Vegetable planting area monitoring system and method for wide-range cultivated land

Abstract

The invention belongs to the technical field of remote sensing monitoring, and particularly relates to a vegetable planting area monitoring system and method for wide-range cultivated land. Comprising the following steps: S1, acquiring multispectral data of each crop in a target area and sample points used for describing crop characteristics; s2, selecting a characteristic phenological time window based on a vegetable phenological period, calculating spectral indexes of pixels corresponding to sample points in the characteristic phenological time window, and taking a set of the spectral indexes of the pixels corresponding to the sample points as spectral index characteristics of the target area; s3, inputting the spectral index features of the target area in different years as training samples into a random forest classifier for training to obtain a trained random forest classification model; and S4, inputting the spectral index characteristics of the to-be-detected target area into the random forest classification model for processing to obtain a vegetable map in the to-be-monitored target area. The method solves the problem that spectral information of vegetables and other crops is similar and cannot be effectively distinguished.

Description

Technical Field

[0001] The present invention belongs to the technical field of remote sensing monitoring, and in particular relates to a vegetable planting area monitoring system and method for large-scale cultivated land. Background Art

[0002] Existing remote sensing monitoring methods primarily rely on low- to medium-resolution satellite imagery (such as Landsat or MODIS) combined with basic spectral classification algorithms, as well as high-resolution drone imagery and deep learning algorithms. However, these algorithms have numerous shortcomings in monitoring vegetable-growing areas: low- and medium-resolution imagery cannot capture the details of small, dispersed vegetable plots, resulting in low classification accuracy, meaning there are significant errors in distinguishing the spectral characteristics of vegetables from other crops (such as rice paddies); drone imagery has limited coverage, is expensive, and is complex to operate; and traditional methods make insufficient use of temporal information, making it impossible to effectively distinguish target crops from environmental disturbances, resulting in a lack of dynamic monitoring capabilities. Specific manifestations are as follows: ① Low classification accuracy: Medium and low-resolution images are difficult to capture the details of scattered and small vegetable planting areas, and traditional classification algorithms are unable to distinguish the spectral characteristics of vegetables from other crops (such as rice fields), resulting in a high error rate.

[0003] ② Poor dynamic monitoring capabilities: Existing remote sensing methods do not make sufficient use of time series information, making it difficult to track changes in planting areas over multiple years and unable to meet the dynamic monitoring needs of vegetable planting management.

[0004] ③ High cost and limited coverage: Although the images collected by drones have high resolution, they are difficult to promote and apply in large areas due to factors such as flight time, coverage and operating costs. Summary of the Invention

[0005] In view of this, the present invention aims to provide a vegetable planting area monitoring system and method for large-scale cultivated land, so as to solve the problem that existing technologies have difficulty in achieving high-precision, low-cost classification and dynamic monitoring in large-scale and diversified vegetable planting areas using remote sensing data, especially when dealing with dispersed small-area vegetable planting areas and distinguishing target crops from other landforms (such as rice fields and weeds). The present invention solves the technical bottlenecks of existing technologies in classification accuracy, dynamic monitoring capabilities and cost efficiency by combining high-resolution remote sensing imagery, phenological characteristics and optimized machine learning algorithms. The present invention solves the problem that the spectral information of vegetables and other crops is similar and cannot be effectively distinguished.

[0006] To achieve the above object, the technical solution created by the present invention is implemented as follows: A method for monitoring vegetable planting areas on large-scale cultivated land comprises the following steps: S1: Obtain multispectral data of each crop in the target area and sample points for describing crop characteristics; S2: Select a characteristic phenological time window based on the vegetable phenological period, calculate the spectral index of the pixel corresponding to each sample point in the characteristic phenological time window, and use the set of spectral indices of the pixel corresponding to each sample point as the spectral index feature of the target area; S3: Repeat steps S1-S2 to obtain the spectral index features of the target area in different years, and input the spectral index features of the target area in different years as training samples into the random forest classifier that integrates multi-temporal data for training to obtain a trained random forest classification model; S4: Repeat steps S1-S2, obtain the spectral index characteristics of the target area to be detected based on the satellite image of the target area to be detected, input the spectral index characteristics of the target area to be detected into the random forest classification model for processing, and obtain a vegetable map within the target area to be monitored.

