Crop type recognition method based on dynamic constraint matching of standard growth curve

Abstract

The application provides a crop type recognition method based on dynamic constraint matching of standard growth curves, and relates to the technical field of agricultural remote sensing. The method comprises the following steps: obtaining multi-source remote sensing observation data of a target area, and constructing an irregular multi-source observation sequence of an object to be recognized according to observation quality weights; mapping each observation data from a calendar time axis to a crop growth phase axis according to remote sensing phenological characteristics and meteorological accumulated temperature characteristics, and reconstructing a spatiotemporal continuous growth curve of the object to be recognized on the crop growth phase axis; constructing a standard growth curve library based on sample growth curves of labeled sample objects, wherein the standard growth curve library comprises standard growth curves of each crop type and corresponding dynamic matching thresholds; for each crop type, matching the spatiotemporal continuous growth curve of the object to be recognized with the standard growth curve in a limited matching space defined by the dynamic matching threshold to obtain a dynamic constraint matching distance; and determining the crop type of the object to be recognized according to the dynamic constraint matching distance. The application can improve the accuracy and stability of crop type recognition under the conditions of sparse observation, cloud pollution, obvious differences in sowing period and growth potential.

Description

Technical Field

[0001] This invention relates to the field of agricultural remote sensing technology, and in particular to a crop type identification method based on dynamic constraint matching of standard growth curves. Background Technology

[0002] The spatial distribution of crop types is a crucial foundation for planting structure surveys, agricultural production management, disaster assessment, and yield prediction. With the continuous accumulation of auxiliary data such as optical remote sensing, synthetic aperture radar, meteorological, topographical, and soil data, crop type identification based on remote sensing time series has become an important means of large-scale agricultural monitoring. Existing methods typically construct calendar time series from reflectance, vegetation indices, or radar backscattering, and then use models such as random forests and recurrent neural networks for discrimination. However, actual observations are often affected by clouds, cloud shadows, sensor observation conditions, and radar noise, resulting in uneven effective observation time, numerous missing phases, and misalignment of growth curves of the same crop on the calendar time axis under different regions, years, and sowing dates. Directly comparing based on fixed date characteristics or fixed time series can easily misjudge normal phenological misalignments or differences in growth as differences in crop type. Meanwhile, traditional standard curve matching often relies on fixed empirical thresholds, making it difficult to simultaneously account for reasonable variations within the same crop type and the separable boundaries between different crop types, leading to insufficient classification stability and interpretability. Summary of the Invention

[0003] This invention provides a crop type identification method based on dynamic constraint matching of standard growth curves, which solves the problem of insufficient classification stability of existing crop type identification methods based on dynamic constraint matching of standard growth curves under conditions of irregular multi-source remote sensing observations, significant phenological misalignment, and empirically dependent matching threshold settings.

[0004] This invention provides a crop type identification method based on dynamic constraint matching of standard growth curves, comprising the following steps.

[0005] Acquire multi-source remote sensing observation data of the target area, and construct an irregular multi-source observation sequence of the object to be identified based on the observation quality weight of each observation data. Based on the remote sensing phenological characteristics and meteorological accumulated temperature characteristics of the object to be identified, the observation data in the irregular multi-source observation sequence are mapped from the calendar time axis to the crop growth phase axis, and the spatiotemporal continuous growth curve of the object to be identified is reconstructed on the crop growth phase axis. Obtain the sample growth curves of sample objects labeled with crop type tags, and construct a standard growth curve library based on the sample growth curves. The standard growth curve library includes: standard growth curves for each crop type and dynamic matching thresholds corresponding to each crop type. For each crop type, the spatiotemporal continuous growth curve is matched with the standard growth curve within a restricted matching space defined by the dynamic matching threshold to obtain the dynamic constraint matching distance between the object to be identified and each crop type. The crop type of the object to be identified is determined based on the dynamic constraint matching distance.

[0006] In one possible implementation, the spatiotemporal continuous growth curve of the object to be identified is reconstructed on the crop growth phase axis, including: Feature unification and scale normalization are performed on remote sensing observation data from various sources to obtain multi-source remote sensing observation values; The coefficients of the spline basis functions are determined based on the observation quality weights of each remote sensing observation data, the fitting error of the spline basis functions, the curve smoothness constraints, the crop growth morphology constraints, and the boundary preservation spatial constraints. Based on the coefficients of the spline basis functions, a spatiotemporal continuous growth curve of the object to be identified on the crop growth phase axis is generated; Based on the observation quality, observation density, consistency of multi-source observations, and spline fitting residuals of the object to be identified at each reproductive phase, the curve uncertainty of the spatiotemporal continuous growth curve at each reproductive phase is determined, and the curve uncertainty is the confidence interval of the spatiotemporal continuous growth curve at each reproductive phase.

