A method and system for automatically extracting the mildew range of winter wheat at maturity

Abstract

The present application provides a method and system for automatically extracting the mildew range of winter wheat during maturity, which belongs to the field of image recognition technology. The method realizes accurate segmentation of normal wheat areas and wheat disaster areas by calculating the mildew remote sensing index of each pixel and combining the threshold segmentation method. Among them, the mildew remote sensing index is constructed using the area method and is defined as the ratio of the first polygon area and the second polygon area. The first polygon area is the sum of the areas of the right-angled trapezoids enclosed by each band and the horizontal axis of the coordinate. The second polygon area is the area enclosed by the maximum value of the surface reflectance and the band with the shortest central wavelength, the band with the longest central wavelength, and the horizontal axis of the coordinate. The mildew remote sensing index is calculated by the ratio of the area, which can enhance the index difference between normal wheat and mildewed wheat, and is conducive to accurately determining the segmentation threshold in the threshold segmentation method, so that the threshold segmentation method can achieve an overall accuracy similar to that of the traditional machine learning method, and improve the stability and convenience of the results.

Description

Technical Field

[0001] The present application relates to the technical field of image recognition, and in particular to a method and system for automatically extracting the mildew range of winter wheat at maturity. Background Art

[0002] At present, remote sensing disaster monitoring and identification methods based on imaging technology include threshold segmentation, machine learning, deep learning, etc.

[0003] Among them, machine learning methods such as support vector machine (SVM) and random forest (RF) are widely used in the field of remote sensing classification because they can achieve higher classification accuracy than threshold segmentation methods. However, as remote sensing data become more complex and diverse, machine learning classification algorithms have the problems of complex calculations, high data processing requirements, and insufficient extraction accuracy in scenarios such as high-dimensional small samples, high-dimensional data imbalance, and feature engineering.

[0004] The idea of ​​the threshold segmentation method is to first calculate the index used to measure the spectral difference of each pixel, and then select the segmentation threshold to classify the pixels, so as to extract the disaster range. Commonly used remote sensing indices include the Normalized Difference Vegetation Index (NDVI), EVI index, HSI (Hue Saturation Intensity) color model and other traditional remote sensing indices. However, when the traditional remote sensing index is applied to the extraction of the range of wheat mold, it is not strong in distinguishing between wheat mold and normal wheat, and it is difficult to accurately determine the segmentation threshold. There is a large gap in the accuracy of the classification results compared with the classification accuracy of the machine learning method. In addition, when the threshold segmentation based on the traditional remote sensing index is applied to remote sensing data of different resolutions in the study area obtained by different types of sensors, there are problems with unstable extraction results accuracy, insufficient adaptability and migration.

[0005] Therefore, it is necessary to provide an improved technical solution to address the above-mentioned deficiencies in the prior art. Summary of the invention

[0006] The purpose of the present application is to provide a method and system for automatically extracting the mildew range of winter wheat at maturity stage, so as to solve or alleviate the problems existing in the above-mentioned prior art.

[0007] In order to achieve the above objectives, this application provides the following technical solutions:

[0008] In a first aspect, the present application provides a method for calculating a wheat mildew remote sensing index, comprising:

[0009] Acquire multispectral remote sensing images;

[0010] With wavelength as the horizontal axis and surface reflectivity as the vertical axis, plotting the surface reflectivity curve of the multispectral remote sensing image in multiple bands;

[0011] According to the interval between the central wavelengths of adjacent bands, the area of ​​the right-angled trapezoid formed by the surface reflectance curve of each band and the horizontal axis of the coordinate is calculated, and the sum is calculated, and the sum result is recorded as the first polygon area;

[0012] Determine the maximum value of the surface reflectance of all pixels in the multispectral remote sensing image, and calculate the area of ​​a rectangle enclosed by the maximum value of the surface reflectance, the central wavelength of the first band, the horizontal axis of the coordinate, and the central wavelength of the second band, and record it as the second polygonal area; the first band is the band with the shortest central wavelength among the multiple bands, and the second band is the band with the longest central wavelength among the multiple bands;

[0013] The ratio of the area of ​​the first polygon to the area of ​​the second polygon is calculated with the area of ​​the first polygon as the numerator and the area of ​​the second polygon as the denominator to obtain the mildew remote sensing index.

[0014] In combination with the first aspect, in some possible implementations, after calculating the ratio of the area of ​​the first polygon to the area of ​​the second polygon with the area of ​​the first polygon as the numerator and the area of ​​the second polygon as the denominator to obtain the mildew remote sensing index, it also includes: assuming that the intervals between the central wavelengths of adjacent bands are equal, simplifying the mildew remote sensing index to obtain a simplified mildew remote sensing index.

[0015] In combination with the first aspect, in some possible implementations, the multispectral remote sensing image includes but is not limited to: spectral information of visible light bands and near-infrared bands, and the multiple bands include but are not limited to: blue, green, red, and near-infrared bands.

[0016] In a second aspect, this embodiment provides a method for automatically extracting the mildew range of winter wheat at maturity using the wheat mildew remote sensing index calculation method provided in any of the above embodiments, comprising:

[0017] Determine the scope of target cultivated land;

[0018] Using the wheat mildew remote sensing index calculation method provided by any of the above embodiments, the mildew remote sensing index value corresponding to each pixel in the multispectral remote sensing image of the target cultivated land range is calculated; wherein the multispectral remote sensing image contains spectral information of mildewed wheat and normal wheat;

[0019] Based on the mildew remote sensing index value, a segmentation threshold is determined using an image segmentation method;

[0020] According to the segmentation threshold, the target cultivated land range is segmented into normal wheat areas and wheat disaster-affected areas to extract the wheat disaster-affected area.

[0021] In conjunction with the second aspect, in some possible implementations, based on the mildew remote sensing index value, an image segmentation method is used to determine a segmentation threshold, including:

[0022] Use multiple candidate image segmentation methods to segment the wheat mildew remote sensing index value, and determine the segmentation threshold value corresponding to each candidate image segmentation method;

[0023] The overall accuracy and Kappa coefficient are used as evaluation indicators, and the segmentation results of each candidate image segmentation method are verified through the confusion matrix to determine the optimal segmentation method; and the threshold corresponding to the optimal segmentation method is used as the segmentation threshold.

[0024] In combination with the second aspect, in some possible implementations, the multiple image segmentation methods to be selected include but are not limited to: Otsu threshold segmentation method, K-means clustering segmentation method, iterative threshold segmentation method, maximum entropy segmentation method and triangular segmentation method.

