A winter field monitoring method and system based on minimization threshold method

Abstract

The present invention discloses a winter field monitoring method and system based on a minimization threshold method, comprising the following steps: selecting a research area; obtaining basic image data based on Google Earth Engine; obtaining a long-time series NDVI curve using a multi-source remote sensing image fusion method to generate time series data; obtaining a crop growth calendar based on the time series data; obtaining winter field remote sensing image sample point annotation data and field survey sample point data of different utilization modes; integrating the winter field remote sensing image sample point annotation data and field survey sample point data to form a sample point data set; dividing the sample point data set into winter climate zones, and constructing a winter fallow field extraction model based on the divided training data set; and using a test data set to evaluate the winter fallow field extraction model, wherein the evaluation data includes calculation accuracy, recall rate indicators, and prediction results of a visualization model. The present application realizes an efficient and real-time winter field utilization analysis method.

Description

Technical Field

[0001] The present invention relates to the technical field of winter idle field monitoring, and in particular to a winter field monitoring method and system based on a minimization threshold method. Background Art

[0002] In southern China, there is a large amount of winter fallow land, which is different from abandoned farmland. It is called winter fallow land. In recent decades, a large amount of double-season rice farmland has been converted to single-season rice farmland, resulting in farmland being idle between the autumn harvest and the growing season of the second year. Some of these fallow lands still have temperature potential for crop growth and development in winter. Although they cannot be planted at present, considering the current rate of climate warming, these lands still have considerable planting potential. It is estimated that under future climate conditions, planting rapeseed on fallow land in winter in the middle and lower reaches of the Yangtze River can reduce soybean imports by about 50%. Increasing the tillage intensity of arable land is an effective way to improve the productivity of arable land. The shortage and spatial imbalance of arable land and water resources in my country restrict the sustainable development of grain production. Therefore, it is crucial to make full use of the production potential of winter fallow land in southern China and increase tillage intensity and grain output.

[0003] In recent years, with the emphasis on food security and the proposal of a new round of "100 billion jin" grain production increase plan. In recent years, the country has repeatedly emphasized the attention and importance of "reserve arable land", "idle land" and "winter idle field utilization". According to relevant reports and research, there are a large number of winter idle fields in southern my country. If these winter idle fields can be fully and reasonably developed and utilized, they will become an important means to increase my country's grain production. In recent years, the middle and lower reaches of the Yangtze River have widely promoted measures such as planting rapeseed and green manure in winter idle fields, which have been very effective. Therefore, real-time monitoring of the scale, duration and utilization of winter idle fields will provide important data support for the precise implementation of policies. However, there is still a certain lag in winter idle field monitoring technology and timeliness. First, the scale of large-scale monitoring is coarse and cannot guide actual production. Traditional winter idle field utilization monitoring usually relies on medium and low resolution images (above 1 km), while the arable land in southern my country is fragmented, and winter idle fields are often not continuous, resulting in a more fragmented and fine distribution of winter idle fields. Therefore, most high-precision winter idle research is at the scale of counties and prefecture-level cities. Monitoring on a national scale is difficult. In addition, the generation of fallow land in winter is subjective, and the possibility of farmers planting in different years varies greatly. In the past, national-scale fallow land monitoring products and research were mostly concentrated before 2015. This has obviously lagged behind today's actual situation.

[0004] Many studies have previously used remote sensing technology to extract China's winter fallow fields. In the large-scale winter fallow field extraction, medium-resolution images such as MODIS are usually used, which can provide images with high temporal resolution. The time interval of winter fallow of cultivated land can be obtained through the vegetation index (VI) curve of the time series. For example, based on MODIS-NDVI images, large-scale extraction of winter fallow fields in southern China can be carried out by adopting the dynamic threshold (DT) method. In addition, image data with different spatial resolutions such as SPOT, GF, Sentinel 2 and Landsat are constantly used to extract winter fallow fields of various scales. Medium-resolution images such as MODIS (250-1000 meters) are mainly used at provincial and larger spatial scales. However, due to the relatively low spatial resolution of MODIS image data, it is not suitable for the identification of winter fallow fields in southern China. This is mainly because my country's cultivated land is mainly manifested as scattered patches or plots, 80% of which are less than 0.6 hectares, or even less than 0.04 hectares. Therefore, from the perspective of spatial resolution, the GF series, Sentinel 2, and Landsat series seem to be more suitable for extracting winter fallow fields in southern China. However, the revisit intervals of these images are very sensitive to atmospheric conditions and cloud effects, so their use will still be limited. In recent years, many studies have combined the advantages of high spatial resolution and high temporal resolution through the spatiotemporal fusion of multi-source images to make up for the shortcomings of time series in high spatial resolution images. Common image fusion methods include GF-SG, IFSDAF, FF, STAIR, STARFMA&ESTARFM, etc. Fusion algorithms are often complex and require a lot of computing resources, so computational efficiency is also an important factor in method selection.

