A method for constructing spectral features of early two-increases and one-decrease of winter wheat

Abstract

The application discloses a kind of winter wheat early two increases one reduces spectral feature construction method, belongs to spectral feature construction technical field, for spectral feature construction, including obtaining the multispectral remote sensing image of experimental area and pre-processing to obtain the image to be processed, the normalized difference vegetation index of tillering stage, overwintering period, green return period is calculated respectively;Two increases one reduces spectral feature is constructed, the two increases one reduces spectral feature of all ground objects in the image to be processed is calculated, and the screening threshold of the two increases one reduces spectral feature of winter wheat is determined.The new index feature of the application can obtain and identify winter wheat, can solve the problem of distinguishing winter wheat planting area from crops similar to winter wheat spectrum feature (such as garlic), so as to more effectively extract the pixel of winter wheat area.

Description

Technical Field

[0001] This invention discloses a method for constructing spectral features of early winter wheat that show two increases and one decrease, belonging to the field of spectral feature construction technology. Background Technology

[0002] The cultivation of winter wheat helps optimize agriculture, improve agricultural production efficiency, and promote sustainable agricultural development. Therefore, establishing a farmland resource survey and monitoring system with high monitoring accuracy and good spatiotemporal scalability is of great significance for the cultivation and production of winter wheat. Existing technologies for crop classification and identification based on remote sensing technology mainly fall into three categories: time-series crop remote sensing classification, which is based on the analysis of multi-temporal remote sensing data and combined with other methods to classify crops; multidimensional data fusion, which integrates multiple remote sensing data and uses statistical learning methods or machine learning algorithms to classify different crops; and spectral feature parameter methods, which extract spectral feature parameters from image data, such as vegetation indices (e.g., NDVI), soil modulation ratio (SI), and vegetation moisture index (VSWI), and use the differences in these parameters to classify and identify crops. Existing technologies have the following drawbacks: weak spectral characteristics in the early stages of winter wheat growth, resulting in weak vegetation characteristics and making early identification difficult; and the phenomenon of similar spectral characteristics between different crops, where winter wheat and other crops may have similar characteristics in certain spectral bands, leading to spectral confusion and making accurate classification difficult. Summary of the Invention

[0003] The purpose of this invention is to provide a method for constructing spectral features of winter wheat in the early stages of growth (two increases and one decrease) to solve the problem of inaccurate classification of winter wheat in the prior art.

[0004] A method for constructing spectral features of early-stage winter wheat showing two increases and one decrease includes:

[0005] S1 obtains multispectral remote sensing images of the experimental area and performs preprocessing to obtain the image to be processed;

[0006] S2 calculates the normalized vegetation index for the tillering stage, overwintering stage, and greening stage, respectively;

[0007] S3 constructs spectral features with two increases and one decrease. ;

[0008] S4 calculates the spectral characteristics of two increases and one decrease for all ground features in the image to be processed;

[0009] S5 determines the screening threshold for the two increases and one decrease in the spectral characteristics of winter wheat.

[0010] Preprocessing in S1 includes radiometric correction, declouding, mosaicking, and clipping.

[0011] S2 includes setting the tillering period as The overwintering period is The period, the greening period is Expect, , , They are Expect, Expect, The corresponding time series dataset:

[0012] ;

[0013] ;

[0014] ;

[0015] In the formula, for End of period The value, for The beginning stage The value, for End of period The value, for The beginning stage The value, for End of period The value, for The beginning stage The value, It is the Normalized Difference Vegetation Index.

[0016] for:

[0017] ;

[0018] In the formula, This refers to the reflectance value in the near-infrared band. This represents the reflectance value in the red light band.

[0019] S3 includes:

[0020] ;

[0021] In the formula, These are the spectral characteristic coefficients.

[0022] .

[0023] S5 involves determining the screening threshold by adding or subtracting a multiple of the standard deviation from the average of the two increased and one decreased spectral characteristics of winter wheat.

[0024] S5 involves adding twice the standard deviation to the average of the two increased and one decreased spectral characteristics of winter wheat to determine the upper limit of the screening threshold.

[0025] S5 involves subtracting twice the standard deviation from the average of the two increased and one decreased spectral characteristics of winter wheat to determine the lower limit of the screening threshold.

[0026] The winter wheat in the image to be processed is determined by screening thresholds.