[0007] In another possible implementation, the standard growth curve library can be constructed in the following ways: Obtain the sample growth curve of sample objects with crop type labels; The growth curves of each sample object are screened for quality and abnormal curves are removed. The growth curves of the same crop type after screening are phase aligned. Based on the phase-aligned sample growth curves of the same crop type, determine the standard growth curve, key phenological nodes, curve uncertainty, and dynamic matching threshold for the crop type. Dynamic matching thresholds include phase offset thresholds and amplitude tolerance thresholds.

[0008] In another possible implementation, the dynamic matching threshold includes a phase offset threshold, which is determined based on a preset high quantile of the phase offset of the sample growth curves of the same crop type at key phenological nodes, and the phase separation distance of the standard growth curves of different crop types at the same key phenological nodes.

[0009] In another possible implementation, the dynamic matching threshold includes an amplitude tolerance threshold, which is determined based on the preset high quantile of the amplitude offset of the sample growth curves of the same crop type at each growth phase, and the amplitude separation distance of the standard growth curves of different crop types at the same growth phase.

[0010] In another possible implementation, the spatiotemporally continuous growth curve is matched with a standard growth curve within a matching space defined by a dynamic matching threshold, including: Based on the standard phases of the key phenological nodes of the object to be identified and the key phenological nodes corresponding to the crop type, the translation amount of the spatiotemporal continuous growth curve in the direction of the crop growth phase is determined, and the translation amount does not exceed the phase offset threshold corresponding to the crop type. The shifted spatiotemporal continuous growth curve is compared with the standard growth curve of the crop type to determine the amplitude deviation at each growth phase. Crop types whose amplitude deviation does not exceed the amplitude tolerance threshold corresponding to the crop type are identified as valid candidate crop types.

[0011] In another possible implementation, the crop type of the object to be identified is determined based on the dynamic constraint matching distance, including: The crop type with the smallest dynamic constraint matching distance is determined as the first candidate crop type, and the crop type with the second smallest dynamic constraint matching distance is determined as the second candidate crop type. When the dynamic constraint matching distance of the first candidate crop type is less than or equal to the matching distance threshold corresponding to the first candidate crop type, and the difference between the dynamic constraint matching distance of the second candidate crop type and the first candidate crop type is greater than the preset category interval threshold, and the effective observation ratio of the object to be identified in the key reproductive phase meets the preset sufficiency condition, the first candidate crop type is determined as the crop type of the object to be identified.

[0012] The present invention also provides a crop type identification device based on dynamic constraint matching of standard growth curves, comprising the following modules: The acquisition module is used to acquire multi-source remote sensing observation data of the target area and construct an irregular multi-source observation sequence of the object to be identified based on the observation quality weight of each observation data. The processing module is used to map the observation data in the irregular multi-source observation sequence from the calendar time axis to the crop growth phase axis based on the remote sensing phenological characteristics and meteorological accumulated temperature characteristics of the object to be identified, and to reconstruct the spatiotemporal continuous growth curve of the object to be identified on the crop growth phase axis. The processing module is also used to obtain the sample growth curves of sample objects labeled with crop type labels, and to construct a standard growth curve library based on the sample growth curves. The standard growth curve library includes: standard growth curves of each crop type and dynamic matching thresholds corresponding to each crop type. The processing module is also used to match the spatiotemporal continuous growth curve with the standard growth curve for each crop type within a limited matching space defined by the dynamic matching threshold, so as to obtain the dynamic constraint matching distance between the object to be identified and each crop type. The processing module is also used to determine the crop type of the object to be identified based on the dynamic constraint matching distance.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-described crop type identification methods based on dynamic constraint matching of standard growth curves.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described crop type identification methods based on dynamic constraint matching of standard growth curves.

[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements any of the above-described crop type identification methods based on dynamic constraint matching of standard growth curves.

[0016] The crop type identification method based on dynamic constraint matching of standard growth curves provided by this invention maps remote sensing observation data from a calendar time axis to a crop growth phase axis, which can reduce the impact of phenological misalignment caused by different regions, years, and sowing dates. By constructing irregular multi-source observation sequences using observation quality weights and reconstructing spatiotemporally continuous growth curves on the growth phase axis, the influence of missing data and noise on subsequent discrimination can be reduced. By constructing a standard growth curve library containing dynamic matching thresholds and performing curve matching within the restricted matching space defined by the dynamic matching thresholds, reasonable differences in sowing date and growth vigor of the same crop can be accommodated simultaneously, while maintaining the separable boundaries between different crop types. Therefore, the classification results are jointly constrained by the consistency of the growth process, the consistency of key phenological nodes, the tolerance of curve amplitude, and the category interval, thereby improving the accuracy, stability, and verifiability of crop type identification. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the present invention, the accompanying drawings used in the embodiments are briefly described below. The following drawings are only used to illustrate some embodiments of the present invention; those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0018] Figure 1 This is one of the flowcharts of the crop type identification method based on dynamic constraint matching of standard growth curves provided by the present invention.

