Phenology-adaptive corn automatic drawing method

Abstract

The invention discloses a phenological self-adaptive corn automatic mapping method, which comprises the following steps of: 1, screening to obtain Sentinel-2 time sequence satellite image data of a target area, preprocessing, and synthesizing a time sequence clear sky image data set with one scene in a half month; 2, filtering the time sequence clear sky image data set to obtain a denoised and smoothed time sequence satellite image data set; step 3, based on peak and valley values of the NDVI index of the filtered image, adaptively acquiring a corn identification key phenological period of the target area; 4, analyzing typical spectrum differences of the corn and other main background ground objects, providing a corrected red edge normalized vegetation index, and forming a typical index data set of the corn identification key phenological period with an NDVI index and an LSWI index; and step 5, automatically realizing rapid mapping of the corn in the target area by utilizing the division rule of the typical index data of the key phenological period of the corn and the main background ground feature, and realizing a corn planting distribution monitoring effect with higher precision and efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of crop remote sensing mapping, and particularly relates to a method for automatically mapping maize with phenological adaptability. Background Technique

[0002] At present, using remote sensing technology to carry out monitoring and automatic identification of crop planting distribution is an important research direction in agricultural informatization management and cultivated land resource investigation. With the enrichment of remote sensing image resources with high spatio-temporal resolution, the application of time-series remote sensing data in crop growth monitoring has been gradually popularized, especially in the remote sensing identification and mapping of major food crops such as maize, showing great potential. Traditional maize remote sensing identification methods mostly rely on single-temporal or multi-temporal combined threshold discrimination methods based on vegetation indices (such as the normalized difference vegetation index NDVI), or rely on image classification methods in specific periods.

[0003] These methods can obtain relatively ideal identification results under certain conditions, but there are also many problems: the existing methods select remote sensing images for classification based on a fixed time window, ignoring the temporal differences in the phenological stages of maize in different years and different regions. Such a static setting often leads to a decrease in identification accuracy under conditions of large climate fluctuations or significant regional differences; and in optical remote sensing images, the presence of clouds and cloud shadows is an important factor affecting identification accuracy. Existing conventional processing methods such as linear interpolation or simple maximum value synthesis are difficult to effectively restore the continuous vegetation index change curve in cloudy areas, affecting the accurate extraction of phenological information. Moreover, the expression ability of vegetation indices is also limited. Although existing indices such as NDVI and LSWI are the most widely used vegetation indices, in the case of spectral confusion between maize and background ground objects such as water bodies, forests, and other crops, the discrimination ability of existing indices is limited and it is difficult to meet the requirements of high-precision classification; at the same time, most existing methods rely on manually setting feature extraction rules, and it is difficult to dynamically adjust the identification model according to the growth characteristics of regional crops, lacking the ability of automatic and standardized rapid mapping. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for automatically mapping maize with phenological adaptability. Based on the peak and valley values of the normalized difference vegetation index of the filtered image, the key phenological periods for maize identification in the target area are adaptively obtained, the typical spectral differences between maize and other main background ground objects are analyzed, a modified red-edge normalized difference vegetation index beneficial to maize identification is proposed, and the rapid mapping of maize in the target area is automatically realized by using the division rules of typical index data of maize and main background ground objects in the key phenological periods, so as to solve the problems raised in the above background technique.

[0005] To solve the above technical problems, the present invention provides the following technical solution: A method for automatically mapping maize with phenological adaptability, comprising the following steps:

[0006] Step 1: Filter and obtain Sentinel-2 time-series satellite image data of the target area, perform preprocessing, and synthesize a half-moon time-series clear sky image dataset;

[0007] Preferably, the preprocessing process includes: cloud mask processing, NDVI index calculation, and LSWI index calculation, wherein the cloud mask processing is performed using the QA band of the Sentinel-2 satellite image.

[0008] Preferably, the calculation formulas for the NDVI index and the LSWI index are:

[0009]

[0010]

[0011] Wherein, NIR is the near infrared band, RED is the red light band, SWIR1 is the shortwave infrared 1 band, NDVI and LSWI index are superimposed on the optical image as two bands, and the optical images after all cloud mask processing are synthesized using the 15-day average synthesis method. The 15-day average synthesis method is as follows: the Sentinel-2 satellite images are synthesized for 15 days to obtain a time series image.

[0012] Step 2: Filter the time series clear sky image dataset to obtain a denoised and smoothed time series satellite image dataset;

[0013] Preferably, filtering the time series clear sky image data set includes: using upper envelope constrained WS filtering, wherein the WS filtering represents each data point as a two-dimensional grid of a space-time grid, uses a Gaussian kernel function to calculate the weight of each data point in the grid, calculates the smoothing value of each data point by weighted average, selects the larger value between the original value and the value after WS filtering, and obtains a denoised and smoothed time series satellite image data set.