[0005] There are two main methods for extracting fallow land in winter: First, at the provincial and larger spatial scales, the threshold method is mainly used to extract fallow land in winter. The threshold method is simple and efficient, but the threshold is difficult to determine, which usually directly affects the results; second, at a spatial scale smaller than the provincial level, based on high spatial resolution images, supervised classification methods such as random forests and decision trees can achieve higher accuracy. These methods do not require multi-temporal images, but require high-precision images of key time periods and a large number of training samples.

[0006] In summary, the current winter fallow field monitoring has the following problems:

[0007] (1) Subjectivity and experience: The dynamic threshold method (DTM) is one of the mainstream methods for vegetation phenology retrieval. However, it has the following two disadvantages. First, the threshold determined by using the subjective threshold percentage of the phenological node based on the user's experience threshold is not universal. Second, a common threshold is used to identify large-area phenology, and the physiological characteristics of different vegetation types are rarely considered, thereby blurring the phenological variability between vegetation types. In the past, large-scale monitoring mainly used empirical dynamic thresholds to identify winter fallow fields. The thresholds were mainly obtained by directly taking the thresholds from other fields such as forests and other countries or regions, ignoring the geographical environment and vegetation physiological differences. Although a certain degree of accuracy has been achieved, it is far from enough to guide production.

[0008] (2) Low resolution: In the past, the identification of fallow fields at the provincial level and above mostly used MODIS image sources with a resolution of 250-1000 meters, which is of medium resolution. However, the distribution of farmland in my country is fragmented, especially in the hilly south, where the farmland is more fragmented. The occurrence of fallow fields is subjective and does not exist in large areas. In this case, the pixel mixing effect of medium-resolution images for fallow field identification is strong, the error is large, and it is difficult to guide the planning of cultivated land in actual production.

[0009] (3) Low efficiency and high cost: Generally, fine resolution refers to 30 meters or less, and Landsat\Sentinel series images are commonly used. Due to limited computing resources, high-precision winter fallow field identification is mostly applied at the prefecture level or below, and only provides single-period results, which limits the monitoring frequency and scale.

[0010] (4) Timeliness and real-time: The occurrence and monitoring of fallow land in winter require timeliness to guide decision-making. The identification of large-scale fallow land in winter has a long time span and lags, and lacks real-time identification and updating. This delay leads to insufficient support for timely intervention and decision-making on the utilization of winter fields. Summary of the invention

[0011] In view of the above technical problems in the related art, the present invention provides a winter field monitoring method and system based on a minimization threshold method, which can solve the above problems.

[0012] To achieve the above technical objectives, the technical solution of the present invention is implemented as follows:

[0013] A winter field monitoring method based on a minimization threshold method comprises the following steps:

[0014] S1. Select the research area;

[0015] S2, obtain basic image data based on Google Earth Engine;

[0016] S3, using multi-source remote sensing image fusion method to obtain long-term NDVI curve and generate time series data;

[0017] S4, obtaining a crop growth calendar based on time series data;

[0018] S5. Obtain the annotation data of remote sensing image sample points and field survey sample point data of winter fields with different utilization modes. The categories of remote sensing image sample points include winter fallow fields and winter crop-planted fields. The field survey sample point data are obtained through field surveys and include winter fallow field data and farmland crop data.

[0019] S6. Based on the winter field remote sensing image sample point annotation data and the field survey sample point data, a sample point data set is formed, wherein the sample point data set includes a training data set and a test data set;

[0020] S7, dividing the sample data set into winter climate zones, and constructing a winter fallow field extraction model based on the divided training data set;

[0021] S8. Use the test data set to evaluate the fallow field extraction model. The evaluation data includes the calculation of accuracy and recall rate indicators, as well as the prediction results of the visualization model.