[0019] Figure 2 This is a schematic diagram of the crop type identification device based on dynamic constraint matching of standard growth curves provided by the present invention.

[0020] Figure 3This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be further described below with reference to the accompanying drawings. The described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative effort are all within the protection scope of the present invention.

[0022] The terms “comprising” and “having”, and any variations thereof, used in the description of this invention are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or modules is not limited to the steps or modules listed, but may optionally include other steps or modules not listed, or may optionally include other steps or modules inherent to such processes, methods, products, or apparatus.

[0023] Furthermore, in this invention, expressions such as "for example" are used to describe specific embodiments and should not be construed as limiting the superiority or inferiority of the embodiments or the scope of protection.

[0024] To address the aforementioned issues, this invention transforms remote sensing observations from a calendar timeline to a crop growth phase axis to mitigate the impact of phenological misalignment on classification. It reconstructs quality-weighted spatiotemporally continuous growth curves on the growth phase axis to reduce interference from cloud contamination, missing data, and anomalous observations on curve morphology. Furthermore, it constructs a standard growth curve library incorporating phase shift thresholds and amplitude tolerance thresholds, and calculates dynamically constrained matching distances within a limited matching space defined by dynamic thresholds. This method transforms crop classification from simple time-series similarity comparisons into reliable discrimination of growth processes under category structure constraints.

[0025] The crop type identification method based on dynamic constraint matching of standard growth curves provided by this invention can be executed by a crop type identification device based on dynamic constraint matching of standard growth curves. This device can be deployed in an electronic device or implemented by a processor or crop type identification module within the electronic device. The method of this invention will be described below using the crop type identification device based on dynamic constraint matching of standard growth curves as the executing entity.

[0026] Figure 1 This is one of the flowcharts illustrating the crop type identification method based on dynamic constraint matching of standard growth curves provided by this invention, such as... Figure 1 As shown, the method includes the following steps: Step 101: Obtain multi-source remote sensing observation data of the target area, and construct an irregular multi-source observation sequence of the object to be identified based on the observation quality weight of each observation data.

[0027] Multi-source remote sensing data may include one or more of the following: optical remote sensing data, synthetic aperture radar remote sensing data, meteorological data, topographic data, soil data, and field management auxiliary data.

[0028] For example, optical remote sensing data may include reflectance in the visible, near-infrared, red-edge, and short-wave infrared bands, as well as vegetation indices, red-edge indices, short-wave infrared moisture indices, and canopy structure indices constructed from them; synthetic aperture radar remote sensing data may include VV polarization, VH polarization, polarization ratio, polarization difference, coherence, or texture features; meteorological data may include air temperature, precipitation, radiation, accumulated temperature, and evapotranspiration demand; topographic and soil data may include elevation, slope, aspect, soil texture, soil organic matter, and field water holding capacity; and field management auxiliary data may include planting systems, irrigation conditions, plot boundaries, historical planting records, or information reported by farmers.

[0029] The object to be identified can be a pixel, field, plot, grid cell, or other spatial unit that requires crop type identification; the target area can be a county, city, province, farm, irrigation district, agricultural park, or other agricultural planting area.

[0030] The observation quality weight is used to characterize the reliability of a single observation in subsequent curve reconstruction and matching. The higher the weight, the more reliable the observation; the lower the weight, the more significantly the observation is affected by clouds, noise, missing data, or anomalous conditions.

[0031] In some embodiments, the observation quality weight for each observation can be determined based on at least one of cloud probability, cloud shadow identification, aerosol quality identification, sensor quality identification, radar incident angle, rainfall impact, observation angle, signal-to-noise ratio, spatial neighborhood consistency, and historical curve consistency.

[0032] The irregular multi-source observation sequence constructed in this way can be represented as: in, Indicates the object to be identified Irregular multi-source observation sequences, Indicates the first Calendar time of the observation Indicates the source of observation or sensor type. Represents the observed value or observed feature vector. Indicates the observation quality weight. Indicates the effectiveness of the observation mask. Indicates auxiliary variables related to weather, topography, soil, or field management. Indicates the object to be identified The number of observations.

[0033] Step 102: Based on the remote sensing phenological characteristics and meteorological accumulated temperature characteristics of the object to be identified, map each observation data in the irregular multi-source observation sequence from the calendar time axis to the crop growth phase axis, and reconstruct the spatiotemporal continuous growth curve of the object to be identified on the crop growth phase axis.

