A method for monitoring the pod setting period of soybeans in the growing season based on remote sensing data

Abstract

The application relates to a soybean pod setting period monitoring method based on remote sensing data in a growth season, which comprises the following steps: step 1, obtaining soybean planting area data and historical soybean pod setting period time; step 2, obtaining historical and current available VIIRS ground reflectivity data, and calculating historical EVI2 time sequence images and current EVI2 time sequence images; step 3, obtaining a historical average time sequence curve; step 4, matching and extracting the soybean pod setting period; matching and fitting the current EVI2 time sequence curve of the soybean planting pixel with the reference EVI2 time sequence curve, and predicting the possible pod setting period time of the current year; step 5, extracting the result evaluation; comparing the matching result with the soybean pod setting period time of the current ground investigation, analyzing the difference between the two, and calculating the deviation and the accuracy. The application has high monitoring accuracy, is simple and convenient, and saves manpower, material resources and time.

Description

TECHNICAL FIELD

[0001] The application relates to a soybean pod setting period monitoring method based on remote sensing data in a growth season, and mainly applies to soybean pod setting period monitoring in a soybean growth season. BACKGROUND

[0002] Accurate and timely estimation of the soybean pod setting period is crucial for effective soybean planting management and soybean model modeling. Traditional soybean pod setting period observation needs to be carried out in the field, and it is difficult to realize regional-scale soybean pod setting period phenology observation, so the research on real-time or near real-time regional soybean growth season monitoring and prediction is still less. SUMMARY

[0003] The application solves the technical problem of overcoming the above-mentioned deficiencies in the prior art, and provides a soybean pod setting period monitoring method based on remote sensing data, which is high in monitoring accuracy and simple and convenient. The application uses satellite observation period and phenology algorithm, and soybean phenology can be obtained through vegetation index (VI) curve. Compared with the traditional method, regional-scale soybean pod setting period can be quickly obtained.

[0004] The technical solution adopted by the application to solve the above technical problem is: a soybean pod setting period monitoring method based on remote sensing data in a growth season, characterized by comprising the following steps:

[0005] Step 1, obtaining soybean planting area data and soybean pod setting period time in previous years;

[0006] According to the spatial distribution map of soybean planting in previous years, the planting frequency of all soybean planting pixels is calculated, and the soybean pod setting period time in previous years is obtained for all soybean planting pixels, which is used as the basis for predicting the soybean pod setting period of the soybean planting pixel in this year;

[0007] Step 2, obtaining the VIIRS surface reflectivity data in previous years and this year, and calculating the EVI2 time series image;

[0008] Obtain all available VIIRS surface reflectivity data in previous years and this year, and calculate the EVI2 time series image in previous years and the EVI2 time series image before the current time point in this year;

[0009] Step 3, obtaining the average time series curve in previous years;

[0010] In order to monitor the remote sensing of soybean growth in the study area, the EVI2 time series image dataset in previous years in the study area is calculated, and the mean value of the EVI2 time series dataset of each soybean planting pixel in the study area is used as the average level of soybean growth in the whole study area, so as to obtain the reference EVI2 time series curve (referred to as reference curve) of the whole study area;

[0011] Step 4, matching to extract the soybean pod setting period;

[0012] In order to monitor the soybean pod setting period time of each soybean planting pixel, matching is performed according to the EVI2 time series data obtained in the current year. Specifically, after obtaining new EVI2 time series data for each soybean planting pixel, the current EVI2 time series curve of the soybean planting pixel is matched with the soybean reference EVI2 time series curve of the study area (i.e., soybean growth stage time series curve matching), and the possible soybean pod setting period time of the soybean planting pixel in the current year is predicted by comparing the vegetation growth state;

[0013] Step 5, evaluation of the extraction result;

[0014] The matching result is compared with the soybean pod setting period time obtained by ground investigation in the current year, and the differences in the three data (i.e., the start day of the pod setting period, the end day of the pod setting period, and the total duration of the pod setting period) are analyzed and compared to evaluate the deviation and accuracy of the monitored and matched pod setting period time from the actual ground observation result, to further understand the feasibility and limitations of remote sensing monitoring, and to provide important reference for achieving more accurate soybean pod setting period monitoring.

[0015] The calculation formula of the dual-band enhanced vegetation index (hereinafter referred to as EVI2) time series image is:

[0016] EVI2 = 2.5 × ((NIR-Red) / (NIR+2.4×Red+1))

[0017] Wherein, NIR represents the near-infrared band reflectivity, and Red represents the red light band reflectivity.

