Satellite remote sensing monitoring method for crop phenological stage

Abstract

The application discloses a kind of crop phenophase satellite remote sensing monitoring methods, utilize Sentinel-2 satellite remote sensing data, calculate crop from sowing to harvest period positive NDVI time series data and positive NDVI peak value, reverse NDVI time series data and reverse NDVI peak value;Set positive reverse NDVI difference threshold;In threshold range, take the time point corresponding to the maximum NDVI value as final NDVI peak point;At final NDVI peak point, positive, reverse NDVI time series data are spliced and segmented, and are reconstructed into continuous NDVI time series data;Using S-G filter smoothing, using Logistic function fitting, combine curve curvature method and dynamic threshold method to extract the key phenophase of crop.The application is characterized in that the improved maximum synthesis method is used, and the NDVI data is smoothed by combining S-G filtering, which greatly improves the accuracy of crop phenophase monitoring in the process of remote sensing detection, cloud, atmospheric interference and vegetation coverage change.

Description

Technical Field

[0001] This invention relates to the field of satellite remote sensing monitoring technology, and is particularly applicable to satellite remote sensing monitoring methods for crop phenological periods. Background Technology

[0002] In agricultural production, monitoring crop phenological stages is crucial for crop condition assessment, field management, and climate response research. Traditional phenological monitoring methods, such as field observation and accumulated temperature methods, while simple to operate, are limited to point-based observations and cannot effectively reflect growth changes over large areas of farmland, thus making them unsuitable for long-term monitoring and regional analysis. In recent years, with the rapid development of satellite remote sensing technology, using satellite remote sensing data to monitor crop growth and phenological characteristics has become a research hotspot. In particular, Sentine2 data, with its narrow band characteristics that effectively reduce the impact of water vapor absorption and high spatiotemporal resolution, has become the most widely used data source for phenological stage monitoring.

[0003] The Normalized Difference Vegetation Index (NDVI) calculated using Sentine2 data is a commonly used indicator for monitoring crop growth and photosynthetic intensity, reflecting crop greenness and growth dynamics over a large area. However, remote sensing is often affected by cloud cover, atmospheric interference, and changes in vegetation cover, leading to noise and discontinuity in NDVI time-series data, making it more difficult to extract phenological information from NDVI.

[0004] Currently, NDVI noise reduction methods mainly include: Maximum Value Combination (MVC), Median / Mean Filtering, and Optimal Exponential Slope Extraction. While the traditional Maximum Value Combination method is simple to operate, it is prone to significant monitoring errors under prolonged cloudy weather conditions. Summary of the Invention

[0005] The purpose of this invention is to provide a satellite remote sensing monitoring method for crop phenological periods, which improves the existing maximum value synthesis method and combines SG filtering to smooth NDVI data, thereby improving the accuracy of phenological period monitoring.

[0006] To achieve the above objectives, the satellite remote sensing monitoring method for crop phenological periods described in this invention includes the following steps: S1, Data Acquisition: Using Sentinel-2 satellite remote sensing data, select 10 m spatial resolution images, revisit every 5 days, with no less than 4 to 6 images per month, and perform format conversion, projection transformation and cropping in map software; S2, Positive Maximum Synthesis: Calculate the positive NDVI time series data of crops from sowing to harvest, compare the NDVI values ​​in fixed time intervals, take the larger value as the NDVI value of the current time, until the NDVI rises to the peak and remains unchanged, and record the current time point as the positive NDVI peak value; S3, Inverse Maximum Synthesis: Calculate the inverse NDVI time series data of crops from harvest to sowing, compare the NDVI values ​​in fixed time intervals, take the larger value as the NDVI value of the current time, until the NDVI rises to the peak and remains unchanged, and record the current time point as the inverse NDVI peak value; S4, NDVI Peak Detection and Difference Calculation: When the NDVI peak values ​​of the forward and reverse NDVI time series data do not overlap, calculate the NDVI difference between the forward and reverse NDVI time series data; S5, NDVI peak point determination and NDVI time series data reconstruction: Set a threshold for the NDVI difference; within the threshold range, take the time point corresponding to the maximum NDVI value as the final NDVI peak point; at the final NDVI peak point, splice and segment the positive and negative NDVI time series data to reconstruct continuous NDVI time series data. S6 uses SG filtering to smooth the reconstructed NDVI time series data, and uses the Logistic function to fit the NDVI curve. The key phenological stages of crops are extracted by combining the curve curvature method and the dynamic threshold method.