[0025] In combination with the second aspect, in some possible implementations, the optimal segmentation method is Otsu threshold segmentation method and iterative threshold segmentation method.

[0026] In a third aspect, this embodiment provides a wheat mildew remote sensing index calculation system, comprising:

[0027] a data acquisition unit configured to acquire multispectral remote sensing images;

[0028] A curve drawing unit is configured to draw a surface reflectance curve of the multispectral remote sensing image in multiple bands with wavelength as the horizontal axis and surface reflectance as the vertical axis;

[0029] A first polygon area solving unit is configured to calculate the area of ​​a right-angled trapezoid formed by the surface reflectance curve of each band and the horizontal axis of the coordinate according to the interval between the central wavelengths of adjacent bands, and sum them up, and record the summation result as the first polygon area;

[0030] The second polygon area solving unit is configured to determine the maximum surface reflectance of all pixels in the multispectral remote sensing image, and calculate the area of ​​a rectangle enclosed by the maximum surface reflectance, the central wavelength of the first band, the horizontal axis of the coordinate, and the central wavelength of the second band, which is recorded as the second polygon area; the first band is the band with the shortest central wavelength among the multiple bands, and the second band is the band with the longest central wavelength among the multiple bands;

[0031] The ratio calculation unit is configured to calculate the ratio of the first polygon area to the second polygon area with the first polygon area as the numerator and the second polygon area as the denominator to obtain the mildew remote sensing index.

[0032] In a fourth aspect, this embodiment provides a system for automatically extracting the mildew range of winter wheat at maturity, comprising:

[0033] a range determination unit configured to determine a range of target cultivated land;

[0034] An index calculation unit is configured to calculate the mildew remote sensing index value corresponding to each pixel in the multispectral remote sensing image of the target cultivated land range by using the wheat mildew remote sensing index calculation method provided in any of the above embodiments; wherein the multispectral remote sensing image contains spectral information of mildewed wheat and normal wheat;

[0035] A threshold calculation unit configured to determine a segmentation threshold using an image segmentation method based on the mildew remote sensing index value;

[0036] The disaster-affected extraction unit is configured to segment the target cultivated land range into normal wheat areas and wheat-affected areas according to the segmentation threshold, so as to extract the wheat-affected area.

[0037] In a fifth aspect, the present embodiment provides an electronic device, comprising a memory having instructions stored thereon, which, when executed on a computer, enables the computer to execute the wheat mildew remote sensing index calculation method provided in any of the above-mentioned embodiments, or to execute the automatic extraction method for the mildew range of winter wheat at maturity described in any of the embodiments.

[0038] The technical solution of the embodiment of the present application has the following beneficial effects:

[0039] The present application first provides a method for calculating the mildew remote sensing index of wheat. The method constructs the mildew remote sensing index of wheat by drawing the surface reflectance curve of spectral remote sensing images in multiple bands using the area method. The mildew remote sensing index is defined as the ratio of two areas. The first polygon area is the sum of the areas of the right-angled trapezoids enclosed by each band and the horizontal axis of the coordinate. The second polygon area is the area enclosed by the maximum surface reflectance and the band with the shortest central wavelength, the band with the longest central wavelength, and the horizontal axis of the coordinate. The method for calculating the mildew remote sensing index by the ratio of the area can effectively enhance the index difference between normal wheat and mildewed wheat, which is conducive to accurately determining the segmentation threshold in the threshold segmentation method. On this basis, the automatic extraction method for the mildew range of winter wheat at maturity provided in this embodiment calculates the mildew remote sensing index of each pixel within the target cultivated land range, combines the threshold segmentation method, and realizes the accurate segmentation of the normal wheat area and the wheat disaster area through the segmentation threshold. Experimental data show that the extraction accuracy of the automatic extraction method for the mildew range of winter wheat at maturity can achieve an effect similar to the overall accuracy of the machine learning method, and the extraction process is effectively simplified. Moreover, this method has strong applicability to the source and resolution of remote sensing images, and the deviation of the extraction result area for remote sensing data of different types and resolutions is small. Whether it is medium-resolution multispectral remote sensing data (such as Sentinel-2 and Landsat 9 satellite data) or high-resolution (such as high-resolution series satellite) multispectral remote sensing data, the extraction accuracy is relatively stable, indicating that the mildew remote sensing index has good generalization ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a schematic diagram of the principle of a method for calculating a wheat mildew remote sensing index according to some embodiments of the present application.

[0041] Figure 2 A grayscale histogram of a mildew remote sensing index provided according to some embodiments of the present application.

[0042] Figure 3 Comparison of segmentation results of different threshold segmentation methods Figure 1 .

[0043] Figure 4 Comparison of segmentation results of different threshold segmentation methods Figure 2 .

[0044] Figure 5 This is a schematic diagram comparing the classification results of the machine learning method and the classification results of the mildew remote sensing index provided in this embodiment.

[0045] Figure 6 This is an illustration of the results of extracting the mold range from different remote sensing data sources using the mold remote sensing index. Figure 1 .

[0046] Figure 7 This is an illustration of the results of extracting the mold range from different remote sensing data sources using the mold remote sensing index. Figure 2 .

[0047] Figure 8 This is a schematic diagram of the results of extracting the mold range in different regions during the same period using the mold remote sensing index. DETAILED DESCRIPTION

[0048] The terms "first", "second", "third" and "fourth" etc. in the specification and claims of the present application and the drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0049] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0050] The embodiments of the present application are described below in conjunction with the accompanying drawings.

[0051] Example 1

[0052] This embodiment provides a method for calculating a wheat mildew remote sensing index, the method comprising the following steps:

[0053] S1: Acquire multispectral remote sensing images;

[0054] S2: With wavelength as the horizontal axis and surface reflectivity as the vertical axis, draw the surface reflectivity curve of the multispectral remote sensing image in multiple bands;

[0055] S3: Calculate the area of ​​the right-angled trapezoid formed by the surface reflectance curve of each band and the horizontal axis of the coordinate according to the interval between the central wavelengths of adjacent bands, and sum them up, and record the sum result as the first polygon area;

[0056] S4: Determine the maximum value of the surface reflectance of all pixels in the multispectral remote sensing image, and calculate the area of ​​a rectangle enclosed by the maximum value of the surface reflectance, the central wavelength of the first band, the horizontal axis of the coordinate, and the central wavelength of the second band, and record it as the second polygonal area; the first band is the band with the shortest central wavelength among the multiple bands, and the second band is the band with the longest central wavelength among the multiple bands;

[0057] S5: Taking the area of ​​the first polygon as a numerator and the area of ​​the second polygon as a denominator, calculate the ratio of the area of ​​the first polygon to the area of ​​the second polygon to obtain the mildew remote sensing index.