[0007] Furthermore, in step S1 , the multispectral data includes spectral information of a red light band, a green light band, a blue light band, a near infrared band, and a short-wave infrared band.

[0008] Furthermore, in step S1, satellite images of the target area are collected, and the location information of the target area is recorded using the global positioning system; based on the location information of the target area and the image features contained in the satellite images of the target area, sample points of different crops are selected on Google Earth, and the sample points are used to describe the crop characteristics, and the spectral information of the sample points contained in each crop is extracted through Arcgis software as the multispectral data of each crop.

[0009] Furthermore, in step S2, the spectral index includes green normalized difference vegetation index, enhanced vegetation index, ratio vegetation index, modified normalized difference moisture index, normalized difference accumulation index and surface water index, wherein, Green Normalized Difference Vegetation Index The calculation formula is: ; in, The green band represents the Sentinel-2 satellite image. Represents the near-infrared band of Sentinel-2 satellite imagery; Enhanced Vegetation Index The calculation formula is: ; in, The red band represents the Sentinel-2 satellite image. The blue band represents the Sentinel-2 satellite image; Ratio Vegetation Index The calculation formula is: ; Modified Normalized Difference Water Index The calculation formula is: ; in, Represents the shortwave infrared band of Sentinel-2 satellite imagery; Normalized Difference Stacking Index The calculation formula is: ; Surface Water Index for: .

[0010] Furthermore, step S3 specifically includes the following steps: S31: In the target area, the time range corresponding to the vigorous stage of vegetable phenology is used as the characteristic phenological time window; S32: Within the characteristic phenological time window, based on the value, Value and value, and divide the target area into vegetation area and non-vegetation area; will satisfy both The value is between 0.2 and 0.7. The value is between 0.2 and 1.0. The area where the pixel value is greater than 0.2 is located is considered as the vegetation area, which will satisfy both Values below 0.2, Values below 0.2, The area where the pixels with a value less than 0.2 are located is considered as non-vegetation area; S33: In the vegetation area, based on the value, Value and value, and divide the vegetation area into vegetable growing area and non-vegetable growing area; will satisfy both The value is between 0.1 and 0.4. The value is between 0.5 and 0.8. The area where the pixels with values between 0.2 and 0.6 are located is regarded as the vegetable planting area, and the pixel set other than the vegetable planting area is regarded as the non-vegetable planting area.

[0011] Furthermore, in step S3, the classification probability formula of the random forest classification model is: ; in, is the probability that data point x belongs to category c, n is the number of trees in the random forest model, is the prediction result of the i-th tree for the data point x, I is the indicator function, and n is the total number of decision trees in the random forest classification model.

[0012] Furthermore, the formula for evaluating the accuracy of the random forest classification model is: ; ; Among them, OA is the ratio of the number of correctly classified samples to the total number of samples in the mapping area, Kappa is the Kappa coefficient, TP is the number of correctly identified positive samples, FP is the number of negative samples that are incorrectly marked as positive samples, TN is the number of correctly identified negative samples, and FN is the number of unrecognized positive samples. is the ratio of the sum of the number of correctly classified samples in each category to the total number of samples, that is, the overall accuracy rate, is the expected probability of agreement under random classification.

[0013] A vegetable planting area monitoring system for large-scale cultivated land, comprising: Data acquisition module: acquires multispectral data of each crop in the target area and sample points used to describe crop characteristics; Spectral index calculation module: selects a characteristic phenological time window based on the vegetable phenological period, calculates the spectral index of the pixel corresponding to each sample point in the characteristic phenological time window, and uses the set of spectral indices of the pixel corresponding to each sample point as the spectral index feature of the target area; The classifier training module obtains the spectral index characteristics of the target area in different years and inputs the spectral index characteristics of the target area in different years as training samples into the random forest classifier that integrates multi-temporal data for training to obtain a trained random forest classification model; The monitoring module obtains the spectral index characteristics of the target area to be detected based on the satellite image of the target area to be detected, inputs the spectral index characteristics of the target area to be detected into the random forest classification model for processing, and obtains the vegetable map within the target area to be monitored.