[0027] Compared with the prior art, the present invention has the following beneficial effects: The present invention obtains new index features for identifying winter wheat, which can solve the problem of distinguishing between crops with similar spectral features to winter wheat (such as garlic) and winter wheat planting areas, thereby extracting pixels from winter wheat areas more effectively. Attached Figure Description

[0028] Figure 1 This is a technical flowchart of the present invention;

[0029] Figure 2 It is an NDVI curve map of various land features in the study area;

[0030] Figure 3 This is a hub-and-distribution graph for verifying IDIVI;

[0031] Figure 4 This is a mapping project for winter wheat identification in Experimental Zone A;

[0032] Figure 5 This is a map for identifying winter wheat in Experimental Zone B. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention are described clearly and completely below. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0034] A method for constructing spectral features of early-stage winter wheat showing two increases and one decrease includes:

[0035] S1 obtains multispectral remote sensing images of the experimental area and performs preprocessing to obtain the image to be processed;

[0036] S2 calculates the normalized vegetation index for the tillering stage, overwintering stage, and greening stage, respectively;

[0037] S3 constructs spectral features with two increases and one decrease. ;

[0038] S4 calculates the spectral characteristics of two increases and one decrease for all ground features in the image to be processed;

[0039] S5 determines the screening threshold for the two increases and one decrease in the spectral characteristics of winter wheat.

[0040] Preprocessing in S1 includes radiometric correction, declouding, mosaicking, and clipping.

[0041] S2 includes setting the tillering period as The overwintering period is The period, the greening period is Expect, , , They are Expect, Expect, The corresponding time series dataset:

[0042] ;

[0043] ;

[0044] ;

[0045] In the formula, for End of period The value, for The beginning stage The value, for End of period The value, for The beginning stage The value, for End of period The value, for The beginning stage The value, It is the Normalized Difference Vegetation Index.

[0046] for:

[0047] ;

[0048] In the formula, This refers to the reflectance value in the near-infrared band. This represents the reflectance value in the red light band.

[0049] S3 includes:

[0050] ;

[0051] In the formula, These are the spectral characteristic coefficients.

[0052] .

[0053] S5 involves determining the screening threshold by adding or subtracting a multiple of the standard deviation from the average of the two increased and one decreased spectral characteristics of winter wheat.

[0054] S5 involves adding twice the standard deviation to the average of the two increased and one decreased spectral characteristics of winter wheat to determine the upper limit of the screening threshold.

[0055] S5 involves subtracting twice the standard deviation from the average of the two increased and one decreased spectral characteristics of winter wheat to determine the lower limit of the screening threshold.

[0056] The winter wheat in the image to be processed is determined by screening thresholds.

[0057] This invention addresses the challenge of weak and difficult-to-distinguish early-stage vegetation spectral characteristics in winter wheat. It also considers the unique spectral variation pattern of winter wheat in its early stages: an increase in spectral characteristic values ​​during the sowing-overwintering period, a decrease during the overwintering-greening period, and an increase after the greening period. The invention attempts to construct an "Increase-Decrease-Increase Vegetation Index" (IDIVI) for winter wheat and perform classification and mapping to effectively identify winter wheat planting areas in the early stages of growth. The technical process is as follows: Figure 1 As shown, remote sensing images of winter wheat at a specific growth stage were selected, and the images were preprocessed. Then, based on the VI values ​​of the remote sensing images, the vegetation index of "two increases and one decrease" was calculated, and based on the VI difference of the spectral characteristics of winter wheat, the winter wheat identification and mapping were carried out.

[0058] The required multispectral remote sensing data, including satellite / aerial photography / UAV time series data, was acquired, and processed with radiometric correction and cloud removal. The data was then mosaicked and cropped to obtain images of the study area. The image selection period is generally from mid-September to the end of March of the following year during the winter wheat growing season. Sentinel-2 satellite imagery data from September 10, 2023 to March 29, 2024 was acquired for the study.

[0059] Statistical analysis of preprocessed data from the corresponding time period in the study area revealed that the reflectance changes were most pronounced in the red and near-infrared bands among the 12 visible and infrared bands of the Sentinel-2 data used. Further band selection showed that the band combination with the best spatial resolution and the most significant changes was the red band (B4) with a wavelength center of 665 nm and the near-infrared band (B8) with a wavelength center of 842 nm. The red and near-infrared bands are commonly used to construct various vegetation indices. This invention uses the commonly used Normalized Difference Vegetation Index (NDVI) as an example to generate a time-series dataset of NDVI for each pixel within the study area, and then obtains the vegetation index curves of winter wheat in the study area, such as... Figure 2 As shown, it was found that during the study period, in the aforementioned C1, C2, and C3 periods, the NDVI values ​​of winter wheat increased in C1 and C3, while the NDVI values ​​of winter wheat decreased in C2. Figure 2 The jointing period in the example is only a portion of the jointing period of the time series studied.

[0060] Based on the obtained winter wheat index spectral curve and combined with prior phenological knowledge of winter wheat, all pixels in the study area were pre-classified, filtering out some pixels that were clearly not winter wheat, while identifying some winter wheat pixels, and the remaining pixels were used as candidate pixels for winter wheat. Figure 2 This study reveals that forest land, roads, buildings, water bodies, and garlic do not exhibit the characteristic vegetation index variations of the C1, C2, and C3 stages during the winter wheat growth period studied. A time-series dataset with normalized vegetation index differences for each pixel within the study area for these three stages is generated.