[0058] Taking into account the accessibility of multispectral remote sensing images and the diversity of data sources, in some embodiments, the multispectral remote sensing images include but are not limited to: spectral information of visible light bands and near-infrared bands, and the multiple bands include but are not limited to: blue, green, red, and near-infrared bands.

[0059] Refer to the following Figure 1 The principle of the method for calculating the wheat mildew remote sensing index provided in this embodiment is explained.

[0060] Figure 1 The surface reflectance comparison of normal wheat and moldy wheat in the mature stage in multiple spectral bands is shown, and the reflectance mean curves of moldy wheat and normal wheat are also shown. Among them, the X-axis (i.e., the horizontal axis of the coordinate) is the wavelength, and the Y-axis represents the surface reflectance. Exemplarily, the multiple bands include: blue, green, red, and near-infrared bands. The orange dots represent the surface reflectance of the moldy wheat at the location of a certain pixel in the multispectral image, and the green dots represent the surface reflectance of normal wheat. The orange triangle and its dotted line represent the reflectance mean curve of moldy wheat. The triangle is located at the center wavelength position of different bands. The green square and dotted line represent the reflectance mean curve of normal wheat. The square is located at the center wavelength position. Between every two adjacent center wavelengths, the horizontal axis of the coordinate and the reflectance curve form an approximate right-angle trapezoid. From Figure 1 It can be seen that in all bands, the surface reflectance of normal wheat is generally higher than that of moldy wheat, and this difference increases with the increase of wavelength, which is particularly obvious in the near-infrared band. On the other hand, with the change of bands and the increase of wavelength, the surface reflectance of moldy wheat is relatively stable, while the surface reflectance of normal wheat shows an upward trend with the increase of wavelength, especially in the near-infrared band.

[0061] Analyzing its mechanism, the reason for the above characteristics is that there are differences in the distribution characteristics of moldy wheat and normal wheat in the visible light and near-infrared bands, that is, their spectral reflectance values ​​in these bands are different. Specifically, after normal wheat matures, the chlorophyll content decreases, the absorption of blue and red light decreases, the reflectance increases, and it appears yellow. The strong absorption of the red band weakens. Compared with wheat in other periods, there is no obvious trough on its surface reflectance curve. After the wheat becomes moldy, mold grows on the surface of the plant and black spots appear. At the same time, the leaves of the wheat plant gradually wither, the reflection of the spectral energy of each band decreases, and the absorption increases. Therefore, the surface reflectance of moldy wheat is generally lower than that of normal wheat. With the stable growth and reproduction of mold on the wheat plant, under the influence of mold, the moldy wheat shows indiscriminate strong absorption of the spectrum of each band (even the near-infrared band with strong reflection of healthy vegetation is no exception), resulting in its low spectral reflectance and a relatively flat spectral reflectance curve, and the difference between bands is not obvious.

[0062] Since the spectral characteristics of moldy wheat are manifested in the indiscriminate absorption of each band and the overall low reflectivity, which contrasts with the high reflectivity of normal wheat, especially the difference between the high reflectivity of normal wheat in the near-infrared band and the blue light band, while moldy wheat presents a "flat and low" spectral reflectance curve, the above characteristics can be used as a basis for distinguishing normal wheat from moldy wheat.

[0063] This embodiment uses the results of the above analysis to construct a mildew remote sensing index for mature wheat. The mildew remote sensing index is constructed using the area method and is defined as the ratio of two areas. The area is used to increase the difference in spectral characteristics between normal wheat and mildewed wheat. The reflectance difference between normal wheat and mildewed wheat in each band of multispectral images is comprehensively utilized to fully utilize the multispectral image information to enhance the spectral difference between normal wheat and mildewed wheat, making it easier to separate the two. After the wheat matures, it encounters continuous rainy weather, and the wheat surface is invaded by mold. A large number of molds grow and reproduce, resulting in a large number of black spots on the surface of the wheat.

[0064] Furthermore, the ratio of the two areas of the mildew remote sensing index is that the numerator is the area of ​​the first polygon, which is the sum of the adjacent band values ​​(i.e., reflectivity values) of each pixel, the center wavelength interval, and the area of ​​the right-angled trapezoid formed by the coordinate axis, and the denominator is the area of ​​the second polygon, which is a rectangle formed by the difference between the longest center wavelength and the shortest center wavelength of the spectral reflectivity in the multi-band remote sensing image and the maximum value of the surface reflectivity. Its function is to construct a rectangular frame (i.e. Figure 1 The red dashed box in the middle is used as a reference to provide a standardized scale for calculation, so that the index can eliminate the scale differences between bands and improve the comparability and stability of the index.

[0065] Taking multispectral remote sensing images including blue, green, red and near-infrared bands as an example, the mold remote sensing index for a certain pixel can be calculated using the following formula:

[0066]

[0067] Where: WMI is the mildew remote sensing index, R blue , R green , R red , R nir is the surface reflectivity value of blue, green, red and near infrared bands, λ blue , green , red , nir The central wavelengths corresponding to the blue, green, red, and near-infrared bands, R max is the maximum value of surface reflectivity.

[0068] It should be noted that when the traditional threshold segmentation method extracts the affected area of ​​wheat mildew, it also obtains the index value by selecting the remote sensing index and using mathematical operations between bands. However, the traditional remote sensing index has poor ability to distinguish between mildewed wheat and normal wheat:

[0069] Taking NDVI as an example, from the calculation method, only the near-infrared and red bands are considered in the calculation process of NDVI, and the information in other bands is ignored. Wheat mold is often caused by local or systematic changes caused by pathogens or other pathological factors. This change is not only reflected in the red and near-infrared bands, but also in other bands such as blue light and green light. The NDVI index only uses the difference between the near-infrared and red bands and the sum of the near-infrared and red bands as a ratio, resulting in insufficient sensitivity in identifying moldy wheat, especially early or mild moldy conditions are more difficult to fully identify. Secondly, from the design goal, the design purpose of the NDVI index is to focus on the overall vitality of vegetation, rather than directly distinguishing the pathological state of vegetation. This leads to its ability to more accurately measure the growth status and greenness of vegetation, but has limited ability to identify details such as whether the vegetation is sick, suffering from disease or moldy. For example, the reduction of NDVI may be caused only by environmental factors such as reduced vegetation cover or drought, that is, affected by background vegetation or soil, and may not accurately indicate the problem of wheat mold.