[0034] Crop growth phases are used to characterize the relative growth process of a crop from the start of growth to maturity. The sowing date, emergence date, rapid growth period, peak growth period, senescence period, and maturity period of the same crop can vary significantly in different regions, years, varieties, and management conditions. Directly comparing crop curves on a calendar timeline can easily misinterpret normal phenological differences as differences in crop type. Therefore, this invention transforms irregular multi-source observation sequences from a calendar timeline to a crop growth phase axis, changing curve comparisons from ordinary time series comparisons to comparisons of crop growth processes.

[0035] In some embodiments, key phenological nodes such as the growth initiation period, peak period, and maturity period can be extracted based on remote sensing time-series curves, and phase normalization can be performed on each observation time in combination with the cumulative growth days.

[0036] In some embodiments, the value range of the crop growth phase can be set to 0 to 1, or it can be divided into multiple discrete phases according to the crop growth stage. The crop growth phase can be determined by a piecewise mapping method: in, Indicates the initial stage of growth. Indicates the peak period. Indicates maturity. Indicates the time from the start of growth to The cumulative growth of the days, This indicates the position of the peak period on the reproductive phase axis. By using cumulative growth days instead of simple calendar dates, the impact of regional temperature differences on the expression of phenological processes can be reduced.

[0037] The specific process of reconstructing the spatiotemporal continuous growth curve of the object to be identified on the crop growth phase axis includes: First, feature unification and scale normalization are performed on remote sensing observation data from various sources to obtain multi-source remote sensing observation values.

[0038] Secondly, the coefficients of the spline basis functions are determined based on the observation quality weights of each remote sensing observation data, the fitting error of the spline basis functions, the curve smoothness constraints, the crop growth morphology constraints, and the boundary preservation spatial constraints.

[0039] The spline coefficients are obtained through quality-weighted constraint optimization. This optimization process simultaneously considers the fitting error of multi-source observations, curve smoothness, crop growth morphology constraints, and boundary-preserving spatial constraints, so that the reconstructed curves can closely resemble high-quality remote sensing observations while avoiding curve breaks and unreasonable fluctuations caused by cloud pollution, sparse observations, or abnormal observations.

[0040] In one possible implementation, spline coefficients Determined by the following objective function: in, Let represent the optimal spline coefficients of the object p to be identified, Lobs represent the multi-source observation quality-weighted fitting error, Lsmooth represent the curve smoothing constraint, Rshape represent the crop growth morphology constraint, Lspace represent the boundary-preserving spatial constraint, and η1, η2, and η3 represent the weights of the curve smoothing constraint, crop growth morphology constraint, and boundary-preserving spatial constraint, respectively. For example, the above objective function can be solved using the cubic B-spline basis function fitting method to obtain the coefficients of the spline basis function.

[0041] Next, based on the coefficients of the spline basis functions, a spatiotemporal continuous growth curve of the object to be identified on the crop growth phase axis is generated. This spatiotemporal continuous growth curve can be a univariate curve or a multivariate curve, and the curve variables include at least one of the following: vegetation index, red edge index, shortwave infrared moisture index, radar polarization characteristics, or comprehensive growth state. The spatiotemporal continuous growth curve of the object to be identified can be expressed as: in, Indicates the object to be identified In the phase of fertility The spatiotemporal continuous growth curve or growth state vector on the surface. Indicates the first spline basis functions Indicates the object to be identified Corresponding to the The coefficient vector of each spline basis function This indicates the number of spline basis functions.

[0042] Finally, based on the observation quality, observation density, multi-source observation consistency, and spline fitting residuals of the target object at each reproductive phase, the curve uncertainty of the spatiotemporal continuous growth curve at each reproductive phase is determined. The curve uncertainty is the confidence interval of the spatiotemporal continuous growth curve at each reproductive phase.

[0043] Specifically, for fertility phases with dense and high-quality effective observations, the curve uncertainty is low; while for fertility phases with severe cloud contamination, sparse observations, significant conflicts between multi-source observations, or those primarily reliant on curve completion, the curve uncertainty is high. Curve uncertainty is further used in subsequent dynamic constraint matching distance calculations to reduce the excessive influence of low-quality or high-uncertainty phases on crop classification results.

[0044] Step 103: Obtain the sample growth curves of the sample objects labeled with crop type tags, and construct a standard growth curve library based on the sample growth curves.

[0045] The standard growth curve library includes standard growth curves for each crop type and corresponding dynamic matching thresholds. The dynamic matching thresholds include phase offset thresholds and amplitude tolerance thresholds. Samples can be obtained from historical crop distribution maps, field surveys, agricultural statistical auxiliary data, farmer-reported data, high-confidence remote sensing samples, or multi-source cross-validation samples.

[0046] The specific methods for constructing a standard growth curve library include: First, obtain the sample growth curves of sample objects labeled with crop type. For each sample object, convert its multi-source remote sensing observation sequence to the crop growth phase axis to obtain the sample growth curve.