[0018] The matching to extract the soybean pod setting period matches the current EVI2 time series curve of the soybean planting pixel with the soybean reference EVI2 time series curve, and the soybean reference EVI2 time series curve is locally translated and scaled to obtain the current time point soybean fitting EVI2 time series curve (this year's predicted EVI2 time series curve, referred to as fitting curve) of the soybean planting area, and the calculation method is as follows:

[0019]

[0020] Wherein, represents the current year's soybean fitting EVI2 time series curve after translation and scaling transformation within the half time window before and after the reference curve pod setting period (the total duration of the reference curve pod setting period w, each half time window width is added before and after the total duration of the reference curve pod setting period w, and the total time window width is 2w), g(x) represents the soybean reference curve, xscale and t shiftrespectively represent the scaling and translation parameters in the x-axis (time axis) direction, and p0 represents the soybean flowering period on the soybean reference curve.

[0021] The correlation calculation formula between the fitted EVI2 time series curve (fitting curve) of the soybean planting area at the current time point and the local actual EVI2 time series curve within the half-time window before and after the pod setting period of the soybean reference curve of the soybean planting area is as follows:

[0022]

[0023] x∈(p0-t shift -w,p0-t shift +w)

[0024]

[0025] wherein x and y respectively represent the daily EVI2 time series values, N represents the time series length within the total time window range, and the maximum value of the correlation is taken to determine the best fitting parameter, at which time the fitting effect is best.

[0026] The extraction formula of the predicted pod setting period of the soybean planting pixel is as follows:

[0027] xscale×(x+tshift)+(1--xscale)×p0=p0

[0028] p est =p0-tshift

[0029] p est refers to the predicted pod setting period.

[0030] Compared with the prior art, the present application has the following advantages and effects: high monitoring accuracy, simplicity, convenience, saving of manpower, material resources and time. BRIEF DESCRIPTION OF DRAWINGS

[0031] Fig. 1 is a flowchart of an embodiment of the present application.

[0032] Fig. 2 is a comparison diagram of the soybean target (actual) curve, the historical reference curve and the fitting (monitoring of the present application) curve.

[0033] Fig. 3 is a monitoring result diagram of the soybean pod setting period in the soybean growing season in Nebraska, USA, realized on the 248th day of the calendar time. DETAILED DESCRIPTION

[0034] The present application will be further described in detail below in combination with the drawings and through embodiments, and the following embodiments are an explanation of the present application but the present application is not limited to the following embodiments.

[0035] Reference Figs. 1-3 The soybean pod setting period monitoring method based on remote sensing data in the growing season of the present application is mainly used to monitor the time interval and state evaluation of the pod setting period of a soybean planting area (study area) in a certain soybean growing season (for example, the middle of the growing season, the flowering period, the pod setting period) by using the soybean growth state data of the soybean planting area in the past years and the soybean growth state at a certain stage (for example, the middle of the growing season, the flowering period, the pod setting period) in the current growing season, and finally comparing the monitoring results with the measured results to evaluate the monitoring accuracy.

[0036] The main steps of the present application are described below:

[0037] Step 1, obtain soybean planting area data and soybean pod setting period time in previous years (at least including the previous year, usually including 3-10 years);

[0038] According to the spatial distribution map of soybean planting in previous years, the planting frequency of all soybean planting pixels (the ratio of the number of soybean planting years to the number of sampling years) is calculated. According to the rotation rule or high-frequency planting pixels, the possible soybean planting area in this year is determined, and the soybean pod setting period time in previous years is obtained for all soybean planting pixels as the basis for predicting the soybean pod setting period in this year.

[0039] Step 2, obtain the VIIRS surface reflectance data of previous years and the current year, calculate the EVI2 time series image and perform masking;

[0040] In order to carry out remote sensing monitoring of soybean planting, we need to obtain all the VIIRS surface reflectance data (such as 8-day composite data product VNP09H1) of previous years and the current year. Then, according to the spatial distribution data of soybean in previous years, the EVI2 time series image of the current year is masked and processed to filter out non-soybean planting areas, so as to obtain more accurate remote sensing monitoring results.

[0041] Step 3, obtain the average time series curve (referred to as reference curve or standard curve) of previous years;

[0042] In order to carry out remote sensing monitoring of soybean growth in the study area, the EVI2 time series image data set of previous years in the study area is calculated. Specifically, we need to calculate the mean value of the EVI2 time series data set of each soybean planting pixel in the study area to obtain the average level of soybean growth, so as to obtain the reference EVI2 time series curve of the entire study area. Then, we need to perform masking to filter out all the blocks in the study area that do not plant soybeans, and obtain more accurate remote sensing monitoring results.