[0061] This invention uses remote sensing sentinel data from three periods—October 31, 2023 to November 30, 2023; November 30, 2023 to February 13, 2024; and February 13, 2024 to March 29, 2024—to calculate the difference in NDVI using an index. Within the study area, land cover types are categorized as: woodland, garlic, winter wheat fields, built-up land, water bodies, and roads. Training sample sets were established for these five types, and statistics were then performed. For winter wheat, sample values ​​DC1 were greater than 0, DC2 was less than 0, and DC3 was greater than 0. The sets of sample points for forests, roads, water bodies, and buildings along the DC1, DC2, and DC3 dimensions were empty. The sets of garlic sample distributions along the three dimensions intersected with those of winter wheat. Analyzing the absolute values ​​of DC1, DC2, and DC3, the values ​​for garlic were all lower than those for wheat. For other land cover types, the absolute values ​​of DC1 and DC3 were greater than those for wheat in the corresponding dimensions. However, comparing the absolute values ​​of DC2, the value for winter wheat was greater than that for other land cover samples. Therefore, to amplify the differences between winter wheat and other crops, an exponent combining the absolute values ​​of DC1, DC2, and DC3 was constructed as the basis for the spectral feature amplification. A K was introduced to further amplify the exponent values, facilitating statistical analysis.

[0062] A validation sample set was established based on the above categories to analyze the results of the "two increases and one decrease" spectral feature model. Figure 3 The "two increases and one decrease" spectral characteristic model shows that winter wheat has good distinguishability from the other five land cover types. Except for garlic, the IDIVI values ​​of water bodies, forest land, roads, and built-up land are around 0, with the largest standard deviation not exceeding 4. The mean values ​​of the four land cover samples are -4, -3, -4, and -1, respectively, with a distribution range of (-14, 3). Taking experimental area A as an example in the mapping, the average IDIVI value of the winter wheat samples is 35, the standard deviation is 10, the maximum value is 63, and the minimum value is 19. For unit data, a simple method to determine the threshold is to use the mean of the data plus or minus an appropriate multiple of the standard deviation to determine the threshold range. The threshold range corresponding to one standard deviation is (25, 45), and the threshold range corresponding to two standard deviations is (15, 65). In this study area, the threshold range corresponding to two standard deviations is optimal. The obtained regional distribution results of winter wheat in experimental area A are as follows. Figure 4 As shown. Similarly, in experimental area B, since garlic cultivation is also on a large scale, in addition to determining the threshold for winter wheat using the same method, the threshold range for garlic also needs to be considered. Ultimately, the threshold for winter wheat in this study area was determined to be (18, 59), and the threshold for garlic was determined to be (5, 13). The results of the winter wheat and garlic distribution in experimental area B are shown below. Figure 5 As shown.

[0063] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing spectral features of early-stage winter wheat showing two increases and one decrease, characterized in that, include: S1 obtains multispectral remote sensing images of the experimental area and performs preprocessing to obtain the image to be processed; S2 calculates the normalized vegetation index for the tillering stage, overwintering stage, and greening stage, respectively; S2 includes setting the tillering period as The overwintering period is The period, the greening period is Expect, , , They are Expect, Expect, The corresponding time series dataset: ； ； ； In the formula, for End of period The value, for The beginning stage The value, for End of period The value, for The beginning stage The value, for End of period The value, for The beginning stage The value, It is the normalized vegetation index; for: ； In the formula, This refers to the reflectance value in the near-infrared band. This represents the reflectance value in the red light band. S3 constructs spectral features with two increases and one decrease. ; S3 include: ； In the formula, These are spectral characteristic coefficients. ; S4 calculates the spectral characteristics of two increases and one decrease for all ground features in the image to be processed; S5 determines the screening threshold for the two increases and one decrease in the spectral characteristics of winter wheat.

2. The method for constructing spectral features of early-stage winter wheat with two increases and one decrease, as described in claim 1, is characterized in that, Preprocessing in S1 includes radiometric correction, declouding, mosaicking, and clipping.

3. The method for constructing spectral features of early-stage winter wheat with two increases and one decrease, as described in claim 2, is characterized in that... S5 involves determining the screening threshold by adding or subtracting a multiple of the standard deviation from the average of the two increased and one decreased spectral characteristics of winter wheat.

4. The method for constructing spectral features of early-stage winter wheat with two increases and one decrease, as described in claim 3, is characterized in that... S5 involves adding twice the standard deviation to the average of the two increased and one decreased spectral characteristics of winter wheat to determine the upper limit of the screening threshold.

5. The method for constructing spectral features of early-stage winter wheat with two increases and one decrease, as described in claim 4, is characterized in that... S5 involves subtracting twice the standard deviation from the average of the two increased and one decreased spectral characteristics of winter wheat to determine the lower limit of the screening threshold.

6. The method for constructing spectral features of early-stage winter wheat with two increases and one decrease, as described in claim 5, is characterized in that... The winter wheat in the image to be processed is determined by screening thresholds.

Images