[0047] Secondly, the growth curves of each sample object are subjected to quality screening and abnormal curve removal, and the growth curves of the same crop type after screening are phase aligned. In one possible implementation, the sample curves are quality screened to remove abnormal curves, mixed pixel curves, low-quality observation curves, abnormal curves obviously affected by disasters, or curves inconsistent with the normal growth process of the crop.

[0048] Finally, based on the sample growth curves of the same crop type after phase alignment, the standard growth curve, key phenological nodes, curve uncertainty, and dynamic matching threshold of the crop type are determined.

[0049] For example, the standard growth curve for a crop type can be estimated from the weighted mean, robust mean, median curve, or probability model of similar high-confidence sample curves. The weights can be determined by sample quality, observational completeness, field reliability, spatial purity, and historical consistency.

[0050] The phase shift threshold is determined based on the preset high quantile of the phase shift of sample growth curves of the same crop type at key phenological nodes, and the phase separation distance of standard growth curves of different crop types at the same key phenological nodes. For crop types... The For each key phenological node, the phase offset threshold satisfies: in, Indicate crop type In the Permissible phase offset range at key phenological nodes Indicates the first Crop type The sample in the Fertility phases at key phenological nodes Indicate crop type Standard key phenological node phases High quantiles representing sample phase shift (e.g.) Desirable That is, take quantiles Represents the inter-class separation constraint coefficient (e.g., can be taken as...) to The values ​​between these ranges are used to limit the phase offset threshold from crossing the separable boundaries of adjacent crop types. Indicates different Other crop types. This threshold is not a fixed empirical value, but rather varies dynamically with crop type, key phenological nodes, and regional sample structure.

[0051] The amplitude tolerance threshold is determined based on the preset high quantile of the amplitude offset of the sample growth curves of the same crop type at each growth phase, and the amplitude separation distance of the standard growth curves of different crop types at the same growth phase. For crop type... In the phase of fertility The amplitude tolerance threshold on the above satisfies: in, Indicate crop type In the phase of fertility The allowable amplitude offset range, Indicates the first Crop type The growth curve of the sample. Indicate crop type Standard growth curve, Standard growth curves representing other crop types, This represents the distance metric between the eigenvectors of a curve. High quantiles representing amplitude offsets of similar samples (e.g.) Desirable ), This represents the inter-class separation constraint coefficient. The resulting amplitude tolerance threshold is not a fixed constant throughout the season, but rather a dynamic tolerance curve that varies with the fertility phase.

[0052] Step 104: For each crop type, match the spatiotemporal continuous growth curve with the standard growth curve within the restricted matching space defined by the dynamic matching threshold to obtain the dynamic constraint matching distance between the object to be identified and each crop type.

[0053] The matching process does not involve unconstrained stretching or arbitrary translation of the curve. Instead, it seeks a reasonable phase alignment within a phase offset threshold range and calculates the curve deviation within a standard curve confidence tube defined by an amplitude tolerance threshold. Specifically, this includes: First, based on the standard phases of the key phenological nodes of the object to be identified and the key phenological nodes corresponding to the crop type, the translation amount of the spatiotemporal continuous growth curve in the crop growth phase direction is determined, and the translation amount does not exceed the phase offset threshold corresponding to the crop type.

[0054] In some embodiments, for candidate crop types, the phase offset of the object to be identified at the first key phenological node needs to satisfy: in, Indicates the object to be identified The A key phenological node phase, This indicates the phase shift corresponding to the key phenological node. Indicate crop type Standard key phenological node phases Indicate crop type In the Phase offset thresholds at key phenological nodes.

[0055] Secondly, the phase-adjusted spatiotemporal continuous growth curve is compared with the standard growth curve of the crop type, and the amplitude deviation at each growth phase is calculated.

[0056] Then, crop types whose amplitude deviation does not exceed the amplitude tolerance threshold corresponding to their crop type are identified as valid candidate crop types. That is, valid candidate categories must simultaneously satisfy both phase and amplitude constraints. in, Indicate crop type In the phase of fertility The amplitude tolerance threshold is applied. A crop type is considered a valid candidate category only if the offset of the key phenological node of the object to be identified does not exceed the phase offset threshold for the corresponding crop type, and its curve amplitude deviation falls within the confidence tube of the standard growth curve defined by the amplitude tolerance threshold at the corresponding growth phase.

[0057] For candidate categories that still meet the dynamic threshold constraints, calculate the dynamic constraint matching distance. The smaller the matching distance, the more consistent the curve to be identified is with the standard growth curve of that crop type. The dynamic constraint matching distance can be expressed as: in, Indicates the object to be identified With crop type Dynamic constraint matching distance, Represents the phase shift function. Indicates crop type The set of permissible translations limited by the phase offset threshold. This represents the set of reproductive phases that participated in the matching. This indicates the weighting of the reproductive phase (e.g., higher weights can be assigned to the rapid growth phase and peak growth phase). Indicate crop type In the phase of fertility The amplitude tolerance threshold on the above This indicates the uncertainty of the curve of the object to be identified in that phase. This indicates a numerically stable term that avoids a denominator of zero (e.g., a value that can be taken as zero). ).