[0043] Step 4, match and extract the soybean pod setting period;

[0044] In order to monitor the pod setting period of each soybean planting pixel, we can match the available EVI2 time series data for the current year. Specifically, after each soybean planting pixel obtains new EVI2 time series data, the current time series curve of the soybean planting pixel (target curve) is matched with the reference curve of the study area, and by comparing the trend and change of vegetation growth, the possible pod setting period of the soybean planting pixel in the current year is determined. This realizes accurate monitoring of soybeans and provides important decision support for relevant personnel and industries.

[0045] Step 5, evaluation of extraction results;

[0046] In order to evaluate the matching accuracy of soybean pod setting period monitoring, we can compare the monitoring matching results with the soybean pod setting period time obtained by ground investigation in the current year. Specifically, we can analyze and compare the differences in the three data of pod setting period (i.e. the starting day of pod setting period, the ending day of pod setting period and the total duration of pod setting period) to evaluate the deviation and accuracy of the monitored pod setting period time from the actual ground observation results. Such evaluation analysis can help us better understand the feasibility and limitations of remote sensing monitoring, and provide important reference for achieving more accurate soybean pod setting period monitoring.

[0047] The application uses a specific vegetation index, namely the dual-band enhanced vegetation index (EVI2), which is obtained by subtracting the blue band from EVI, solving the problem of some remote sensing satellites not containing the blue band. The calculation formula of EVI2 is as follows:

[0048]

[0049] where ρ NIR represents the reflectivity of near-infrared band, ρ red represents the reflectivity of red band. This index can more accurately reflect the growth state of vegetation, because it can eliminate the influence between vegetation and background, and has higher sensitivity to parameters such as chlorophyll content and vegetation coverage. Therefore, EVI2 is widely used in the fields of vegetation remote sensing monitoring and vegetation growth condition evaluation. The EVI2 used in the application has the advantages of simple calculation and convenient data acquisition, which can provide important technical support for remote sensing monitoring and precision agriculture.

[0050] The step S3, the need for the years of EVI2 time series image data set to obtain and calculate. In order to improve the calculation accuracy, the need for the study area of the years of EVI2 time series image using soybean spatial distribution data mask, screening out the soybean pure pixel (refers to only soybean planting pixel), this can improve the accuracy of subsequent processing. Then, the average value of all soybean planting pixels is calculated, and the standard curve of the soybean growth state in the years (time series curve, prior art) is generated, so that the generated vegetation index time series curve is closer to the standard curve of the soybean crop in the study area. This can more accurately monitor the soybean pod setting period time, and provide a benchmark for comparing the corresponding time series curve in subsequent years.

[0051] The step S4, we use shape model method for time series matching to monitor the soybean pod setting period time. Specifically, each soybean planting pixel after obtaining new EVI2 time series data, we need to match the current time series curve of the soybean planting pixel with the reference curve, compare the trend and change of vegetation growth, in order to predict the pod setting period time of soybean. The matching fitting process of shape model adopts the following formula, which translates and scales the soybean time series curve in the years to obtain the vegetation index time series curve of the soybean planting area in the year (this year's predicted time series curve, referred to as fitting curve):

[0052]

[0053] Wherein, The reference pod setting period before and after half the time window width after translation and scaling transformation of the soybean vegetation index time series curve in the year, g(x) represents the soybean reference vegetation index time series curve, xscale and t shift The scaling and translation parameters in the x axis (time axis) direction, respectively, p0 represents the soybean flowering period on the soybean reference curve.

[0054] Then, we use the Pearson correlation coefficient to process the correlation between the fitting curve after scaling and translation transformation and the local actual time series curve of the soybean planting area around the predicted pod setting period in the year, the specific formula is as follows:

[0055]

[0056] x∈(p0-t shift -w, p0-t shift +w)

[0057]

[0058] where x and y represent daily time series values, w represents the total length of the podding period of the reference curve, and N represents the length of the time series within the total time window. Finally, the best fitting parameter is determined according to the maximum value of the correlation, at which time the fitting effect is best. The soybean planting pixel prediction podding period extraction formula is as follows:

[0059] xscate×(x+tshift)+(1-xscale)×p0=p0

[0060] p est =p0-tshift

[0061] In the example of Nebraska, the above steps are described as follows.