[0014] Compared with the prior art, the present invention can achieve the following beneficial effects: (1) The present invention creates a vegetable planting area monitoring system and method for large-scale cultivated land, and proposes a crop classification method based on phenological characteristics. By extracting VPC (Vegetable Phenological Characteristics) and analyzing the phenological characteristics of vegetables, the time period with the most significant spectral reflectance difference (such as the peak growth period or the maturity period) in the vegetable growth cycle is selected for remote sensing data collection and processing. This can effectively enhance the classification algorithm's ability to distinguish target crops from other landforms and significantly reduce the error rate. The present invention combines multi-time series remote sensing data to significantly improve classification accuracy, and for the first time combines phenological characteristics with high-resolution remote sensing images (10-meter resolution) for large-scale classification of vegetable planting areas.

[0015] (2) The present invention creates a vegetable planting area monitoring system and method for large-scale cultivated land, which combines vegetation indices and constructs a comprehensive and scalable classification framework by comprehensively analyzing the time series characteristics and spatial distribution patterns of vegetation indices (GNDVI, EVI, mNDWI, RVI, NDBI and LSWI). This effectively solves the classification problems of traditional methods in mixed planting and complex backgrounds, and can be used for precise management and planning of large-scale agricultural production.

[0016] (3) The present invention creates a system and method for monitoring vegetable planting areas across large arable land, generating high-resolution vegetable distribution maps. The present invention can generate large-scale vegetable planting distribution maps using 10-meter resolution remote sensing imagery, which is suitable for small, dispersed plots. The present invention achieves 10-meter resolution regional vegetable planting map generation in the black soil region of Northeast China. Classification using an RF machine learning classifier achieves an overall accuracy (OA) of 0.83 and a Kappa coefficient of 0.93.

[0017] (4) The present invention creates a system and method for monitoring vegetable planting areas across a wide range of cultivated land, creating an efficient remote sensing system suitable for large-scale monitoring. This system relies on open-source satellite data (e.g., Sentinel-2) and automated processing to reduce monitoring costs. By optimizing the system architecture, the present invention reduces data collection costs and reliance on high-cost equipment (e.g., drones), while achieving high classification accuracy and dynamic monitoring capabilities, thus avoiding the limited coverage of drone technology.

[0018] (5) The vegetable planting area monitoring system and method for large-scale cultivated land created by the present invention extracts characteristic time windows through analysis of time series information, classifies vegetables from other crops, and uses multi-year data to map vegetable planting areas, providing long-term data support for agricultural policy formulation and precise management. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which constitute part of the present invention, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings: Figure 1 A flow chart of a method for monitoring vegetable planting areas on a large scale of cultivated land according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a vegetable planting area monitoring system for large-scale cultivated land according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the phenological periods of various crops described in the embodiments of the present invention; Figure 4 Schematic diagram of the curve of the spectral index characteristics of each crop described in the embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not constitute a limitation of the present invention.

[0021] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other.

[0022] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second" and the like are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, features defined as "first", "second" and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0023] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to internal connections between two components. Those skilled in the art can understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0024] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0025] like Figure 1 As shown, the vegetable planting area monitoring method for large-scale cultivated land proposed in the present invention specifically includes the following steps: S1: obtaining multispectral data of each crop in the target area and sample points for describing crop characteristics; S2: selecting a characteristic phenological time window based on the vegetable phenological period, calculating the spectral index of the pixel corresponding to each sample point in the characteristic phenological time window, and taking the set of spectral indices of the pixel corresponding to each sample point as the spectral index feature of the target area; S3: repeating steps S1-S2 to obtain the spectral index features of the target area in different years, and inputting the spectral index features of the target area in different years as training samples into the random forest classifier that integrates multi-temporal data for training to obtain a trained random forest classification model; S4: repeating steps S1-S2 to obtain the spectral index features of the target area to be detected based on the satellite image of the target area to be detected, and inputting the spectral index features of the target area to be detected into the random forest classification model for processing to obtain a vegetable map within the target area to be monitored.

[0026] By combining VPC (vegetable phenological characteristics) with multi-time series data, the present invention accurately identifies vegetable growing areas and precisely captures dispersed small plots, thereby improving the classification accuracy of vegetable growing areas from 0.75 in the prior art to over 0.87. VPC refers to specific biological and physical characteristics based on different stages of the vegetable growth cycle (such as germination, flowering, fruiting, and maturity). These characteristics are captured and analyzed through remote sensing data to reflect the growth patterns and spatiotemporal distribution characteristics of crops. The core of the present invention is to utilize the differences in crop phenology and spectral performance to accurately distinguish target crops (vegetables) from other crops (such as rice and corn) or natural vegetation (such as weeds).