[0058] The dynamic constraint matching distance has the following functions: First, it allows the translation set to limit the phase adjustment range, avoiding excessive translation or distortion of the curve, so that the matching results still retain the meaning of the crop growth process; Second, the amplitude tolerance threshold allows the matching tolerance of different growth phases to be dynamically determined by the standard growth curve library, rather than being set manually; Third, the curve uncertainty automatically reduces the impact of low-quality or high-uncertainty growth phases, avoiding excessive influence of cloud pollution, sparse observations, or completion errors on the classification results.

[0059] Step 105: Determine the crop type of the object to be identified based on the dynamic constraint matching distance.

[0060] Specifically, the crop type with the smallest dynamic constraint matching distance is determined as the first candidate crop type, and the crop type with the second smallest dynamic constraint matching distance is determined as the second candidate crop type.

[0061] For example, let the first candidate crop type be c1 and the second candidate crop type be c2.

[0062] When the dynamic constraint matching distance of the first candidate crop type is less than or equal to the matching distance threshold corresponding to the first candidate crop type, and the difference between the dynamic constraint matching distance of the second candidate crop type and the first candidate crop type is greater than the preset category interval threshold, and the effective observation ratio of the object to be identified in the key reproductive phase meets the matching sufficiency condition, the first candidate crop type is determined as the crop type of the object to be identified.

[0063] In some embodiments, the matching distance threshold can be determined based on the distribution of matching distances dynamically constrained among similar samples, for example, by taking its high quantile; the category interval threshold can be determined based on the distance distribution between different crop types and operational accuracy requirements. Matching sufficiency conditions may include: key growth phases having no less than a preset number of valid observations, at least one of the rapid growth period, peak period, or senescence period having high-quality observations, the proportion of curve completion not exceeding a preset ratio, or the consistency between multi-source observations reaching a preset level.

[0064] The first candidate crop type will be determined as the crop type of the object to be identified when the following conditions are met: in, Indicates the object to be identified The classification results This represents the dynamic constraint matching distance for the first candidate crop type. This represents the dynamic constraint matching distance for the second candidate crop type. This represents the matching distance threshold for the first candidate crop type. Indicates the category interval threshold. This indicates the proportion of valid observations or a sufficiency index of the subjects to be identified at key reproductive phases. This represents the threshold for matching sufficiency.

[0065] If the above conditions are not met, a set of candidate crop types is output or a check tag is generated. This processing method avoids forcibly outputting a single category when information is insufficient or the category interval is insufficient, thus improving the reliability of the classification results.

[0066] The final output may include crop classification map, candidate category map, matching confidence map, dynamic constraint matching distance map, category interval map, effective observation sufficiency map, or marker map to be verified.

[0067] Based on the above technical solutions, this invention maps remote sensing observation data from a calendar time axis to a crop growth phase axis, which can mitigate the impact of phenological misalignment caused by different regions, years, and sowing dates. By constructing irregular multi-source observation sequences using observation quality weights and reconstructing spatiotemporally continuous growth curves on the growth phase axis, it can reduce the impact of missing data and noise on curve morphology under conditions of cloud pollution and sparse observation. By constructing a standard growth curve library containing dynamic matching thresholds and performing curve matching within a limited matching space defined by the dynamic matching thresholds, it can simultaneously accommodate reasonable differences in sowing dates and growth patterns of the same crop while maintaining separable boundaries between different crop types. Therefore, the classification results are jointly constrained by the consistency of the growth process, the consistency of key phenological nodes, the tolerance of curve amplitude, the category interval, and the sufficiency of matching, thereby improving the accuracy, stability, and verifiability of crop type identification.

[0068] The crop type identification device based on dynamic constraint matching of standard growth curves provided by this invention will be described below. This device embodiment and the above method embodiment are based on the same inventive concept, and related technical features can be referred to each other. To achieve the above functions, the electronic device may include corresponding hardware structures and / or software modules; the specific implementation method—hardware, software, or a combination of hardware and software—can be determined according to the application scenario and design requirements.

[0069] Figure 2 This is a schematic diagram of the crop type identification device based on dynamic constraint matching of standard growth curves provided by the present invention, as shown in the figure. Figure 2 As shown, the crop type identification device based on dynamic constraint matching of standard growth curves includes the following modules: acquisition module 201 and processing module 202.