[0014] Step 3: Based on the peak and valley values ​​of the NDVI index of the filtered image, adaptively obtain the key phenological period for corn identification in the target area;

[0015] Preferably, adaptively acquiring corn in the target area and identifying key phenological periods include:

[0016] The minimum NDVI value that appears during the rising phase of the NDVI curve is recorded as ;

[0017] The minimum NDVI value that appears in the descending stage of the NDVI curve is recorded as ;

[0018] The maximum value of the recorded NDVI curve is ;

[0019] Normalize the NDVI values at each stage of the rising phase of the NDVI curve using the following formula:

[0020] ;

[0021] Normalize the NDVI values at each stage of the falling phase of the NDVI curve using the following formula:

[0022] ;

[0023] Set .1 as the sowing-emergence period of maize;

[0024] Set .5 as the rapid growth period of maize;

[0025] Set 1.0 and 1.0 as the heading period of maize;

[0026] Set 0.9 as the milk-ripening to full-ripening period of maize;

[0027] Set 0.2 as the maturity period of maize.

[0028] Step 4: Analyze the typical spectral differences between maize and the main background features, and propose a modified red-edge normalized difference vegetation index for maize recognition , and form a typical index dataset for the key phenological periods of maize recognition together with the NDVI index and the LSWI index;

[0029] Preferably, the calculation formula of the modified red-edge normalized difference vegetation index for maize recognition is as follows:

[0030]

[0031] In the formula, is the red-edge band near the 865 nm wavelength, is the red-edge band near the 740 nm wavelength. By normalizing the value of the red-edge band near the 865 nm wavelength and the red-edge band near the 740 nm wavelength, the red-edge band near the 865 nm wavelength is RE4, and the red-edge band near the 740 nm wavelength is RE2.

[0032] Step 5: Use the partitioning rules of the typical index data for the key phenological periods of maize and the main background features to automatically map the maize in the target area.

[0033] Preferably, automatically mapping the maize in the target area includes:

[0034] Sowing period: , ;

[0035] Jointing stage: ;

[0036] Heading stage: ;

[0037] Milky - dough stage: ;

[0038] Maturity stage: ;

[0039] Among them, is a representation of a threshold value, and the threshold value is dynamically adjusted based on the recommended value according to different regions and different recognition accuracy metrics.

[0040] Preferably, the recommended value is 0.2, the recommended value is 0.15, the recommended value is 0.15, the recommended value is 0.2, the recommended value is 0.3, the recommended value is 0.

[0041] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: The present invention proposes a phenology - adaptive automatic maize mapping technology. Based on the half - monthly synthesized time - series clear - sky remote sensing images, the upper - envelope - constrained Whittaker smoothing filter algorithm (UE - WS) is used for denoising to ensure the smoothness and integrity of index curves such as NDVI, effectively suppressing the interference of cloud cover and data loss on the remote - sensing curves; then, based on the peak - valley value changes of the NDVI curve, the key phenological periods of each ground object are automatically extracted, improving the adaptability of maize recognition to phenological differences, and constructing an index combination of NDVI, LSWI, and the innovatively designed modified red - edge normalized difference vegetation index (mNDVIre), effectively enhancing the feature separability between maize and background ground objects; finally, according to the numerical change rules of various indexes in the phenological stages, a dynamic threshold division method is adopted to achieve automatic maize recognition and rapid mapping. Compared with the traditional sample - based maize mapping technology, the present invention has the advantages of no need for samples, high automation degree, and fast extraction speed. Description of the Drawings

[0042] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0043] Figure 1 is a schematic flow chart of the method steps provided by the embodiment of the present invention;

[0044] Figure 2 is a comparison diagram of upper envelope constrained WS filtering provided by an embodiment of the present invention;

[0045] Figure 3 It is a schematic diagram of corn mapping in the target area provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the 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 without creative work are within the scope of protection of the present invention.