[0022] Furthermore, step S2 also includes image orthorectification, image stitching, cloud removal, and filtering and smoothing processing on the acquired basic image data.

[0023] Furthermore, the multi-source remote sensing image fusion method in step S3 specifically includes: using MODIS image data and Landsat image data to fuse to reconstruct high spatiotemporal resolution images, performing cloud removal and time series linear interpolation on MODIS image data and Landsat image data respectively, filling in images with missing time, using the GF-SG filtering method to perform time series reconstruction to obtain a relatively complete time series curve, and performing time series reconstruction on a grid-by-grid basis based on Google Earth Engine computing resources.

[0024] Furthermore, step S4 specifically includes: calculating the first-order and second-order derivatives of the NDVI curve to obtain extreme points, the maximum point is the peak, the minimum point is the trough, and the two troughs corresponding to each peak are considered to be a crop planting cycle.

[0025] Furthermore, the winter field remote sensing image sample point annotation data and field survey sample point data in step S5 also include the following contents: geographical location information, key phenological information, variety information, and canopy spectral information of fields with different utilization modes.

[0026] Furthermore, step S6 includes the following steps: obtaining object feature data of different utilization conditions in winter based on winter field remote sensing image sample point annotation data and field survey sample point data, wherein the object feature data is composed of time and space relationship features, the time feature is composed of the length and duration of the crop growing period, and the canopy texture and spectral changes corresponding to different growing periods, and the spatial relationship feature is composed of the relative position and distribution pattern of the field.

[0027] Furthermore, the winter climate zone division in step S7 specifically includes: dividing the entire study area into 0℃, 5℃, 10℃, 15℃, and 20℃ climate zones according to the winter average temperature, extracting the NDVI distribution of different climate zones, and establishing different NDVI probability density functions according to different climate zones to ensure that the probability density function finally constructed can adapt to different climate zones.

[0028] Furthermore, based on the climate zones where the winter field remote sensing image samples and the field survey samples were located, the probability density function and Gaussian distribution curve of the winter fallow field and non-winter fallow field categories were constructed for the NDVI values ​​of the samples in the same zone. The intersection of the Gaussian curves of different land types was used as the optimal classification threshold, and the classification threshold was used as the probability threshold for winter field classification. The probability threshold was substituted into the winter period in the crop growth calendar grid by grid using Google Earth Engine to identify unused land in winter.

[0029] A winter field monitoring system based on a minimization threshold method, comprising:

[0030] Basic image data module, used to obtain image data of winter fields located in the study area;

[0031] An image data processing module, used for performing data estimation processing on the acquired image data;

[0032] The crop calendar monitoring module obtains the crop growth calendar based on time series data;

[0033] Real-time monitoring module, used for real-time monitoring of winter field image data;

[0034] Field survey data module, used to obtain field survey sample data;

[0035] Winter field image annotation data module, used to obtain winter field remote sensing image sample point annotation data;

[0036] Winter field object feature data module, which obtains object feature data of different utilization conditions in winter based on winter field remote sensing image sample point annotation data and field survey sample point data;

[0037] A sensor module is used to collect temperature information of the sample point area;

[0038] Farmland climate resource calculation module, used to divide the sample data set into winter climate zones;

[0039] The probability threshold calculation update module extracts the NDVI distribution in different climate zones and establishes different NDVI probability density functions according to different climate zones;

[0040] Winter field utilization analysis module, which analyzes winter field utilization based on the winter fallow field extraction model;

[0041] The grid worker feedback module provides grid workers with feedback on the analysis of winter field utilization.

[0042] Beneficial effects of the invention: This application is based on the principle that winter fields under different utilization conditions have different representations on remote sensing images, and constructs probability density functions of winter idle and non-winter idle sample points based on a large number of samples, thereby realizing an efficient and real-time winter field utilization analysis method. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] The present invention is further described in detail below with reference to the accompanying drawings.