[0070] The acquisition module 201 is used to acquire multi-source remote sensing observation data of the target area and construct an irregular multi-source observation sequence of the object to be identified based on the observation quality weight of each observation data. Processing module 202 is used to map each observation data in the irregular multi-source observation sequence from the calendar time axis to the crop growth phase axis according to the remote sensing phenological characteristics and meteorological accumulated temperature characteristics of the object to be identified, and to reconstruct the spatiotemporal continuous growth curve of the object to be identified on the crop growth phase axis. The processing module 202 is also used to obtain the sample growth curve of the sample object labeled with crop type, and to construct a standard growth curve library based on the sample growth curve. The standard growth curve library includes: the standard growth curve of each crop type and the dynamic matching threshold corresponding to each crop type. The processing module 202 is also used to match the spatiotemporal continuous growth curve with the standard growth curve for each crop type within a limited matching space defined by the dynamic matching threshold, so as to obtain the dynamic constraint matching distance between the object to be identified and each crop type. The processing module 202 is also used to determine the crop type of the object to be identified based on the dynamic constraint matching distance.

[0071] Figure 3 An example of the physical structure of an electronic device is given. For example... Figure 3 As shown, the electronic device may include a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, communications interface 320, and memory 330 communicate with each other through the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute the above-described crop type identification method based on dynamic constraint matching of standard growth curves. The method includes: acquiring multi-source remote sensing observation data of the target area and constructing an irregular multi-source observation sequence of the object to be identified according to the observation quality weight; mapping the irregular multi-source observation sequence to the crop growth phase axis according to the remote sensing phenological characteristics and meteorological accumulated temperature characteristics, and reconstructing a spatiotemporal continuous growth curve; constructing a standard growth curve library and calculating the dynamic constraint matching distance within the restricted matching space defined by the dynamic matching threshold; and determining the crop type of the object to be identified according to the dynamic constraint matching distance, category interval, and matching sufficiency conditions.

[0072] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the part of the technical solution of the present invention that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device to execute all or part of the steps of the methods described in the embodiments of the present invention. The storage medium includes media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0073] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the crop type identification method based on dynamic constraint matching of standard growth curves provided by the above methods. This method includes: acquiring multi-source remote sensing observation data of a target area, and constructing an irregular multi-source observation sequence of the object to be identified according to the observation quality weights of each observation data point; mapping each observation data point in the irregular multi-source observation sequence from a calendar time axis to a crop growth phase axis according to the remote sensing phenological characteristics and meteorological accumulated temperature characteristics of the object to be identified, and reconstructing the spatiotemporal continuous growth curve of the object to be identified on the crop growth phase axis; acquiring sample growth curves of sample objects labeled with crop type tags, and constructing a standard growth curve library based on the sample growth curves. The standard growth curve library includes: standard growth curves for each crop type and dynamic matching thresholds corresponding to each crop type; and matching the spatiotemporal continuous growth curve with the standard growth curve within a matching space defined by the dynamic matching thresholds for each crop type to obtain the dynamic constraint matching distance between the object to be identified and each crop type. The crop type of the object to be identified is determined based on the dynamic constraint matching distance.

[0074] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements any of the above-described crop type identification methods based on dynamic constraint matching of standard growth curves.

[0075] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; the components shown as units may or may not be physical units, and may be located in one location or distributed across multiple network units. Some or all modules can be selected to implement the solution of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0076] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software and necessary general-purpose hardware platforms, or it can be implemented using hardware. Based on this understanding, the parts of the above technical solutions that contribute to the prior art can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium and includes several instructions for causing a computer device to execute the methods described in each embodiment or parts thereof.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some technical features; these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A crop type identification method based on dynamic constraint matching of standard growth curves, characterized in that, include: Acquire multi-source remote sensing observation data of the target area, and construct an irregular multi-source observation sequence of the object to be identified based on the observation quality weight of each observation data. Based on the remote sensing phenological characteristics and meteorological accumulated temperature characteristics of the object to be identified, the observation data in the irregular multi-source observation sequence are mapped from the calendar time axis to the crop growth phase axis, and the spatiotemporal continuous growth curve of the object to be identified is reconstructed on the crop growth phase axis. Obtain the sample growth curves of sample objects labeled with crop type tags, and construct a standard growth curve library based on the sample growth curves. The standard growth curve library includes: standard growth curves for each crop type and dynamic matching thresholds corresponding to each crop type. For each crop type, the spatiotemporal continuous growth curve is matched with the standard growth curve within the restricted matching space defined by the dynamic matching threshold to obtain the dynamic constraint matching distance between the object to be identified and each crop type. The crop type of the object to be identified is determined based on the dynamic constraint matching distance.