[0047] Embodiments of the present invention are combined Figures 1 to 3 , among which, Figure 1 As shown, Figure 1 This is a flow chart of the automatic mapping technology of corn based on phenological adaptation of the present invention. First, the Sentinel-2 time-series satellite image data of the target area is screened and obtained, and preprocessed to synthesize a half-moon time-series clear sky image data set; the time-series clear sky image data set is subjected to upper envelope constraint WS filtering to obtain a denoised and smoothed time-series satellite image data set; the key phenological period for corn identification in the target area is adaptively obtained through the NDVI index peak and valley values; the typical spectral differences between corn and other major background objects are analyzed, and the proposed modified red-edge normalized vegetation index mNDVIre is combined with the NDVI index and the normalized water body index LSWI index to form a typical index data set for corn identification in the key phenological period; the division rules of the typical index data of the key phenological period of corn and major background objects are used to automatically realize the rapid mapping of corn in the target area. That is, combined with Figure 1 This embodiment specifically provides the following technical solution: a phenological adaptive corn automatic mapping method, comprising the following steps:

[0048] Step 1: Filter and obtain Sentinel-2 time-series satellite image data of the target area, perform preprocessing, and synthesize a half-moon time-series clear sky image dataset;

[0049] In this embodiment, a Sentinel-2 satellite image of the target area is obtained, and the preprocessing includes: cloud mask, NDVI index calculation, and LSWI index calculation, wherein the cloud mask processing is performed using the QA band of the Sentinel-2 satellite image;

[0050] For example, the calculation formulas for NDVI and LSWI index are as follows:

[0051]

[0052]

[0053] In the formula, NIR is the near-infrared band, RED is the red band, and SWIR1 is the short-wave infrared 1 band. The NDVI and LSWI indices are superimposed into the optical image as two bands. For all the optical images after cloud masking, a 15-day composite method with daily mean values is used, that is, by performing a 15-day mean composite on the Sentinel-2 satellite images, a time-series image is obtained.

[0054] Step 2: Filter the time-series clear-sky image dataset to obtain a denoised and smoothed time-series satellite image dataset;

[0055] In this embodiment, the filtering method is the upper envelope-constrained WS filter. The core idea of the WS filter is to perform a weighted average on each data point, and the weight coefficient is calculated based on the time and space distances. Specifically, the WS filter algorithm represents each data point as a two-dimensional grid (i.e., a space-time grid), and uses a Gaussian kernel function to calculate the weight of each data point within the grid, and then calculates the smoothed value of each data point through weighted averaging;

[0056] Exemplarily, the upper envelope algorithm is to select the larger value between the original value and the value after WS filtering to obtain a denoised and smoothed time-series satellite image dataset. The specific curve is combined Figure 2 as shown. The original NDVI value is the original curve, the smoothed NDVI value is the WS filter curve, and the NDVI upper envelope is the upper envelope-constrained WS filter curve.

[0057] Step 3: Based on the peak and valley values of the NDVI index of the filtered image, adaptively obtain the key phenological periods for corn recognition in the target area;

[0058] In this embodiment, the time node when NDVI reaches the maximum value is defined as the heading stage of corn. The time node when the minimum NDVI appears in the rising stage of the NDVI curve represents the sowing stage of corn, and the time node when the minimum NDVI appears in the falling stage of the NDVI curve represents the harvesting stage of corn. By recording the minimum NDVI value that appears in the rising stage of the NDVI curve, it is denoted as ; Recording the minimum NDVI value that appears in the falling stage of the NDVI curve, it is denoted as ; Recording the maximum value of the NDVI curve as ;

[0059] Exemplarily, the NDVI values of each period in the rising stage are normalized using the following formula:

[0060] ;

[0061] The NDVI values in each period of the decline stage are normalized using the following formula:

[0062] ;

[0063] Set .1 as the sowing-emergence period of maize;

[0064] Set .5 as the rapid growth period (jointing stage) of maize;

[0065] Set 1.0 and 1.0 as the heading stage of maize;

[0066] Set 0.9 as the milk-ripe to full-ripe period of maize;

[0067] Set 0.2 as the mature period of maize;

[0068] In the target area, combined with the image quality, the sowing-emergence period of maize is selected as the first half of May, the jointing stage as the first and second half of June, the heading stage as the first and second half of August, the milk-ripe to full-ripe period as the first and second half of September, and the mature period as the first and second half of October.

[0069] Step 4: Analyze the typical spectral differences between maize and other main background ground objects, propose a modified red-edge normalized difference vegetation index (Modified Red-edge NDVI, mNDVIre) that is beneficial to maize recognition, and form a typical index dataset for key phenological periods of maize recognition with the NDVI index and the LSWI index;

[0070] In this embodiment, the calculation method of the modified red-edge normalized difference vegetation index (Modified Red-edge NDVI, mNDVIre) is the normalized value of the red-edge band (RE4) near the 865 nm wavelength and the red-edge band (RE2) near the 740 nm wavelength. The calculation formula is:

[0071] .

[0072] Step 5: Automatically realize the rapid mapping of maize in the target area by using the division rules of typical index data for key phenological periods of maize and main background ground objects.

[0073] In this embodiment, the method of automatic maize mapping is: sowing period ( , ), jointing stage ( ), heading stage ( ), milk-ripe to full-ripe period ( ), maturity stage ( ). Among them is a representation of a threshold value, which is based on the recommended value and can be dynamically adjusted according to different regions based on the recognition accuracy index.