[0062] In step S1, the specific process of obtaining soybean spatial distribution data and the specific time of soybean podding period of all soybean planting pixels over the years is as follows:

[0063] Download the Crop Data Layers data of Nebraska in the past 8 years from the National Agricultural Statistics Service of the US Department of Agriculture. Crop the range of Nebraska. For the cropped data, extract the soybean planting pixels (prior art) from other pixels by using a classifier, threshold method, etc. For each pixel, count the number of years of soybean planting in the past 8 years and divide the result by 8 to obtain the planting frequency. For some specific areas, the possible spatial distribution of soybean planting this year can be determined according to the rotation rule or high-frequency planting pixels. Select the pixels with a soybean planting frequency of more than 75%, and resample them to 500m resolution using density aggregation to obtain more accurate spatial distribution data. Obtain the crop progress report of Nebraska over the years, and extract the soybean podding period time. It can be obtained by querying the relevant database or directly contacting the local agricultural department. Integrate the soybean planting frequency data and historical soybean podding period time data obtained in the above steps to obtain the possible spatial distribution of soybean planting this year and the historical soybean podding period time.

[0064] In step S2, obtain remote sensing time series observation data.

[0065] The method for obtaining remote sensing time series observation data is as follows:

[0066] In order to carry out remote sensing monitoring of soybean planting, all available VIIRS land surface reflectance data (8-day composite data product VNP09H1) over the years and this year need to be obtained. After obtaining the VIIRS land surface reflectance data, EVI2 calculation and time series image construction are needed. EVI2 is a vegetation index that can be used to reflect the growth status of vegetation. The calculation formula of EVI2 is as follows:

[0067] EVI2 = 2.5 x ((NIR - Red) / (NIR + 2.4 x Red + 1))

[0068] Where NIR represents the reflectance of near-infrared band, Red represents the reflectance of red light band. After calculating EVI2, we can mask the previous years' EVI2 time series data according to the historical spatial distribution data of soybeans to obtain the EVI2 time series data containing only the soybean planting area. At the same time, we can also mask the current EVI2 time series image according to the soybean planting frequency distribution map to filter out non-soybean planting areas to obtain more accurate remote sensing monitoring results. After masking, we can obtain the reference EVI2 time series image dataset of the mean value in the study area, and after obtaining new data each time, we can mask according to the high-frequency distribution map of soybean planting, match the current time series curve with the reference curve to monitor the soybean pod setting period.

[0069] In step S3, the method for obtaining the average time series curve of previous years is as follows:

[0070] In order to monitor the remote sensing of soybean growth in Nebraska, we need to calculate and process the EVI2 time series image dataset of previous years. Specifically, we can first obtain the EVI2 time series data of previous years, and mask the EVI2 time series data of previous years according to the historical spatial distribution data of soybeans to obtain the EVI2 time series dataset containing only the soybean planting area. Then we need to calculate this EVI2 time series dataset to obtain the average level of soybean growth. We can use a simple method to calculate the average value of all masked EVI2 time series data, which can obtain the average level of soybean growth. According to the historical spatial distribution data of soybeans, mask the EVI2 time series data, take the average of the pixels with a planting density close to 100%, and obtain the average EVI2 time series curve. This time series curve will be used as the reference time series curve for all soybean planting pixels in the subsequent steps for matching and analysis.

[0071] In step S4, the available vegetation index time series image is matched with the historical time series curve to extract (monitor) the soybean pod setting period, and the matching method of the soybean pod setting period is as follows:

[0072] After obtaining the new EVI2 time series data of Nebraska each time, it is added to the current EVI2 time series image to establish the current EVI2 time series image. The reference vegetation index time series curve is established using historical EVI2 time series data as the growth reference curve of soybeans. The current time series curve of each soybean planting pixel is matched with the reference curve to monitor the soybean pod setting period time. Specifically, the reference time series curve needs to be translated and scaled in the x-axis direction, and then the EVI2 subsequence of a certain time range around the reference curve fall period is compared with the subsequence in the soybean planting pixel. The translation amount and scaling amount are optimized in the specific range to maximize the correlation of the two subsequences.