[0007] Furthermore, the SG filtering smoothing process is a Savitzky-Golay filter with 10 iterations.

[0008] Furthermore, the Logistic fitting parameters are dynamically set according to the growth characteristics of different crops in different regions.

[0009] Furthermore, the phenological period includes the seedling emergence period, jointing period, tasseling period, and maturity period.

[0010] The advantage of this invention lies in the use of an improved maximum value synthesis method, combined with SG filtering to smooth NDVI data, which greatly improves the accuracy of monitoring crop phenological stages by cloud cover, atmospheric interference, and vegetation cover changes during remote sensing. Attached Figure Description

[0011] Figure 1 is a flowchart of the method described in this invention.

[0012] Figure 2 is a time series data diagram of NDVI synthesized using traditional MVC.

[0013] Figure 3 is a diagram of the NDVI bidirectional maximum value synthesis process in the method described in this invention.

[0014] Figure 4 is a time series data diagram of summer maize NDVI after bidirectional maximum value synthesis in the method described in this invention.

[0015] Figure 5 is a schematic diagram of the point where the curvature reaches its maximum value (Logistic rising curve) on the summer maize growth curve fitted by the Logistic function.

[0016] Figure 6 is a schematic diagram of a minimum curvature point (Logistic descent curve) obtained on the summer maize growth curve fitted by the Logistic function.

[0017] Figure 7 is a schematic diagram of the rising stage of the crop growth curve, where the distance from the minimum vegetation index is 10% and 90% of the curve increase. Detailed Implementation

[0018] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] The satellite remote sensing monitoring method for crop phenological periods described in this invention, such as... Figure 1 As shown, it includes the following steps: S1. Data Acquisition and Preprocessing: Sentinel-2 satellite remote sensing data is selected as the input data source, preferably multi-period imagery with a spatial resolution of 10 meters and a revisit period of 5 days. Four to six images are acquired monthly, covering the entire crop growth cycle, such as April to October. The acquired imagery data is then processed using ENVI map software for format conversion, projection transformation, and cropping to obtain the raw NDVI (Normalized Difference Vegetation Index) time-series data. Figure 2 The image shows the original NDVI data.

[0020] S2, NDVI bidirectional maximum value synthesis. Calculate the positive NDVI time series data of crops from sowing to harvest, compare the NDVI values ​​within fixed time intervals, and take the larger value as the NDVI value at the current time. Continue until the NDVI rises to its peak and remains unchanged, then record the current time point as the positive NDVI peak value.

[0021] S3, NDVI inverse maximum value synthesis. Calculate the inverse NDVI time-series data of crops from harvest to sowing, comparing NDVI values ​​at fixed time intervals, taking the largest value as the NDVI value for the current time, until the NDVI rises to a peak and then remains constant. The current time point is then recorded as the inverse NDVI peak value. For example... Figure 3 As shown in the figure, the blue curve represents the positive NDVI time series data synthesized using the maximum value, and the yellow curve represents the inverse NDVI time series data synthesized using the maximum value. S4, NDVI Peak Detection and Difference Calculation: Ideally, the NDVI peak times of the positive and negative NDVI time-series data coincide. However, due to cloud cover and other factors, the NDVI peak times of the positive and negative NDVI time-series data intersect. When the NDVI peak times of the positive and negative NDVI time-series data do not coincide, the NDVI difference between the two data is calculated. In the above formula, The peak value of NDVI in positive NDVI time series data. The peak value of NDVI is the inverse NDVI time series data. This is the set NDVI difference threshold.

[0022] S5, NDVI Peak Point Determination and NDVI Time Series Data Reconstruction: Based on a set threshold for NDVI difference; within the threshold range, the time point corresponding to the maximum NDVI value is taken as the final NDVI peak point; at the final NDVI peak point, the positive and negative NDVI time series data are spliced ​​and segmented to reconstruct continuous NDVI time series data. For example... Figure 4 As shown in the figure, the blue curve on the left comes from positive NDVI timing data, and the yellow curve on the right comes from negative NDVI timing data.

[0023] S6 uses SG filtering to smooth the reconstructed NDVI time series data. The output value of each iteration is used as the input value of the next iteration. After 10 iterations, the error approaches zero, the curve converges, and the NDVI curve achieves a good smoothing effect.