[0070] The Enhanced Vegetation Index (EVI) is another commonly used remote sensing index, which reduces the impact of atmospheric scattering by enhancing the reflectance difference between near-infrared and red light and introducing the blue light band in the calculation process. However, the main spectral characteristics of moldy wheat are the overall strong absorption of each band (including the enhanced absorption of the near-infrared band) and the flatness of the spectral reflectance curve, while EVI emphasizes the reflectance difference between near-infrared and red light. Moldy wheat will show lower reflectance in each band (especially the near-infrared band) due to mold coverage, making the reflectance difference between near-infrared and red light not obvious, and the change of EVI value is not significant, which cannot effectively identify the moldy situation. Secondly, although the EVI index introduces the blue light band, its function is to suppress the interference of atmospheric aerosols and soil background, and the weight of the blue light band is low, which cannot effectively capture the spectral characteristics of moldy wheat.

[0071] The mildew remote sensing index (WMI) constructed in this embodiment makes full use of the characteristics that wheat strongly absorbs each band spectrum indiscriminately after mildew, resulting in a large difference between it and the normal wheat reflectance spectrum, and enhances this difference by the sum of the surface reflectance of each band pixel and the area surrounded by the horizontal axis of the coordinate, and selects the rectangle formed by the maximum value of the pixel reflectance and the horizontal axis as the denominator for normalization, comprehensively considers the information of each band, and uses the concept of area to enhance the sensitivity of the index to mildewed wheat. In addition, since the mildew remote sensing index (WMI) only needs to implement area calculation to obtain the index value, it is simple to apply and intuitive to calculate, and does not require complex data pre-processing and model training. Compared with the machine learning method, it is convenient for practical application and promotion.

[0072] From the principle of constructing the mildew remote sensing index (WMI), we can know that the theoretical maximum value of the mildew remote sensing index (WMI) of wheat is 1 and the minimum value is 0. Since the reflectivity of mildew wheat in each band is lower than that of normal wheat, the mildew remote sensing index (WMI) of mildew wheat is closer to 0, while normal wheat has a stronger reflection of the spectrum and is closer to 1 on the mildew remote sensing index (WMI). Therefore, the closer the wheat mildew remote sensing index (WMI) is to 1, the lower the possibility of wheat mildew. Conversely, the closer the wheat mildew remote sensing index (WMI) is to 0, the higher the possibility of wheat mildew.

[0073] The experiment on improving the identification of moldy wheat by using the mold remote sensing index (WMI) will be discussed in detail below.

[0074] Since the center wavelength of each band is used in the mildew remote sensing index (WMI) of wheat, and the center wavelength is used as a scalar in the mildew remote sensing index, its function is to calculate the crop reflectance and the area of ​​the coordinate axis. The physical quantity that actually reflects whether the wheat is mildewed is still the spectral reflectance of the wheat. Therefore, in some embodiments, the step of simplifying the index is also included, that is, after taking the area of ​​the first polygon as the numerator and the area of ​​the second polygon as the denominator, calculating the ratio of the area of ​​the first polygon to the area of ​​the second polygon, and obtaining the mildew remote sensing index, it also includes: assuming that the intervals between the center wavelengths of adjacent bands are equal, simplifying the mildew remote sensing index, and obtaining a simplified mildew remote sensing index.

[0075] Taking multispectral remote sensing images including blue, green, red and near-infrared bands as an example, the simplified mold remote sensing index expression is as follows:

[0076]

[0077] Where: WMI1 is the simplified mildew remote sensing index, R blue , R green , R red , R nir is the surface reflectivity value of blue, green, red and near infrared bands, R max is the maximum value of surface reflectivity.

[0078] The above simplification assumes that the intervals between the central wavelengths of each band are equal. By ignoring the differences in the central wavelengths of each band of different remote sensing sensors, the simplified wheat mildew index extraction results are similar to the unsimplified extraction results, and the overall accuracy difference is only about 1%. However, the simplified mildew remote sensing index no longer needs to obtain the original central wavelengths of each band, which reduces the initial calculation cost and data processing cost in the actual use of the mildew remote sensing index, improves the efficiency of wheat mildew range identification, and makes actual use simpler and more convenient.

[0079] Example 2

[0080] This embodiment provides a method for automatically extracting the mildew range of winter wheat at maturity using the wheat mildew remote sensing index calculation method provided in any of the above embodiments, comprising the following steps:

[0081] S101: Determine the target cultivated land scope;

[0082] S102: Calculate the mildew remote sensing index value corresponding to each pixel in the multispectral remote sensing image of the target cultivated land range by using the wheat mildew remote sensing index calculation method provided in any of the above embodiments; wherein the multispectral remote sensing image contains spectral information of mildewed wheat and normal wheat;

[0083] S103: Based on the mildew remote sensing index value, determine a segmentation threshold using an image segmentation method;

[0084] S104: Segment the target cultivated land range into normal wheat areas and wheat disaster-affected areas according to the segmentation threshold, so as to extract the wheat disaster-affected area.

[0085] In the above embodiment, GIS technology or remote sensing data processing technology can be used to first extract the cultivated land range from the multispectral remote sensing image of the study area, and the non-cultivated land range does not participate in the subsequent calculation to reduce the amount of calculation. The value of the mildew remote sensing index is calculated pixel by pixel. The mildew index value of each pixel is calculated through multispectral remote sensing image data to quantify and distinguish between mildewed wheat and normal wheat.

[0086] Among them, multispectral images contain data from multiple bands (such as blue, green, red, near-infrared, etc.), which can capture the different spectral reflectance characteristics of wheat on cultivated land. Within the scope of cultivated land, these bands record the spectral information of wheat at different wavelengths, including the reflectance difference between moldy wheat and normal wheat.

[0087] Pixel is the smallest unit of remote sensing images, and each pixel corresponds to a small area on the farmland. The reflectivity of each pixel in different bands can be used to calculate the value of the mildew remote sensing index.

[0088] Image segmentation is to divide an image into several regions in order to distinguish regions with different characteristics. By analyzing the distribution of the mildew index value, a segmentation algorithm (such as Otsu method, K-means clustering, etc.) is used to determine a suitable segmentation threshold, which is used to distinguish between normal wheat and wheat areas affected by mildew.

