effective identification of small-area evergreen vegetation patches. The final evergreen vegetation distribution map not only has high classification accuracy but also maintains good spatial integrity, laying a solid foundation for the subsequent fine differentiation between tea gardens and other evergreen vegetation.

[0090] The "radar timing, optical classification" collaborative framework proposed by the present application realizes the functional complementation of multi-source remote sensing data in theory. Unlike traditional simple data superposition or feature level fusion, the present application divides tasks in the time and space dimensions. Radar data assumes the function of defining the phenological period in the time dimension, and optical data is responsible for the classification task in the space dimension. The core advantage of this design is to fully exert the inherent characteristics of different data sources, that is, the all-weather observation capability of SAR data ensures the accurate capture of the key growth period, and the rich spectral information of optical data provides the basis for fine classification. This division of labor mode avoids the common problems of information redundancy and noise accumulation in multi-source data fusion. The application of radar data in the phenological analysis stage is essentially to use its sensitivity to changes in vegetation structure to construct a time constraint, providing an optimal time window for subsequent feature extraction of optical data. The introduction of this time sequence constraint significantly improves the discrimination of optical indexes in tea garden identification, verifying the effectiveness of multi-source data collaboration.

[0091] In addition, the present application intentionally selects an ensemble learning model based on traditional machine learning rather than using a deep learning method. This is because although the latter has strong feature extraction capability, it has large computational load and high resource consumption, making it difficult to support efficient classification and real-time monitoring of large-scale tea gardens. In practical application scenarios, especially in areas with limited computing resources, an algorithm with high efficiency and low resource consumption is more valuable. Therefore, the present application effectively balances the advancement of the classification method and the feasibility of implementation under limited infrastructure conditions.

[0092] The present application not only achieves innovation in technical methods, but also has important practical value in the field of agricultural and forestry remote sensing applications. The method provides a new idea for solving the technical problem of fine identification of economic crops and has a positive significance for promoting the in-depth application of remote sensing technology in agriculture and forestry. From the perspective of industrial application, accurate tea garden distribution information plays an important supporting role in tea industry planning, market analysis, insurance evaluation, etc. The method can provide reliable basic data for relevant government departments and enterprises to support industrial policy making and business decision making. In addition, the crop identification idea based on phenological characteristics explored by the present application provides a new methodological framework for crop remote sensing monitoring, which has strong theoretical innovation and method promotion value. From the perspective of engineering application, future development should focus on the standardized deployment and industrial application of the algorithm. The entire algorithm process is deployed directly to the cloud computing platform, and distributed computing resources are used to process large-scale remote sensing data to solve the problem of limited computing resources. Cloud platform deployment not only solves the limitation of single computer computing power, but also realizes the standardized service of the algorithm, provides unified tea garden identification capability for different users, and supports regional, national and even global scale tea garden dynamic monitoring. At the same time, the present application helps to establish a perfect technical standard system and a business operation mechanism, including standardized data preprocessing process, parameter setting specification, precision evaluation standard, etc., to develop user-friendly software tools and data service interfaces, and to promote the wide application and industrialization of the technology.

[0093] The above description shows and describes several preferred embodiments of the present application, but as mentioned previously, it is to be understood that the application is not limited to the disclosed forms, and should not be considered as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the inventive concept described in the specification, by the above teaching or related technical or knowledge. Any modification and change made by those skilled in the art without departing from the spirit and scope of the present application shall be within the protection scope of the appended claims of the present application.

Claims

1. A method for identifying tea gardens based on multi-source remote sensing data, characterized in that, The method includes the following steps: S1: Acquire temporal radar data and optical remote sensing data of the target area; S2: Extract key growth periods of tea gardens based on radar data, including the beginning and end of the growing season; S3: Calculate vegetation index (VI) and water index (WI) using optical data, and select the VI and WI combination with the best discrimination by time-weighted dynamic time warping method. S4: Determine the tea garden identification index as follows: Where Ω1 is a linear operation, Ω2 is a nonlinear operation, T represents the tea garden sample, G represents the evergreen vegetation sample, and DTW is the dynamic time warping method. S5: Tea garden identification is achieved based on the tea garden identification index and ensemble learning model.

2. The tea garden identification method based on multi-source remote sensing data as described in claim 1, characterized in that, Key growth stages in tea gardens, extracted based on radar data, include: a) Perform radiometric calibration, geometric correction and filtering on radar data to construct a time series of VV / VH polarization backscattering coefficients; b) For the VV / VH polarization backscattering coefficient time series, filter smoothing is performed. The formula for calculating the filter smoothing is: ， in, These are the values ​​after filtering and smoothing. For the original value, The convolution coefficients are... The normalization coefficient is... is the half-width of the filtering window, and i is the offset relative to the center j of the sliding window, with a value ranging from -m to m; The start-of-season SOS and end-of-season EOS are determined based on the extreme points of the first derivative, as follows: ， , The argmax function is used to find the maximum value of a function within a given range. This is the change in the backscattering coefficient. This represents the corresponding time interval.