[0027] In some embodiments, in step S1 , the multispectral data includes spectral information of a red light band, a green light band, a blue light band, a near infrared band, and a short-wave infrared band.

[0028] It should be noted that the green band is 560 nanometers, the red band is 665 nanometers, the blue band is 490 nanometers, the near-infrared band is 842 nanometers, and the short-wave infrared band is 1610 nanometers.

[0029] Furthermore, during the data preparation process in step S1, multispectral data of the target area (including five bands: red, green, blue, and near-infrared) was obtained. A Global Positioning System (GPS) device was used to record vegetable field information (target area location information), and Sentinel-2 satellite imagery of the field was obtained. Based on the spectral characteristics of the target area sample points in the satellite imagery of the target area, sample points of different crops were selected on Google Earth (obtained through visual interpretation). These sample points were used to describe the crop characteristics and to train and test the results generated by the random forest classification model.

[0030] In some embodiments, the spectral index includes a green normalized difference vegetation index, an enhanced vegetation index, a ratio vegetation index, a modified normalized difference water index, a normalized difference accumulation index, and a surface water index, wherein Green Normalized Difference Vegetation Index The calculation formula is: ; in, The green band represents the Sentinel-2 satellite image. Represents the near-infrared band of Sentinel-2 satellite imagery; Enhanced Vegetation Index The calculation formula is: ; in, The red band represents the Sentinel-2 satellite image. The blue band represents the Sentinel-2 satellite image; Ratio Vegetation Index The calculation formula is: ; Modified Normalized Difference Water Index The calculation formula is: ; in, Represents the shortwave infrared band of Sentinel-2 satellite imagery; Normalized Difference Stacking Index The calculation formula is: ; Surface Water Index for: .

[0031] It should be noted that the spectral index of each sample point is calculated to process the spectral information of the sample point. GNDVI increases with the chlorophyll concentration of the plant; mNDWI provides auxiliary information for distinguishing the characteristics of these paddy fields from vegetable fields; NDBI includes near-infrared and short-wave infrared bands, which can reflect the moisture content of leaves, canopies, and soil, and can effectively distinguish vegetable planting areas from the surrounding background; the RVI index is closely related to the leaf area index (LAI), leaf dry biomass (DM), and chlorophyll content, which helps to exclude crops with relatively low chlorophyll content or obvious growth cycles, such as certain early-maturing or late-maturing crops; LSWI can maximize the differences between vegetable fields and rice fields.

[0032] In some embodiments, step S3 specifically includes the following steps: S31: In the target area, the time range corresponding to the vigorous stage of vegetable phenology is used as the characteristic phenological time window; S32: Within the characteristic phenological time window, based on the value, Value and value, and divide the target area into vegetation area and non-vegetation area; will satisfy both The value is between 0.2 and 0.7. The value is between 0.2 and 1.0. The area where the pixel value is greater than 0.2 is located is considered as the vegetation area, which will satisfy both Values below 0.2, Values below 0.2, The area where the pixels with a value less than 0.2 are located is considered as non-vegetation area; S33: In the vegetation area, based on the value, Value and value, and divide the vegetation area into vegetable growing area and non-vegetable growing area; will satisfy both The value is between 0.1 and 0.4. The value is between 0.5 and 0.8. The area where the pixels with values between 0.2 and 0.6 are located is regarded as the vegetable planting area, and the pixel set other than the vegetable planting area is regarded as the non-vegetable planting area.