[0089] Specifically, when the mildew index value of a pixel exceeds the segmentation threshold, it is judged as a disaster-stricken area, otherwise it is judged as a normal area. The spatial distribution range of the wheat-disaster area can be extracted through the segmentation result, so as to quickly and accurately extract the disaster-stricken range of wheat and realize automatic monitoring.

[0090] Optionally, based on the mildew remote sensing index value, an image segmentation method is used to determine a segmentation threshold, including:

[0091] Use multiple candidate image segmentation methods to segment the wheat mildew remote sensing index value, and determine the segmentation threshold value corresponding to each candidate image segmentation method;

[0092] The overall accuracy and Kappa coefficient are used as evaluation indicators, and the segmentation results of each candidate image segmentation method are verified through the confusion matrix to determine the optimal segmentation method; and the threshold corresponding to the optimal segmentation method is used as the segmentation threshold.

[0093] Since different segmentation methods perform differently in dealing with spectral feature differences and noise, by using multiple segmentation methods, the effect of each method in wheat mold index segmentation can be compared, so as to select the segmentation method that most accurately reflects the moldy area, reduce mis-segmentation and missed segmentation, and improve the accuracy of the final segmentation. At the same time, by comparing multiple methods, we can find methods that perform well for a variety of mold conditions, improve the robustness of segmentation under different conditions, and make the model more universal.

[0094] The above scheme uses overall accuracy and Kappa coefficient as evaluation criteria to ensure that the selected segmentation method is effective. These indicators are based on the confusion matrix, comparing the segmentation results and the true values, reflecting the consistency and accuracy of the classification, avoiding relying solely on subjective judgment, and making the method selection more objective.

[0095] Furthermore, the multiple image segmentation methods to be selected include but are not limited to: Otsu threshold segmentation method, K-means clustering segmentation method, iterative threshold segmentation method, entropy value segmentation method and triangular segmentation method.

[0096] Furthermore, the optimal segmentation method is Otsu threshold segmentation method and iterative threshold segmentation method.

[0097] The Otsu threshold segmentation method is particularly suitable for binary classification problems, that is, dividing images into two categories (such as moldy wheat and normal wheat), while the iterative threshold method determines the optimal threshold by recursion, and can also provide good segmentation effects for mold identification tasks with obvious spectral differences. Therefore, these two methods are stable and reliable in wheat mold segmentation.

[0098] In order to verify the effectiveness of the wheat mildew remote sensing index calculation method and the winter wheat mildew range automatic extraction method provided in this application, exemplary applications and comparative experimental results are introduced below.

[0099] This embodiment selects the main wheat producing area in the mature stage with continuous rainy weather as the study area. First, the multispectral image data of the study area is collected by the high-resolution (GF for short) satellite, the spectral characteristics of moldy wheat and normal wheat are analyzed, the moldy remote sensing index value is calculated, and the normal wheat and the moldy wheat are automatically segmented by five segmentation methods such as Otsu threshold and cluster segmentation. Secondly, as a comparison, this embodiment also uses traditional machine learning classification methods such as random forest (RF) and support vector machine (SVM), and the classification samples obtained by measured data are used to classify and extract normal wheat and moldy wheat. Finally, the extraction results of the wheat mold remote sensing index value are compared and analyzed with the traditional machine learning classification results to verify the feasibility and effectiveness of the provided wheat mold remote sensing index calculation method and the automatic extraction method of the moldy range of winter wheat during maturity.

[0100] Specifically, the experiment used a scene of GF-2 PMS remote sensing image data on June 1, 2023 and a scene of GF-6 remote sensing image data on April 18, 2023. The GF-2 and GF-6 remote sensing multispectral images were preprocessed by radiometric calibration, atmospheric correction, image registration, etc. Since the GF-2 images have higher resolution and spectral information, they are used to calculate the mold remote sensing index and extract moldy wheat using machine learning methods.

[0101] When determining the scope of the target cultivated land, in order to extract the moldy wheat information more accurately, the normalized difference vegetation index (NDVI) was calculated using the Gaofen-2 and Gaofen-6 remote sensing images of wheat growth periods at different times to set a reasonable threshold, extract the wheat cultivated land plot information, and then extract the range of moldy wheat within the cultivated land.

[0102] In order to verify the transferability and stability of the mildew remote sensing index, the experiment also used sensor data from different remote sensing platforms in the same area, specifically the Sentinel-2 and Landsat9 remote sensing images on June 1, 2023. The Sentinel-2 and Landsat multispectral remote sensing image data are from Google Earth Engine (GEE), and the data has undergone preprocessing operations such as radiation calibration, orthorectification, image registration, and atmospheric correction.

[0103] After calculating the mold remote sensing index value, since the mold remote sensing index further improves the distinguishability between categories by expanding the difference between moldy wheat and normal wheat, the grayscale histogram of the mold remote sensing index can be used to determine the segmentation threshold, such as Figure 2 As shown in the figure, normal wheat and moldy wheat form two peaks on the grayscale histogram respectively. Using the grayscale histogram to assist in determining the segmentation threshold can avoid the influence of subjective factors on the threshold selection.

[0104] In order to find the optimal segmentation method, five segmentation methods, including Otsu Threshold Segmentation, K-means Clustering Segmentation, Iterative Threshold Segmentation, Maximum Entropy Segmentation, and Triangle Segmentation, were used to perform binary automatic threshold segmentation on moldy wheat and normal wheat to obtain the range of moldy wheat. Among them, the principle of Otsu Threshold Segmentation method is to divide the image into multiple gray levels after graying, calculate the probability of each gray level respectively, and finally calculate and find the gray level corresponding to the maximum inter-class variance. This gray level makes the difference between the foreground and background of the image the largest, and the effect of distinguishing the foreground and background is the best. The calculation is simple and is not affected by the brightness and contrast of the image. The K-means clustering segmentation algorithm is based on Euclidean distance, which divides the close distance into one category, recalculates the distance mean and selects the mean as the new seed point, and repeatedly iterates and corrects the cluster center until the criterion function reaches the optimal value to obtain the optimal segmentation threshold. The basic idea of ​​the iterative threshold segmentation method is to obtain the segmentation threshold by continuously iterating and judging whether the mean of the average grayscale values ​​of the foreground and background is equal to the threshold before the iteration or the threshold change between the last two iterations is less than the set value. The iterative threshold segmentation method has high practicality and robustness, and can effectively deal with image segmentation problems under complex and diverse backgrounds. The maximum entropy segmentation algorithm uses the maximum information entropy of the image to distinguish the foreground from the background. The triangular segmentation algorithm assumes that the image grayscale histogram has only one peak (single-peak histogram). At this time, it is only necessary to find the maximum vertical distance from the straight line between the highest point of the grayscale histogram and the leftmost point or the rightmost point to each point on the histogram. The grayscale value of this point is the segmentation threshold. The triangular threshold segmentation has a good segmentation effect on image types with obvious single peaks in the grayscale histogram.