[0073]

[0074]

[0075] x∈(p0-t shift -w,p0-t shift +w)

[0076] After determining the two scaling parameters, the corresponding pod setting period time can be estimated as follows:

[0077]

[0078] Wherein, the reference fall period time can be obtained by observation, and the scaling parameter is obtained by optimization calculation in step S3. In specific implementation, we can set the appropriate local VI time series half-time window W width according to the growth of soybeans in Nebraska, and use the related formula for matching and estimation. In addition, we can also adjust and optimize the parameters according to the actual situation to improve the monitoring accuracy and efficiency. The formula for extracting the soybean pod setting period is as follows:

[0079] xscale×(x+tshift)+(1--xscale)×p0=p0

[0080] p est =p0-tshift

[0081] In step S5, the soybean pod setting period monitoring result is evaluated, and the evaluation method is as follows:

[0082] The evaluation method of the soybean pod setting period monitoring result is: comparing the extracted soybean pod setting period time with the ground survey result of the same year to evaluate the extraction accuracy. The ground survey is a commonly used measurement method, which usually obtains the information of crop growth and development, including the soybean pod setting period time, by observing, recording and sampling the crops in the field. Comparing the soybean pod setting period time extracted by the embodiment with the ground survey result can evaluate the accuracy, robustness and reliability of the embodiment. The error index is the root mean square error (RMSE) and the like, which can reflect the difference between the extraction results of the two methods.

[0083] The present application uses the soybean planting and growth data of previous years combined with remote sensing technology to realize accurate monitoring of the growth process of soybeans. At the same time, through the processing and matching extraction of time series curves of previous years, the pod setting period time of all soybean planting pixels can be accurately obtained, providing important support for decision-making. This technology uses the available VIIRS land surface reflectance data, saving the cost and time consumption of field investigation and manual data collection. At the same time, the present application can quickly obtain the spatial distribution and growth state of the soybean planting area, improving the efficiency and accuracy of monitoring and providing important data support for precision agriculture. At the same time, the present application can be combined with other agricultural data, such as weather, soil and water, to realize comprehensive and scientific agricultural management and decision-making, and can provide important data support for precision agriculture.

Claims

1. A method for monitoring the pod-setting stage of soybeans during the growing season based on remote sensing data, characterized in that: The method comprises the following steps: Step 1, obtaining soybean planting area data and historical soybean pod setting period time; Step 2, obtaining historical and current available VIIRS land surface reflectance data, and calculating historical and current EVI2 time series images; the EVI2 time series image calculation formula is: EVI2 = 2.5 * ((NIR-Red) / (NIR+2.4*Red+1)) Wherein, NIR represents near-infrared band reflectivity, and Red represents red light band reflectivity; Step 3, obtaining the average time series curve; The average of the EVI2 time series data set of each soybean planting pixel in the research area is taken as the average level of the whole research area soybean growth, and the reference EVI2 time series curve of the whole research area soybean is obtained. Step 4, matching and extracting soybean pod setting period; After obtaining the new EVI2 time series data of each soybean planting pixel, the current EVI2 time series curve of the soybean planting pixel is matched with the reference EVI2 time series curve, and the possible pod setting period of the soybean planting pixel in the current year is predicted by comparing the vegetation growth state; The matching and extraction of the soybean pod setting period is to match the shape of the current EVI2 time series curve of the soybean planting pixel with the reference EVI2 time series curve, and to translate and scale the reference EVI2 time series curve to the soybean fitting EVI2 time series curve of the target soybean pixel to be extracted, and the shape model matching calculation process adopts the following formula: wherein, represents the soybean fitted EVI2 time series curve after translation and scaling transformation within the reference half-time window width around the podding stage, x represents time, g(x) represents the soybean reference EVI2 time series curve, x scale and t shift respectively represent the scaling and translation parameters in the x-axis direction, p0 represents the soybean flowering stage on the soybean reference EVI2 vegetation index time series curve; Step 5, result evaluation; The matching result is compared with the soybean pod setting period time of the ground survey in the current year, the difference between the two pod setting periods is analyzed, and the deviation and accuracy of the predicted pod setting period time from the actual ground observation result are analyzed.

2. The method for monitoring the pod setting period of soybean in the growing season based on remote sensing data according to claim 1, characterized in that: The current EVI2 data is subjected to mask processing to filter out all the blocks without soybean planting.

3. The method for monitoring the pod setting period of soybean in the growing season based on remote sensing data according to claim 1, characterized in that: Pearson correlation analysis is carried out between the soybean fitting EVI2 time series curve after scaling and translation transformation and the local time series curve of the target pixel of the soybean crop to be extracted within the half time window width before and after the pod setting period of the reference curve, and the calculation formula is as follows: x e (p0-t shift -w, p0-t shift +w) Wherein, x and y represent the daily EVI2 time series value, w represents the total length of the reference curve pod setting period, N represents the time series length within the half time window range before and after the reference curve pod setting period, the maximum value of the correlation is determined to determine the best fitting parameter, and i is a positive integer from 1 to N.

Images