[0024] Savitzky and Golay proposed the SG filter, a local polynomial least squares curve fitting method. The SG filter is generated by the input filter window width and polynomial degree, smoothing the data through convolution of the original sequence. The formula is as follows: in, It is the (j+1)th value in the reconstructed NDVI time series data. It is the j-th fitted value of the filtered and smoothed NDVI sequence. is the coefficient used for filtering the i-th NDVI value in the reconstructed NDVI time series data. N is the number of convolutions, which is also equal to the width of the filtering window, 2m+1. m represents the number of adjacent data points participating in the polynomial fitting calculation when smoothing a point in the reconstructed NDVI time series data.

[0025] Then, the NDVI curve was fitted using the Logistic function, and the key phenological stages of crops were extracted by combining the curve curvature method and the dynamic threshold method.

[0026] The Logistic function has an "S"-shaped curve, which aligns with the growth trend of crops. Therefore, by fitting the SG-filtered NDVI time-series data and determining the transition periods of each phenological stage of the crop on the NDVI time-series curve based on the extreme points of the rate of change of the fitted curve's curvature, the process of crop growth and change can be obtained. The Logistic function can be expressed as... Where t is time, denoted as t, where a and b are the fitting parameters to be determined in the model; d is the initial background value of the vegetation index, generally defined as the minimum stable value of a pixel without water or snow influence within a year; and c+d is the maximum value of the vegetation index.

[0027] like Figure 5 , Figure 6 As shown, on the summer maize growth curve fitted by the Logistic function, we obtain one maximum curvature point (Logistic ascending curve) and one minimum curvature point (Logistic descending curve), which correspond to the jointing stage of summer maize vegetative growth and the start time of the maturity stage of reproductive growth, respectively.

[0028] like Figure 7 As shown, during the rising phase of the crop growth curve, based on the research results of Josson and Eklundhd, the position 10% away from the minimum vegetation index (i.e., 10% of the difference between the minimum and maximum vegetation index) is defined as the beginning of vegetative growth; during the falling phase of the crop growth curve, the position 90% away from the minimum vegetation index (i.e., 90% of the difference between the minimum and maximum vegetation index) is defined as the beginning of reproductive growth.

Claims

1. A satellite remote sensing monitoring method for crop phenological stages, characterized in that, Includes the following steps: S1, Data Acquisition: Using Sentinel-2 satellite remote sensing data, select 10 m spatial resolution images, revisit every 5 days, with no less than 4 to 6 images per month, and perform format conversion, projection transformation and cropping in map software; S2, Positive Maximum Synthesis: Calculate the positive NDVI time series data of crops from sowing to harvest, compare the NDVI values ​​in fixed time intervals, take the larger value as the NDVI value of the current time, until the NDVI rises to the peak and remains unchanged, and record the current time point as the positive NDVI peak value; S3, Inverse Maximum Synthesis: Calculate the inverse NDVI time series data of crops from harvest to sowing, compare the NDVI values ​​in fixed time intervals, take the larger value as the NDVI value of the current time, until the NDVI rises to the peak and remains unchanged, and record the current time point as the inverse NDVI peak value; S4, NDVI Peak Detection and Difference Calculation: When the NDVI peak values ​​of the forward and reverse NDVI time series data do not overlap, calculate the NDVI difference between the forward and reverse NDVI time series data; S5, Determination of NDVI Peak Point and Reconstruction of NDVI Time Series Data: Set a threshold for the NDVI difference; within the threshold range, take the time point corresponding to the maximum NDVI value as the final NDVI peak point; At the final NDVI peak point, the positive and negative NDVI time series data are spliced ​​and segmented to reconstruct continuous NDVI time series data; S6 uses SG filtering to smooth the reconstructed NDVI time series data, and uses the Logistic function to fit the NDVI curve. The key phenological stages of crops are extracted by combining the curve curvature method and the dynamic threshold method.

2. The satellite remote sensing monitoring method for crop phenological periods according to claim 1, characterized in that: The SG filtering smoothing process is a Savitzky-Golay filter with 10 iterations.

3. The satellite remote sensing monitoring method for crop phenological periods according to claim 1, characterized in that: The Logistic fitting parameters are dynamically set according to the growth characteristics of different crops in different regions.

4. The satellite remote sensing monitoring method for crop phenological periods according to claim 1, characterized in that: The phenological periods include the seedling emergence period, jointing period, tasseling period, and maturity period.

Images