[0105] Two sub-areas A and B were selected from the study area to compare the accuracy of the analysis results. Five segmentation methods were used for segmentation. The spatial distribution of the segmentation results is shown in the figure. Figure 3 , Figure 4As shown, (a) is the value of the wheat mildew remote sensing index (WMI), expressed in grayscale values, and (b) is the segmentation result of the Otsu threshold segmentation method (yellow indicates that the identification result is mildewed wheat, and green indicates that the identification result is normal wheat, the same below). (c) is the segmentation result of the K-means clustering segmentation method. (d) is the segmentation result of the iterative threshold segmentation method. (e) is the segmentation result of the maximum entropy segmentation method. (f) is the segmentation result of the triangular segmentation method. The overall accuracy and Kappa coefficient were used to evaluate the accuracy of the above five segmentation methods. The overall accuracy of Otsu threshold segmentation was 93.62%, and the Kappa coefficient was 0.857. The overall accuracy of K-means clustering algorithm was 93.05%, and the Kappa coefficient was 0.843. The overall accuracy of iterative threshold segmentation was 93.37%, and the Kappa coefficient was 0.8509. The overall accuracy of maximum entropy segmentation was 65.53%, and the Kappa coefficient was 0.0506. The overall accuracy of triangular segmentation was 34.84%, and the Kappa coefficient was 0. Among the five segmentation algorithms, the overall accuracy and Kappa coefficient of Otsu threshold segmentation and iterative threshold segmentation were both high. The overall accuracy of both methods was above 93%, and the Kappa coefficient was above 0.84. Based on the above results, and considering the problem of simple and efficient calculation, Otsu threshold segmentation and iterative threshold segmentation were determined as the optimal segmentation methods.

[0106] Two classic machine learning classification methods, random forest and support vector machine, were used to extract moldy wheat in the study area to compare and verify the performance of the proposed wheat mold remote sensing index in extracting moldy wheat. In terms of feature selection, in addition to selecting four bands of blue, green, red and near-infrared, 27 index features were calculated using ENVI software. At the same time, reference was made to a variety of traditional remote sensing indices commonly used in existing wheat disease research, including six remote sensing index features: Greenness Index (GI), Triangular Vegetation Index (TVI), Normalized Pigment Chlorophyll Ratio Index (NPCI), Structural Independent Pigment Index (SIPI), Nitrogen Reflectance Index (NRI) and Green Chlorophyll Index (GCI). The Relief F algorithm was combined with the m RMR algorithm to reduce the dimension of the above 37 features, screen the optimal feature subset for the classification of moldy wheat, and finally select the top five features with the largest feature weights and input them into the random forest and support vector machine classification models for training. The training results were used for random forest and support vector machine classification. The classification results are shown in Figure 2. Figure 5 As shown, (a) is the value of the wheat mildew remote sensing index in the study area, expressed in grayscale, (b) is the random forest classification result, and (c) is the support vector machine classification result. The overall accuracy and Kappa coefficient of the random forest and support vector machine classification results were calculated respectively: the overall accuracy of the random forest classification for extracting mildewed wheat was 92.40% and the Kappa coefficient was 0.8354, and the overall accuracy of the support vector machine classification for extracting mildewed wheat was 92.40% and the Kappa coefficient was 0.8354.

[0107] It can be seen that the automatic extraction method of the moldy range of winter wheat at maturity proposed in this application can achieve an accuracy level comparable to that of random forest and support vector machine classification by combining the mold remote sensing index with the threshold segmentation method. This method does not require sample collection and a large amount of preprocessing and training processes. It only requires simple area and ratio calculations, and threshold segmentation can be used to achieve fast and accurate classification.

[0108] In terms of the mobility of remote sensing platforms, Figure 6 , Figure 7The figure is a schematic diagram of the results of extracting the mold range of different remote sensing data sources using the mold remote sensing index, where: (a) is the mold remote sensing index of Sentinel-2 image (abbreviated as S2); (b) is the segmentation result of the mold remote sensing index of Sentinel-2 image by Otsu threshold segmentation method; (c) is the segmentation result of the mold remote sensing index of Sentinel-2 image by iterative threshold segmentation method; (d) is the mold remote sensing index of Landsat-9 image (abbreviated as L9); (e) is the segmentation result of the mold remote sensing index of Landsat-9 image by Otsu threshold segmentation method; (f) is the segmentation result of the mold remote sensing index of Landsat-9 image by iterative threshold segmentation method. The accuracy evaluation of the segmentation results of the wheat mold remote sensing index of different remote sensing data sources is as follows: the overall accuracy of the segmentation results of the mold remote sensing index of Sentinel-2 image by Otsu threshold segmentation method and iterative threshold segmentation method is 92.94%, and the Kappa coefficient is 0.8461. The overall accuracy of the segmentation results of the mildew remote sensing index of Landsat-9 images using the Otsu threshold segmentation method is 91.40%, and the Kappa coefficient is 0.8151; the overall accuracy of the segmentation results of the mildew remote sensing index of Landsat-9 images using the iterative threshold segmentation method is 91.14%, and the Kappa coefficient is 0.8102. The above results show that the wheat mildew remote sensing index proposed in this application has generalization ability and good migration for different remote sensing sensors. For sensor data from different sources, the extraction accuracy of mildewed wheat is more than 91%, and the Kappa coefficient is above 0.81. It also shows that using the mildew remote sensing index as a segmentation index, the reflectance difference between normal wheat and mildewed wheat is still well preserved on medium and high resolution remote sensing images such as Sentinel-2 and Landsat-9. In other words, for this index, the difference in spectral reflectance between normal wheat and mildewed wheat is not sensitive to remote sensing platforms and remote sensing sensors, so that the index still has good adaptability and generalization ability in the face of different remote sensing image data sources, which greatly expands the application scenarios of the index, and thus can realize high-frequency wheat mildew monitoring using remote sensing images of multiple resolutions.