[0045] Figure 1 It is a structural principle diagram of a winter field monitoring system based on a minimization threshold method according to an embodiment of the present invention;

[0046] Figure 2 is a schematic diagram of a crop growth calendar according to an embodiment of the present invention;

[0047] Figure 3 Schematic diagram of probability density function of fallow fields and non-fallow fields in different climate zones according to an embodiment of the present invention;

[0048] Figure 4 is a time series curve diagram of various sample points described in an embodiment of the present invention;

[0049] Figure 5 It is an example diagram of normal distribution fitting of time series of various sample points described in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0051] like Figure 1-5 As shown, according to the present invention, a winter field monitoring method based on a minimization threshold method is disclosed, comprising the following steps:

[0052] S1. Fifteen provinces, autonomous regions and municipalities in southern my country, including Hunan, Hubei, Jiangxi, Jiangsu, Sichuan, Anhui, Yunnan, Guizhou, Chongqing, Zhejiang, Fujian, Hainan, Guangdong, Guangxi Zhuang Autonomous Region and Shanghai, were selected as the research areas.

[0053] S2, based on Google Earth Engine (GEE), obtain basic image data, perform image orthorectification, image stitching, cloud removal, and filtering and smoothing;

[0054] Image orthorectification is used to eliminate geometric distortion caused by terrain undulation and sensor tilt, including obtaining metadata (including sensor information, shooting time, geographic coordinates, etc.), building models (using geographic information system (GIS) tools to build geometric models of images), applying ground control points (adjusting using ground control points with known coordinates), and resampling (reprojecting images to a standard coordinate system, usually using bilinear or cubic convolution interpolation methods).

[0055] Image stitching is used to combine multiple images into a seamless panoramic image, including feature extraction (using algorithms (such as SIFT, SURF, etc.) to extract feature points in overlapping areas), feature matching (matching feature points in adjacent images), transformation calculation (calculating affine or perspective transformation matrix), image fusion (weighted averaging or other fusion methods for overlapping areas to eliminate seams), etc.

[0056] Declouding is used to remove clouds and shadows from images to improve image quality and usability, including cloud detection (using threshold segmentation, machine learning or deep learning models to identify cloud areas), mask generation (creating masks of clouds and shadows), and declouding (using interpolation, texture synthesis or image restoration technology to fill in cloud-covered areas and restore surface information under the clouds).

[0057] Among them, filtering and smoothing processing is used to reduce image noise and improve image quality, including filter selection (common filters include mean filter, median filter, Gaussian filter, etc.), filter application (apply the selected filter to the image, process each pixel, smooth the image, and reduce noise), parameter adjustment (adjust the filter parameters (such as window size) as needed to achieve the best effect) and other steps.

[0058] S3, using multi-source remote sensing image fusion method to obtain long-term series NDVI (normalized difference vegetation index) curve and generate time series data;

[0059] Normalized Difference Vegetation Index (NDVI), the calculation formula is as follows:

[0060] ;

[0061] Its essence is an indicator that reflects the degree of vegetation coverage. NDVI can be used to quantitatively analyze vegetation growth and understand the changes in vegetation coverage. The general NDVI value range is between -1 and 1. The larger the value, the better the growth condition. A value less than 0 generally indicates a water body. In the above formula, NIR represents the reflectivity of the near-infrared band (845-885nm), and RED represents the reflectivity of the red band (630-680nm). Image sources with different spatial resolutions have their own advantages. MODIS image data has high temporal resolution but low spatial resolution. Landsat image data has high spatial resolution but low temporal resolution. In large-scale applications, it is difficult to ensure temporal continuity due to the influence of cloudy and rainy weather.

[0062] This application uses MODIS and Landsat image fusion to reconstruct high spatiotemporal resolution images. MODIS and Landsat are declouded and time series linear interpolated to fill in the missing time images. GapFilling and Savitzky-Golay (GF-SG) filtering method is used for time series reconstruction to obtain a relatively complete time series curve. Time series reconstruction is performed grid by grid unit based on Google Earth Engine computing resources.

[0063] S4. Obtain the crop growth calendar based on the time series data, as follows: Figure 2 As shown, the first and second order derivatives of the NDVI curve are calculated to obtain the extreme points. The maximum point is the peak, and the minimum point is the trough. The two troughs on both sides of each peak are considered to be a crop planting cycle.

[0064] It is necessary to explain that this application not only identifies the fallow fields in winter, but also obtains the fallow period (starting period and ending period). The acquisition of the fallow period requires combining the obtained probability threshold with the crop growth calendar to obtain the fallow start and ending period, such as Figure 2 As shown, the left bulge is a crop growing season, the right bulge is a growing season, and the middle part is the winter fallow period, during which the probability threshold should be used.