2. The method according to claim 1, characterized in that, Reconstructing the spatiotemporal continuous growth curve of the object to be identified on the crop growth phase axis includes: Feature unification and scale normalization are performed on remote sensing observation data from various sources to obtain multi-source remote sensing observation values; The coefficients of the spline basis functions are determined based on the observation quality weights of each remote sensing observation data, the fitting error of the spline basis functions, the curve smoothness constraints, the crop growth morphology constraints, and the boundary preservation spatial constraints. Based on the coefficients of the spline basis functions, a spatiotemporal continuous growth curve of the object to be identified on the crop growth phase axis is generated; The spatiotemporal continuous growth curve is a univariate curve or a multivariate curve, and the curve variables include at least one of vegetation index, red edge index, shortwave infrared moisture index, radar polarization characteristics, or comprehensive growth status. Based on the observation quality, observation density, multi-source observation consistency, and spline fitting residual of the object to be identified at each reproductive phase, the curve uncertainty of the spatiotemporal continuous growth curve at each reproductive phase is determined, and the curve uncertainty is the confidence interval of the spatiotemporal continuous growth curve at each reproductive phase.

3. The method according to claim 1, characterized in that, The construction methods of the standard growth curve library include: Obtain the sample growth curve of sample objects with crop type labels; The growth curves of each sample object are screened for quality and abnormal curves are removed. The growth curves of the same crop type after screening are phase aligned. Based on the phase-aligned sample growth curves of the same crop type, determine the standard growth curve, key phenological nodes, curve uncertainty, and dynamic matching threshold for that crop type. The dynamic matching threshold includes a phase offset threshold and an amplitude tolerance threshold.

4. The method according to claim 1 or 3, characterized in that, The dynamic matching threshold includes a phase offset threshold, which is determined based on the preset high quantile of the phase offset of the sample growth curve of the same crop type at the key phenological node, and the phase separation distance of the standard growth curves of different crop types at the same key phenological node.

5. The method according to claim 1 or 3, characterized in that, The dynamic matching threshold includes an amplitude tolerance threshold, which is determined based on the preset high quantile of the amplitude offset of the sample growth curve of the same crop type at each growth phase, and the amplitude separation distance of the standard growth curves of different crop types at the same growth phase.

6. The method according to claim 1, characterized in that, Matching the spatiotemporally continuous growth curve with the standard growth curve within a limited matching space defined by the dynamic matching threshold includes: Based on the standard phases of the key phenological nodes of the object to be identified and the key phenological nodes corresponding to the crop type, the translation amount of the spatiotemporal continuous growth curve in the crop growth phase direction is determined, and the translation amount does not exceed the phase offset threshold corresponding to the crop type. The translated spatiotemporal continuous growth curve is compared with the standard growth curve of the crop type to determine the amplitude deviation at each growth phase. Crop types whose amplitude deviation does not exceed the amplitude tolerance threshold corresponding to the crop type are identified as valid candidate crop types.

7. The method according to claim 1, characterized in that, Determining the crop type of the object to be identified based on the dynamic constraint matching distance includes: The crop type with the smallest dynamic constraint matching distance is determined as the first candidate crop type, and the crop type with the second smallest dynamic constraint matching distance is determined as the second candidate crop type. When the dynamic constraint matching distance of the first candidate crop type is less than or equal to the matching distance threshold corresponding to the first candidate crop type, and the difference between the dynamic constraint matching distance of the second candidate crop type and the first candidate crop type is greater than the preset category interval threshold, and the effective observation ratio of the object to be identified in the key reproductive phase meets the preset sufficiency condition, the first candidate crop type is determined as the crop type of the object to be identified.

8. A crop type identification device based on dynamic constraint matching of standard growth curves, characterized in that, The device includes: The acquisition module is used to acquire multi-source remote sensing observation data of the target area and construct an irregular multi-source observation sequence of the object to be identified based on the observation quality weight of each observation data. The processing module is used to map each observation data in the irregular multi-source observation sequence from the calendar time axis to the crop growth phase axis according to the remote sensing phenological characteristics and meteorological accumulated temperature characteristics of the object to be identified, and to reconstruct the spatiotemporal continuous growth curve of the object to be identified on the crop growth phase axis; The processing module is also used to obtain the sample growth curve of the sample object labeled with crop type, and to construct a standard growth curve library based on the sample growth curve. The standard growth curve library includes: standard growth curves of each crop type and dynamic matching thresholds corresponding to each crop type. The processing module is further configured to match the spatiotemporal continuous growth curve with the standard growth curve for each crop type within a restricted matching space defined by the dynamic matching threshold, so as to obtain the dynamic constraint matching distance between the object to be identified and each crop type. The processing module is also used to determine the crop type of the object to be identified based on the dynamic constraint matching distance.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the crop type identification method based on dynamic constraint matching of standard growth curves as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the crop type identification method based on dynamic constraint matching of standard growth curves as described in any one of claims 1 to 7.

Images