[0109] In terms of mobility in different study areas, Figure 8The schematic diagram is the result of extracting the moldy range of different regions in the same period using the moldy remote sensing index. Among them, (a) is the value of the moldy remote sensing index (WMI) of wheat, expressed in grayscale value, (b) is the result of the Otsu threshold segmentation, and (c) is the result of the iterative threshold segmentation method. Based on the high-resolution 2 remote sensing images of different regions in the same period, the moldy situation of wheat is monitored. The overall accuracy of the moldy wheat extraction by Otsu threshold segmentation is 91.76%, the Kappa coefficient is 0.8285, and the overall accuracy of the iterative threshold segmentation extraction is 91.91%, and the Kappa coefficient is 0.8320. Overall, the extraction results of the wheat moldy index in different regions are very close, with an overall accuracy of more than 91% and a Kappa coefficient of more than 0.82. The results show that the difference in spectral reflectance between moldy wheat and normal wheat is universal, and this difference can be captured by remote sensing sensors, and is less affected by regional factors, which further illustrates that the wheat moldy remote sensing index constructed based on the spectral difference between normal wheat and moldy wheat in this application has good applicability in different regions.

[0110] In summary, in this application, by analyzing the difference in the spectrum between moldy wheat and normal wheat, the area method is used to construct the wheat mold remote sensing index to enhance the spectral difference between normal wheat and moldy wheat, and then a variety of segmentation methods are used to achieve automatic segmentation of moldy wheat. The experimental results show that the extraction results based on the wheat mold remote sensing index are similar to the overall accuracy of the extraction results of the traditional random forest and SVM classification methods, but the extraction process is greatly simplified, and the wheat mold remote sensing index has good migration to different remote sensing sensor data, the extraction result accuracy is more stable, and it has good applicability and generalization ability.

[0111] Example 3

[0112] This embodiment provides a wheat mildew remote sensing index calculation system, comprising:

[0113] a data acquisition unit configured to acquire multispectral remote sensing images;

[0114] A curve drawing unit is configured to draw a surface reflectance curve of the multispectral remote sensing image in multiple bands with wavelength as the horizontal axis and surface reflectance as the vertical axis;

[0115] A first polygon area solving unit is configured to calculate the area of ​​a right-angled trapezoid formed by the surface reflectance curve of each band and the horizontal axis of the coordinate according to the interval between the central wavelengths of adjacent bands, and sum them up, and record the summation result as the first polygon area;

[0116] The second polygon area solving unit is configured to determine the maximum surface reflectance of all pixels in the multispectral remote sensing image, and calculate the area of ​​a rectangle enclosed by the maximum surface reflectance, the central wavelength of the first band, the horizontal axis of the coordinate, and the central wavelength of the second band, which is recorded as the second polygon area; the first band is the band with the shortest central wavelength among the multiple bands, and the second band is the band with the longest central wavelength among the multiple bands;

[0117] The ratio calculation unit is configured to calculate the ratio of the first polygon area to the second polygon area with the first polygon area as the numerator and the second polygon area as the denominator to obtain the mildew remote sensing index.

[0118] The wheat mildew remote sensing index calculation system provided in this embodiment can implement the process and steps of the wheat mildew remote sensing index calculation method provided in any of the above embodiments and achieve the same technical effects, which will not be described one by one here.

[0119] Example 4

[0120] This embodiment provides a system for automatically extracting the mildew range of winter wheat at maturity, comprising:

[0121] a range determination unit configured to determine a range of target cultivated land;

[0122] An index calculation unit is configured to calculate the mildew remote sensing index value corresponding to each pixel in the multispectral remote sensing image of the target cultivated land range by using the wheat mildew remote sensing index calculation method provided in any of the above embodiments; wherein the multispectral remote sensing image contains spectral information of mildewed wheat and normal wheat;

[0123] A threshold calculation unit configured to determine a segmentation threshold using an image segmentation method based on the mildew remote sensing index value;

[0124] The disaster-affected extraction unit is configured to segment the target cultivated land range into normal wheat areas and wheat-affected areas according to the segmentation threshold, so as to extract the wheat-affected area.

[0125] The automatic extraction system for the moldy range of winter wheat at maturity provided in this embodiment can implement the process and steps of the automatic extraction method for the moldy range of winter wheat at maturity provided in any of the above embodiments, and achieve the same technical effects, which will not be described in detail here.

[0126] Example 5

[0127] The present embodiment provides an electronic device, which includes a memory, and the memory stores instructions. When the instructions are executed on a computer, the computer executes the wheat mildew remote sensing index calculation method provided in any of the above embodiments, or executes the winter wheat maturity mildew range automatic extraction method provided in any of the above embodiments.

[0128] The electronic device of the embodiment of the present application may be, but is not limited to, a mobile terminal such as a mobile phone, a tablet computer, a handheld computer, a personal digital assistant (PDA), a smart home device such as a smart TV, a smart camera, a wearable device such as a smart bracelet, a smart watch, a smart glasses, or other computer devices such as a desktop, a laptop, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, a smart screen, etc.

[0129] The electronic device may include one or more of the following components: a processor, a computer readable storage medium (memory), a communication interface, and a communication bus. The memory may be connected to the processor via a bus. The bus may transmit data between the processor and the memory. The bus may be divided into an address bus, a data bus, a control bus, etc.

[0130] The processor may include one or more processing cores, and the processor may use various interfaces and lines to connect various parts of the entire electronic device, and execute various functions of the electronic device and process data by running or executing instructions, programs, code sets or instruction sets stored in the memory, and calling data stored in the memory. Exemplarily, the processor may include an application processor (application processor, AP), a modem processor, a CPU, a graphics processor (graphics processing unit, GPU), an image signal processor (image signal processor, ISP), a controller, a video codec, a digital signal processor (digital signal processor, DSP), a field programmable gate array (field-programmable gate array, FPGA), a programmable logic array (programmable logic array, PLA) and / or a neural network processor (neural-network processing unit, NPU), etc. Among them, the CPU mainly processes the operating system, user interface and application programs, etc.; the GPU is responsible for rendering and drawing the content to be displayed; the NPU is used to realize artificial intelligence (artificial intelligence, AI) functions; and the modem is used to process wireless communications. Different processing units can be independent devices or integrated in one or more processors. For example, the multiple processing units shown above are all integrated into one SoC, or the AP is a separate semiconductor chip and the other processing units are integrated into one SoC, which is not limited in the present application.