[0033] It should be noted that in step S31, the principle for selecting the characteristic phenological time window is based on the following: the phenological periods of different crops vary significantly. Vegetables have short growth cycles and peak periods that differ from those of rice fields or other crops. Analysis of multi-year multi-time series spectral data reveals that the peak growth period of vegetables is typically from late September to early October each year, which differs from the maturity period of other crops (such as rice). Therefore, the present invention sets the characteristic phenological time window roughly from late September to early October each year. This period is a critical period for the harvest of major crops and the vigorous growth of vegetable crops. During this time period, the contrast between the target and the background in the captured image is high, effectively filtering out easily confused crops through the vegetation index, thereby maximizing the information difference between vegetable areas and other land cover types and accurately reflecting the growth range of vegetables. Multi-time series spectral reflectance data is: when a light source illuminates an object's surface, the object selectively reflects electromagnetic waves of different wavelengths. Spectral reflectance refers to the ratio of the light flux reflected by an object in a certain wavelength band to the light flux incident on the object, and is an essential property of the object's surface. Multi-time series spectral reflectance data refers to the changes in the reflectivity of ground objects to specific electromagnetic wave bands (such as visible light, near-infrared, and short-wave infrared) recorded at different time points using remote sensing technology. This data is usually presented in the form of a time series, reflecting the spectral characteristics of ground objects (such as vegetation, soil, and water bodies) during their growth cycle or changing environmental conditions.

[0034] To further facilitate understanding of the classification and monitoring methods of the present invention, six spectral indices—GNDVI, EVI, RVI, mNDWI, NDBI, and LSWI—were combined to analyze characteristic changes at rice, corn, soybean, and plant sample points during the growing season. First, a time window (late September to early October) was selected during which spectral differences between vegetables and other crops were significant. Vegetation index differences between vegetables and other crops during this period were screened. Specifically, vegetables had the highest GNDVI values, significantly different from corn. During this period, corn enters maturity, with yellowing leaves, reduced leaf area, and reduced ground cover. In contrast, vegetables are at their peak growth, with the highest biomass content and largest leaf area, resulting in significantly higher GNDVI values for corn than for vegetables. Vegetables, due to their high chlorophyll content and rapid growth, typically exhibit higher EVI and RVI values, effectively distinguishing green vegetation from non-vegetated areas. As part of natural vegetation, vegetables typically have lower NDBI values, which helps distinguish vegetable-grown areas from periods when bulk crops are exposed to bare soil. Significant differences in mNDWI values between vegetables and other crops are observed in the middle and late growing season. Wetlands, rice paddies, and grasslands may be misclassified as vegetable-growing areas due to their similar spectral characteristics. To address this, mNDWI and LSWI should be used to further eliminate interference from areas such as rice paddies. By intersecting the classification results of existing farmland mapping and excluding rice reflectance, rice paddies can be filtered out, thereby excluding irrigation areas and rice paddies and accurately extracting vegetable-growing areas.

[0035] In the present invention, a random forest classifier that integrates multi-temporal data is used for training, and the trained random forest classification model is used to improve the classification accuracy of vegetable planting areas. The random forest classifier is an integrated classification method based on decision trees. The final prediction result is obtained by constructing multiple decision trees and voting or averaging their classification results. Each decision tree uses randomly selected samples and features when constructing, which not only reduces the risk of overfitting but also enhances the robustness of classification. The random forest classifier performs well in processing high-dimensional data and multi-category classification tasks, and is particularly suitable for the classification of complex spectral features in remote sensing data. For example, it can distinguish target crops from other landforms through time series spectral data, and exhibits high classification accuracy and adaptability when dealing with the problem of imbalanced class samples.

[0036] Based on the differences in spectral indices within the characteristic phenological time windows of different crop sample points, spectral features for classification (i.e., spectral features of vegetable-growing areas) were extracted. Training samples from different years were fed into a random forest classifier that integrated multi-temporal data. By determining the parameters of the random forest classification model, the model automatically learned sample features and evaluated the importance of these parameters. This allowed the model to extract vegetable plots and output the distribution of vegetable-growing areas. In complex scenarios (such as areas where vegetables and rice fields alternate), a hierarchical classification approach significantly reduced misclassifications: The first stage eliminated non-vegetated areas; the second stage further distinguished vegetables from other crops.

[0037] In some embodiments, in step S3, the classification probability formula of the random forest classification model is: ; in, is the probability that data point x belongs to category c, n is the number of trees in the random forest model, is the prediction result of the i-th tree for the data point x, I is the indicator function, and n is the total number of decision trees in the random forest classification model.

[0038] To evaluate the accuracy of the vegetable map generated by the present invention, ground sampling data were compared with existing results. The classification accuracy of the random forest classification model within the mapping area was calculated using yearbook statistics.