[0065] S5. Obtain winter field remote sensing image sample annotation data and field survey sample data of different utilization methods. The categories of remote sensing image samples include winter fallow fields and winter crop-planted fields. The field survey sample data are obtained through field surveys and include winter fallow field data and farmland crop data. Field surveys refer to data collection and field investigation activities conducted in actual environments. Farmland crops include green manure and other returning crops, winter wheat and winter rapeseed and other field crops, and short-maturing economic crops such as potatoes and vegetables.

[0066] Winter field remote sensing image sample annotation data and field survey sample data also include the following: geographical location information of fields with different utilization methods, key phenological information, variety information, and canopy spectral information.

[0067] S6. Based on the winter field remote sensing image sample point annotation data and the field survey sample point data, a sample point data set is formed, wherein the sample point data set includes a training data set and a test data set;

[0068] Based on the winter field remote sensing image sample annotation data and field survey sample data, the object feature data of different winter utilization conditions are obtained. The object feature data are composed of time and spatial relationship characteristics. The time characteristics are composed of the length and duration of the crop growing period, and the canopy texture and spectral changes corresponding to different growing periods. The spatial relationship characteristics are composed of the relative position and distribution pattern of the fields.

[0069] In order to generate multiple samples from a single sample, bootstrapping is used as an alternative to cross-validation to perform random sampling with replacement in the original sample to obtain new samples. The samples obtained by bootstrap overlap more than the samples obtained by cross-validation, so their estimates are more dependent, and it is considered to be the best way to resample small data sets. This application uses the bootstrap method to perform multiple repeated samplings with replacement to avoid accidental errors caused by a single division affecting the performance evaluation of the model. The ratio of training set to test set is approximately 63.2% and 36.8%.

[0070] S7, dividing the sample data set into winter climate zones, and constructing a winter fallow field extraction model based on the divided training data set;

[0071] Specific as Figure 3As shown in the figure, the entire study area is divided into 0℃, 5℃, 10℃, 15℃, and 20℃ climate zones according to the average winter temperature (December, January, and February). The NDVI distribution in different climate zones is extracted, and different NDVI probability density functions are established according to different climate zones to ensure that the final constructed probability density function can adapt to different climate zones and minimize the misjudgment caused by NDVI value fluctuations caused by temperature differences.

[0072] S8. Use the test data set to evaluate the winter fallow field extraction model. The evaluation data includes calculating the accuracy and recall rate indicators, as well as the prediction results of the visualization model;

[0073] ;

[0074] Where OA is the overall accuracy, PA is the producer accuracy, UA is the user accuracy, and F1 is the F1 score; pkk represents the number of correct identifications for each category. pii and pjj are the number of correct classifications for category i and category j, respectively, and pi+ and p+j are the total number of true samples of category i and the total number of samples predicted to be category j, respectively.

[0075] like Figure 2 , 4 As shown in Figure 5, based on the above-mentioned zoning based on the climate zones where the remote sensing sample points are marked and the ground survey sample points are located, the probability density function and Gaussian distribution curve of the winter fallow field and non-winter fallow field categories are constructed for the NDVI values ​​of the sample points in the same zone, and the intersection of the Gaussian curves of different land types is used as the optimal classification threshold (i.e., probability threshold). GEE is used to substitute the probability threshold into the winter fallow period identified based on the crop growth calendar to obtain the accurate results of the winter fallow field.

[0076] The present application also provides a winter field monitoring system based on the minimization threshold method, including: a basic image data module, an image data processing module, a field survey data module, a winter field image annotation data module, a winter field object feature data module, a sensor module, a farmland climate resource calculation module, a probability threshold calculation update module, a crop calendar monitoring module, a real-time monitoring module, a winter field utilization analysis module, and a grid worker feedback module.