[0131] The memory may include a random access memory (RAM), a read-only memory (ROM), or a non-transitory computer-readable storage medium. The memory may be used to store instructions, programs, codes, code sets, or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system and instructions for at least one function, such as a three-dimensional rough fracture nonlinear flow field simulation method, a three-dimensional rough fracture long-term dissolution mechanism analysis method, etc.; the data storage area may store data created according to the use of the electronic device, such as input data for flow field simulation, etc.

[0132] In addition, those skilled in the art can understand that the structure of the electronic device shown in the above drawings does not constitute a limitation on the electronic device, and the electronic device may include more or fewer components than shown in the drawings, or combine certain components, or arrange the components differently. For example, the electronic device also includes microphones, speakers, radio frequency circuits, sensors, audio circuits, power supplies, Bluetooth modules and other components, which will not be described in detail here.

[0133] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for calculating wheat mildew remote sensing index, characterized in that: include: Acquire multispectral remote sensing images; With wavelength as the horizontal axis and surface reflectivity as the vertical axis, plotting the surface reflectivity curve of the multispectral remote sensing image in multiple bands; According to the interval between the central wavelengths of adjacent bands, the area of ​​the right-angled trapezoid formed by the surface reflectance curve of each band and the horizontal axis of the coordinate is calculated, and the sum is calculated, and the sum result is recorded as the first polygon area; Determine the maximum value of the surface reflectance of all pixels in the multispectral remote sensing image, and calculate the area of ​​a rectangle enclosed by the maximum value of the surface reflectance, the central wavelength of the first band, the horizontal axis of the coordinate, and the central wavelength of the second band, and record it as the second polygonal area; the first band is the band with the shortest central wavelength among the multiple bands, and the second band is the band with the longest central wavelength among the multiple bands; The ratio of the area of ​​the first polygon to the area of ​​the second polygon is calculated with the area of ​​the first polygon as the numerator and the area of ​​the second polygon as the denominator to obtain the mildew remote sensing index.

2. The method for calculating the wheat mildew remote sensing index according to claim 1, characterized in that: After calculating the ratio of the area of ​​the first polygon to the area of ​​the second polygon with the area of ​​the first polygon as the numerator and the area of ​​the second polygon as the denominator to obtain the mildew remote sensing index, the method further includes: assuming that the intervals between the central wavelengths of adjacent bands are equal, simplifying the mildew remote sensing index to obtain a simplified mildew remote sensing index.

3. The method for calculating the wheat mildew remote sensing index according to claim 1, characterized in that: The multispectral remote sensing image includes but is not limited to: spectral information of visible light band and near infrared band, and the multiple bands include but are not limited to: blue, green, red, and near infrared bands.

4. A method for automatically extracting the mildew range of winter wheat at maturity using the wheat mildew remote sensing index calculation method according to any one of claims 1 to 3, characterized in that: include: Determine the scope of target cultivated land; The method for calculating the wheat mildew remote sensing index according to any one of claims 1 to 3 is used to calculate the mildew remote sensing index value corresponding to each pixel in the multispectral remote sensing image of the target cultivated land range; wherein the multispectral remote sensing image contains spectral information of mildewed wheat and normal wheat; Based on the mildew remote sensing index value, a segmentation threshold is determined using an image segmentation method; According to the segmentation threshold, the target cultivated land range is segmented into normal wheat areas and wheat disaster-affected areas to extract the wheat disaster-affected area.

5. The method for automatically extracting the mildew range of winter wheat at maturity stage according to claim 4, characterized in that: Based on the mildew remote sensing index value, an image segmentation method is used to determine a segmentation threshold, including: Use multiple candidate image segmentation methods to segment the wheat mildew remote sensing index value, and determine the segmentation threshold value corresponding to each candidate image segmentation method; The overall accuracy and Kappa coefficient are used as evaluation indicators, and the segmentation results of each candidate image segmentation method are verified through the confusion matrix to determine the optimal segmentation method; and the threshold corresponding to the optimal segmentation method is used as the segmentation threshold.

6. The method for automatically extracting the mildew range of winter wheat at maturity according to claim 5, characterized in that: The multiple image segmentation methods to be selected include but are not limited to: Otsu threshold segmentation method, K-means clustering segmentation method, iterative threshold segmentation method, maximum entropy segmentation method and triangular segmentation method.

7. The method for automatically extracting the mildew range of winter wheat at maturity stage according to claim 6, characterized in that: The optimal segmentation methods are Otsu threshold segmentation method and iterative threshold segmentation method.

8. A wheat mildew remote sensing index calculation system, characterized in that: include: a data acquisition unit configured to acquire multispectral remote sensing images; A curve drawing unit is configured to draw a surface reflectance curve of the multispectral remote sensing image in multiple bands with wavelength as the horizontal axis and surface reflectance as the vertical axis; A first polygon area solving unit is configured to calculate the area of ​​a right-angled trapezoid formed by the surface reflectance curve of each band and the horizontal axis of the coordinate according to the interval between the central wavelengths of adjacent bands, and sum them up, and record the summation result as the first polygon area; The second polygon area solving unit is configured to determine the maximum surface reflectance of all pixels in the multispectral remote sensing image, and calculate the area of ​​a rectangle enclosed by the maximum surface reflectance, the central wavelength of the first band, the horizontal axis of the coordinate, and the central wavelength of the second band, which is recorded as the second polygon area; the first band is the band with the shortest central wavelength among the multiple bands, and the second band is the band with the longest central wavelength among the multiple bands; The ratio calculation unit is configured to calculate the ratio of the first polygon area to the second polygon area with the first polygon area as the numerator and the second polygon area as the denominator to obtain the mildew remote sensing index.

9. An automatic extraction system for the mildew range of winter wheat at maturity, characterized in that: include: a range determination unit configured to determine a range of target cultivated land; An index calculation unit is configured to calculate the mildew remote sensing index value corresponding to each pixel in the multispectral remote sensing image of the target cultivated land range by using the wheat mildew remote sensing index calculation method according to any one of claims 1 to 3; wherein the multispectral remote sensing image contains spectral information of mildewed wheat and normal wheat; A threshold calculation unit configured to determine a segmentation threshold using an image segmentation method based on the mold remote sensing index value; The disaster-affected extraction unit is configured to segment the target cultivated land range into normal wheat areas and wheat-affected areas according to the segmentation threshold, so as to extract the wheat-affected area.

10. An electronic device, comprising a memory, characterized in that: The memory stores instructions which, when executed on a computer, cause the computer to execute the method for calculating the wheat mildew remote sensing index according to any one of claims 1 to 3, or the method for automatically extracting the mildew range of winter wheat at maturity according to any one of claims 4 to 7.

Images