[0074] Exemplarily, the recommended value is 0.2, the recommended value is 0.15, the recommended value is 0.15, the recommended value is 0.2, the recommended value is 0.3, the recommended value is 0.

[0075] Exemplarily, the specific maize mapping is as Figure 3 shown, and the accuracy verification results show that the overall accuracy of the target area reaches 91.15%, the Kappa is 0.8102, the producer accuracy is 82.02%, and the user accuracy is 94.81%.

[0076] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0077] Finally, it should be noted that: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A phenology - adaptive automatic mapping method for corn, characterized in that: The following steps are involved: Step 1: Filter and obtain Sentinel-2 time-series satellite image data of the target area, perform preprocessing, and synthesize a half-moon time-series clear sky image dataset; Step 2: Filter the time series clear sky image dataset to obtain a denoised and smoothed time series satellite image dataset; Step 3: Based on the peak and valley values ​​of the NDVI index of the filtered image, adaptively obtain the key phenological period for corn identification in the target area; Step 4: Analyze the typical spectral differences between maize and the main background features, and propose a modified red-edge normalized difference vegetation index for maize recognition , and form a typical index dataset for the key phenological periods of maize recognition together with the NDVI index and the LSWI index; Step 5: Automatically map corn in the target area using the division rules of typical index data of key phenological periods of corn and major background objects.

2. The phenology self-adaptive automatic maize mapping method according to claim 1, wherein: The preprocessing includes: cloud mask processing, NDVI index calculation, and LSWI index calculation. The cloud mask processing is performed using the QA band of the Sentinel-2 satellite image.

3. The phenology - adaptive automatic mapping method for corn according to claim 2, characterized in that: The calculation formulas for the NDVI index and the LSWI index are as follows: ； ； Wherein, NIR is the near infrared band, RED is the red light band, SWIR1 is the shortwave infrared 1 band, NDVI and LSWI index are superimposed on the optical image as two bands, and the optical images after all cloud mask processing are synthesized using the 15-day average synthesis method. The 15-day average synthesis method is as follows: the Sentinel-2 satellite images are synthesized for 15 days to obtain a time series image.

4. A phenology - adaptive automatic mapping method for corn according to claim 3, characterized in that: The filtering of the time series clear sky image dataset includes: using upper envelope constrained WS filtering, wherein the WS filtering represents each data point as a two-dimensional grid of a space-time grid, uses a Gaussian kernel function to calculate the weight of each data point in the grid, calculates the smoothing value of each data point by weighted average, selects a larger value between the original value and the value after WS filtering, and obtains a denoised and smoothed time series satellite image dataset.

5. The phenology - adaptive automatic maize mapping method according to claim 1, wherein: The adaptive acquisition of corn in the target area and identification of key phenological periods include: Record the minimum NDVI value that appears during the ascending stage of the NDVI curve as ; Record the minimum NDVI value that appears during the descending stage of the NDVI curve as ; Record the maximum value of the NDVI curve as ; The NDVI values ​​at each period of the rising phase of the NDVI curve were normalized using the following formula: ; The NDVI values ​​at each period of the NDVI curve decline were normalized using the following formula: ; Settings .1 is the period from sowing to emergence of maize; Setting .5 period is the rapid growth stage of corn; Setting 1.0 and The period of 1.0 is the heading stage of corn; Settings The period of 0.9 is the milk-ripe to full-ripe stage of corn; Setting The period of 0.2 is the mature stage of corn.

6. The phenology self - adaptive automatic mapping method for corn according to claim 1, characterized in that: The modified red-edge normalized difference vegetation index based on maize recognition The calculation formula is as follows: ； In the formula, is the red-edge band near the wavelength of 865 nm, is the red-edge band near the wavelength of 740 nm. By normalizing the red-edge band near the wavelength of 865 nm and the red-edge band near the wavelength of 740 nm, the red-edge band near the wavelength of 865 nm is RE4, and the red-edge band near the wavelength of 740 nm is RE2.

7. A phenology-adaptive automatic maize mapping method according to claim 6, characterized in that: The automatic mapping of corn in the target area includes: Sowing date: , ; Jointing stage: ; Heading date: ; Milk-ripe to full-ripe stage: ; Maturity stage: ; Among them, to represent a threshold value, which is dynamically adjusted based on a recommended value according to different regions and different recognition accuracy metrics.

8. A phenology - adaptive automatic mapping method for corn according to claim 7, characterized in that: The recommended value is 0.2, recommended value is 0.15, recommended value is 0.15, recommended value is 0.2, recommended value is 0.3, recommended value is 0.

Images