[0077] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A winter field monitoring method based on the minimization threshold method, characterized in that: The following steps are involved: S1. Select the research area; S2, obtain basic image data based on Google Earth Engine; S3, using multi-source remote sensing image fusion method to obtain long-term NDVI curve and generate time series data; The multi-source remote sensing image fusion method in step S3 specifically includes: using MODIS image data and Landsat image data to fuse to reconstruct high spatiotemporal resolution images, performing cloud removal and time series linear interpolation on MODIS image data and Landsat image data respectively, filling in images with missing time, using GF-SG filtering method to perform time series reconstruction to obtain a relatively complete time series curve, and performing time series reconstruction on a grid-by-grid basis based on Google Earth Engine computing resources; S4, obtaining a crop growth calendar based on time series data; S5. Obtain the annotation data of remote sensing image sample points and field survey sample point data of winter fields with different utilization modes. The categories of remote sensing image sample points include winter fallow fields and winter crop-planted fields. The field survey sample point data are obtained through field surveys and include winter fallow field data and farmland crop data. S6. Based on the winter field remote sensing image sample point annotation data and the field survey sample point data, a sample point data set is formed, wherein the sample point data set includes a training data set and a test data set; S7, dividing the sample data set into winter climate zones, and constructing a winter fallow field extraction model based on the divided training data set; The winter climate zone division in step S7 specifically includes: dividing the entire study area into 0°C, 5°C, 10°C, 15°C, and 20°C climate zones according to the winter average temperature, extracting the NDVI distribution of different climate zones, and establishing different NDVI probability density functions according to different climate zones to ensure that the probability density functions finally constructed can adapt to different climate zones; Based on the climate zones where the winter field remote sensing image samples and field survey samples are located, the probability density function and Gaussian distribution curve of the winter fallow field and non-winter fallow field categories are constructed for the NDVI values ​​of the same partitioned samples. The intersection of the Gaussian curves of different land types is used as the optimal classification threshold. The classification threshold is used as the probability threshold for winter field classification. The probability threshold is substituted into the winter period in the crop growth calendar by Google Earth Engine for winter unused land identification. S8. Use the test data set to evaluate the fallow field extraction model. The evaluation data includes the calculation of accuracy and recall rate indicators, as well as the prediction results of the visualization model.

2. A winter field monitoring method based on the minimization threshold method according to claim 1, characterized in that: Step S2 also includes image orthorectification, image stitching, cloud removal, and filtering and smoothing processing on the acquired basic image data.

3. A winter field monitoring method based on the minimization threshold method according to claim 1, characterized in that: Step S4 specifically includes: calculating the first-order and second-order derivatives of the NDVI curve to obtain extreme points, the maximum point is the peak, the minimum point is the trough, and the two troughs on both sides of each peak are considered to be a crop planting cycle.

4. A winter field monitoring method based on the minimization threshold method according to claim 1, characterized in that: The winter field remote sensing image sample point annotation data and field survey sample point data in step S5 also include the following contents: geographical location information, key phenological information, variety information, and canopy spectrum information of fields with different utilization modes.

5. A winter field monitoring method based on the minimization threshold method according to claim 1, characterized in that: Step S6 includes the following steps: based on the winter field remote sensing image sample point annotation data and the field survey sample point data, the object feature data of different utilization conditions in winter are obtained, and the object feature data are composed of time and space relationship characteristics. The time characteristics are composed of the length and duration of the crop growth period, and the canopy texture and spectral changes corresponding to different growth periods. The spatial relationship characteristics are composed of the relative position and distribution pattern of the fields.

6. A winter field monitoring system based on the minimization threshold method, characterized in that: The system is used to execute the winter field monitoring method based on the minimization threshold method as described in any one of claims 1 to 5, comprising: Basic image data module, used to obtain image data of winter fields located in the study area; An image data processing module, used for performing data estimation processing on the acquired image data; The crop calendar monitoring module obtains the crop growth calendar based on time series data; Real-time monitoring module, used to monitor winter field image data in real time; field survey data module, used to obtain field survey sample data; Winter field image annotation data module, used to obtain winter field remote sensing image sample point annotation data; Winter field object feature data module, which obtains object feature data of different utilization conditions in winter based on winter field remote sensing image sample point annotation data and field survey sample point data; A sensor module is used to collect temperature information of the sample point area; Farmland climate resource calculation module, used to divide the sample data set into winter climate zones; The probability threshold calculation update module extracts the NDVI distribution in different climate zones and establishes different NDVI probability density functions according to different climate zones; Winter field utilization analysis module, which analyzes winter field utilization based on the winter fallow field extraction model; The grid worker feedback module provides grid workers with feedback on the analysis of winter field utilization.

Images