[0039] The formula for evaluating the accuracy of the random forest classification model is: ; ; Among them, OA is the ratio of the number of correctly classified samples to the total number of samples in the mapping area, Kappa is the Kappa coefficient, TP is the number of correctly identified positive samples, FP is the number of negative samples that are incorrectly marked as positive samples, TN is the number of correctly identified negative samples, and FN is the number of unrecognized positive samples. is the ratio of the sum of the number of correctly classified samples in each category to the total number of samples, that is, the overall accuracy rate, is the expected probability of agreement under random classification.

[0040] like Figure 2 As shown, the present invention also provides a vegetable planting area monitoring system for large-scale cultivated land, comprising: Data acquisition module: acquires multispectral data of each crop in the target area and sample points used to describe crop characteristics; Spectral index calculation module: selects a characteristic phenological time window based on the vegetable phenological period, calculates the spectral index of the pixel corresponding to each sample point in the characteristic phenological time window, and uses the set of spectral indices of the pixel corresponding to each sample point as the spectral index feature of the target area; The classifier training module obtains the spectral index characteristics of the target area in different years and inputs the spectral index characteristics of the target area in different years as training samples into the random forest classifier that integrates multi-temporal data for training to obtain a trained random forest classification model; The monitoring module obtains the spectral index characteristics of the target area to be detected based on the satellite image of the target area to be detected, inputs the spectral index characteristics of the target area to be detected into the random forest classification model for processing, and obtains the vegetable map within the target area to be monitored.

[0041] The present invention combines multi-time series high-resolution remote sensing images, vegetable phenological characteristics (VPC) and optimized machine learning algorithms to achieve high-precision classification of vegetable planting areas, long-time series dynamic monitoring and change trend analysis.

[0042] like Figure 3 As shown in the figure, by analyzing the phenological calendars of Chinese cabbage, potatoes and cabbage in Northeast China, it was found that vegetables with bimodal growth curves were significantly different from other crops from June to August, so they were used as effective samples for classification; combined with remote sensing data throughout the year, it can be seen that the end of September to early October is the best identification window period - at this time, major crops are mature (low coverage) and vegetables are at their peak growth (largest leaf area and biomass). Vegetation indices (such as GNDVI and EVI) can significantly enhance image contrast, accurately distinguish vegetable planting areas from easily confused land types (such as rice fields and bare soil), and provide key phenological timing basis for regional crop classification.

[0043] like Figure 4 As shown, an analysis of crop characteristics during the 2023 growing season using six spectral indices revealed that vegetables exhibited significantly higher GNDVI, EVI, and RVI values during the peak season (September-October) due to high chlorophyll content and maximum leaf area, in stark contrast to mature corn (which exhibits yellowing leaves and low cover). mNDWI and LSWI can effectively distinguish vegetables from interfering areas such as rice paddies. Combined with farmland classification and reflectance exclusion, precise identification of vegetable planting areas is possible. mNDWI (focusing on water detection) and LSWI (reflecting vegetation moisture content) complement each other, and their combined use can reduce misidentification of wetlands and rice paddies.

[0044] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved. This is not limited herein.

[0045] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for monitoring vegetable planting areas on large-scale cultivated land, characterized by: The specific steps include: S1: Obtain multispectral data of each crop in the target area and sample points for describing crop characteristics; S2: Select a characteristic phenological time window based on the vegetable phenological period, calculate the spectral index of the pixel corresponding to each sample point in the characteristic phenological time window, and use the set of spectral indices of the pixel corresponding to each sample point as the spectral index feature of the target area; S3: Repeat steps S1-S2 to obtain the spectral index features of the target area in different years, and input the spectral index features of the target area in different years as training samples into the random forest classifier that integrates multi-temporal data for training to obtain a trained random forest classification model; S4: Repeat steps S1-S2, obtain the spectral index characteristics of the target area to be detected based on the satellite image of the target area to be detected, input the spectral index characteristics of the target area to be detected into the random forest classification model for processing, and obtain a vegetable map within the target area to be monitored.

2. The method for monitoring vegetable planting areas on large-scale cultivated land according to claim 1, characterized in that: In step S1, the multispectral data includes spectral information of a red light band, a green light band, a blue light band, a near infrared band, and a short-wave infrared band.