3. The tea garden identification method based on multi-source remote sensing data as described in claim 1, characterized in that, The optimal VI and WI combinations, selected using the time-weighted dynamic time warping method, include: Building a tea garden timeline With evergreen vegetation time sequence The weighted distance matrix is: ,in Point and Weighted distance between them; Calculate the elements of the cumulative distance matrix: ， Construct the cumulative distance matrix ,in This represents the distance from the start of the sequence to the current position. The minimum cumulative distance; The output time-weighted dynamic time warp distance is: .

4. The tea garden identification method based on multi-source remote sensing data as described in claim 1, characterized in that, The vegetation indices include at least one of the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Ratio Vegetation Index (RVI), and the water indices include at least one of the Land Water Index (LSWI), Normalized Difference Water Index (NDWI), and Modified Normalized Difference Water Index (mNDWI).

5. The tea garden identification method based on multi-source remote sensing data according to claim 4, characterized in that, It also includes a pre-extraction step for evergreen vegetation: a) For each of the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Ratio Vegetation Index (RVI), Land Water Index (LSWI), Normalized Water Index (NDWI), and Modified Normalized Difference Water Index (mNDWI), four time-series statistical features are extracted: annual mean, maximum value, minimum value, and annual variation, and a 24-dimensional multi-temporal feature space is constructed. b) Combine random forest, support vector machine and extreme gradient boosting methods to classify land cover in the target area and extract evergreen vegetation areas; c) Morphological processing methods, including opening and closing operations and region connectivity analysis, are used to remove isolated misclassified pixels and enhance the spatial continuity of the results.

6. The tea garden identification method based on multi-source remote sensing data according to claim 1, characterized in that, Using ensemble learning models to identify tea gardens includes: Integrating the probability outputs of three classifiers: Random Forest, Support Vector Machine, and Extreme Gradient Boosting: ， in, It is the final classification probability. is the probability prediction of the i-th classifier, and M is the number of classifiers; A feature space is constructed by combining topographic parameters such as elevation, slope, and aspect.

7. The tea garden identification method based on multi-source remote sensing data according to claim 1, characterized in that, The radar data is Sentinel-1 C-band synthetic aperture radar (SAR) data, and the optical data is Sentinel-2 multispectral data.

8. A tea garden identification system based on multi-source remote sensing data, characterized in that, It includes a data acquisition module, a phenological analysis module, an index construction module, a classification module, and an output module, among which... The data acquisition module is used to acquire time-series radar data and optical remote sensing data of the target area; The phenology analysis module is used to extract key growth stages in tea gardens based on radar data; The index construction module is used to calculate the vegetation index (VI) and water index (WI) using optical data, and to select the VI and WI combination with the best discrimination through a time-weighted dynamic time warping method; and to determine the tea garden identification index according to the following method: Where Ω1 is a linear operation, Ω2 is a nonlinear operation, T represents the tea garden sample, G represents the evergreen vegetation sample, and DTW is the dynamic time warping method. The classification module identifies tea gardens based on a tea garden identification index and an ensemble learning model. The output module is used to generate a spatial distribution map of tea gardens based on the tea garden identification results.

9. The tea garden identification system based on multi-source remote sensing data according to claim 8, characterized in that, The phenological analysis module extracts key growth stages in tea gardens based on radar data, including: a) Perform radiometric calibration, geometric correction and filtering on radar data to construct a time series of VV / VH polarization backscattering coefficients; b) For the VV / VH polarization backscattering coefficient time series, filter smoothing is performed. The formula for calculating the filter smoothing is: ， in, These are the values ​​after filtering and smoothing. For the original value, The convolution coefficients are... The normalization coefficient is... is the half-width of the filtering window, and i is the offset relative to the center j of the sliding window, with a value ranging from -m to m; The start-of-season SOS and end-of-season EOS are determined based on the extreme points of the first derivative, as follows: ， , The argmax function is used to find the maximum value of a function within a given range. This is the change in the backscattering coefficient. This represents the corresponding time interval.

10. The tea garden identification system based on multi-source remote sensing data according to claim 8, characterized in that, The index construction module is used to calculate the vegetation index (VI) and water index (WI) using optical data, and selects the optimal combination of VI and WI using a time-weighted dynamic time warping method, including: Building a tea garden timeline With evergreen vegetation time sequence The weighted distance matrix is: ,in Point and Weighted distance between them; Calculate the elements of the cumulative distance matrix: ， Construct the cumulative distance matrix ,in This represents the distance from the start of the sequence to the current position. The minimum cumulative distance; The output time-weighted dynamic time warp distance is: .

Images