3. The method for monitoring vegetable planting areas on large-scale cultivated land according to claim 1, characterized in that: In step S1, satellite images of the target area are collected, and the location information of the target area is recorded using the global positioning system. Based on the location information of the target area and the image features contained in the satellite images of the target area, sample points of different crops are selected on Google Earth. The sample points are used to describe the crop characteristics, and the spectral information of the sample points contained in each crop is extracted using Arcgis software as the multispectral data of each crop.

4. The method for monitoring vegetable planting areas on large-scale cultivated land according to claim 1, characterized in that: In step S2, the spectral index includes green normalized difference vegetation index, enhanced vegetation index, ratio vegetation index, modified normalized difference moisture index, normalized difference accumulation index and surface water index, wherein, Green Normalized Difference Vegetation Index The calculation formula is: ； in, The green band represents the Sentinel-2 satellite image. Represents the near-infrared band of Sentinel-2 satellite imagery; Enhanced Vegetation Index The calculation formula is: ； in, The red band represents the Sentinel-2 satellite image. The blue band represents the Sentinel-2 satellite image; Ratio Vegetation Index The calculation formula is: ； Modified Normalized Difference Water Index The calculation formula is: ； in, Represents the shortwave infrared band of Sentinel-2 satellite imagery; Normalized Difference Stacking Index The calculation formula is: ； Surface Water Index for: 。 5. The method for monitoring vegetable planting areas on large-scale cultivated land according to claim 4, characterized in that: The step S3 specifically includes the following steps: S31: In the target area, the time range corresponding to the vigorous stage of vegetable phenology is used as the characteristic phenological time window; S32: Within the characteristic phenological time window, based on the value, Value and value, and divide the target area into vegetation area and non-vegetation area; will satisfy both The value is between 0.2 and 0.

7. The value is between 0.2 and 1.

0. The area where the pixel value is greater than 0.2 is located is considered as the vegetation area, which will satisfy both Values below 0.2, Values below 0.2, The area where the pixels with a value less than 0.2 are located is considered as non-vegetation area; S33: In the vegetation area, based on the value, Value and value, and divide the vegetation area into vegetable growing area and non-vegetable growing area; will satisfy both The value is between 0.1 and 0.

4. The value is between 0.5 and 0.

8. The area where the pixels with values between 0.2 and 0.6 are located is regarded as the vegetable planting area, and the pixel set other than the vegetable planting area is regarded as the non-vegetable planting area.

6. The method for monitoring vegetable planting areas on large-scale cultivated land according to claim 1, characterized in that: In step S3, the classification probability formula of the random forest classification model is: ； in, is the probability that data point x belongs to category c, n is the number of trees in the random forest model, is the prediction result of the i-th tree for the data point x, I is the indicator function, and n is the total number of decision trees in the random forest classification model.

7. The method for monitoring vegetable planting areas on large-scale cultivated land according to claim 1, characterized in that: The formula for evaluating the accuracy of the random forest classification model is: ； ； Among them, OA is the ratio of the number of correctly classified samples to the total number of samples in the mapping area, Kappa is the Kappa coefficient, TP is the number of correctly identified positive samples, FP is the number of negative samples that are incorrectly marked as positive samples, TN is the number of correctly identified negative samples, and FN is the number of unrecognized positive samples. is the ratio of the sum of the number of correctly classified samples in each category to the total number of samples, that is, the overall accuracy rate, is the expected probability of agreement under random classification.

8. A vegetable planting area monitoring system for large-scale cultivated land, characterized by: include: Data acquisition module: acquires multispectral data of each crop in the target area and sample points used to describe crop characteristics; Spectral index calculation module: selects a characteristic phenological time window based on the vegetable phenological period, calculates the spectral index of the pixel corresponding to each sample point in the characteristic phenological time window, and uses the set of spectral indices of the pixel corresponding to each sample point as the spectral index feature of the target area; The classifier training module obtains the spectral index characteristics of the target area in different years and inputs the spectral index characteristics of the target area in different years as training samples into the random forest classifier that integrates multi-temporal data for training to obtain a trained random forest classification model; The monitoring module obtains the spectral index characteristics of the target area to be detected based on the satellite image of the target area to be detected, inputs the spectral index characteristics of the target area to be detected into the random forest classification model for processing, and obtains the vegetable map within the target area to be monitored.

Images