Method, system, device and medium for extracting spatial distribution information of agricultural greenhouses

Abstract

The present application relates to a method, system, device and medium for extracting the spatial distribution information of agricultural greenhouses, belonging to the technical field of agricultural remote sensing image monitoring. The extraction method includes: receiving aerospace hyperspectral remote sensing image data and performing preprocessing to obtain surface reflectance data; extracting the spectral subset in the first band interval and performing quadratic fitting to obtain the reflectance fitting coefficient, determining the region where the spectral response conforms to the first absorption feature, and generating the first mask; extracting the spectral subset in the second band interval, calculating the reflectance difference between both ends and performing threshold comparison to generate the second mask; extracting the spectral subset in the third band interval and performing cubic fitting calculation to obtain the reflectance fitting coefficient, determining the region where the spectral response conforms to the second absorption feature, and generating the third mask; performing logical fusion according to the first mask, the second mask and the third mask to obtain the extraction result of the spatial distribution information of agricultural greenhouses. The present application can efficiently and accurately extract the spatial distribution information of agricultural greenhouses.

Description

Technical Field

[0001] The present application relates to the technical field of agricultural remote sensing image monitoring, and in particular to a method, system, device and medium for extracting spatial distribution information of agricultural greenhouses. Background Art

[0002] With the rapid development of agricultural modernization, the management and monitoring of agricultural resources have become increasingly important. Especially in the fields of farmland management, environmental protection, precision agriculture, etc., quickly and accurately obtaining the spatial distribution information of agricultural facilities has become a core requirement. As an important part of modern agriculture, the accurate extraction of the spatial distribution information of agricultural greenhouses is of great significance for the optimal utilization of land resources, agricultural planning and management, disaster assessment, etc.

[0003] Remote sensing technology, especially optical remote sensing technology, has been widely used in agricultural monitoring. However, the resolution of traditional remote sensing image data is relatively low, and the band range is relatively single. Existing remote sensing technologies generally analyze remote sensing data based on fewer bands, which results in poor accuracy and timeliness in the extraction of the spatial distribution information of agricultural greenhouses and cannot make full use of the rich spectral information in spectral remote sensing data.

[0004] In recent years, with the progress of hyperspectral remote sensing technology, hyperspectral images provide higher spectral resolution and can cover a wider band range, making it possible to more finely distinguish the spectral characteristics of different ground objects. However, although hyperspectral remote sensing technology has obvious advantages in theory, in practical applications, due to the high processing complexity and large data volume of hyperspectral images, how to efficiently and accurately extract the spatial distribution information of agricultural greenhouses is still an urgent problem to be solved at present. Summary of the Invention

[0005] In order to efficiently and accurately extract the spatial distribution information of agricultural greenhouses, the present application provides a method, system, device and medium for extracting spatial distribution information of agricultural greenhouses.

[0006] In the first aspect, the present application provides a method for extracting spatial distribution information of agricultural greenhouses, adopting the following technical solution:

[0007] A method for extracting spatial distribution information of agricultural greenhouses, the extraction method comprising:

[0008] Receiving real-time collected aerospace hyperspectral remote sensing image data and performing preprocessing to obtain surface reflectance data;

[0009] Extracting a spectral subset in a first band interval from the surface reflectance data, performing quadratic fitting to obtain a reflectance fitting coefficient, and determining a region where the spectral response conforms to a first absorption characteristic through logical judgment to generate a first mask;

[0010] Extract a spectral subset in the second band interval from the surface reflectance data, calculate the reflectance difference at both ends of the second band interval and perform a threshold comparison to generate a second mask;

[0011] Extract a spectral subset in the third band interval from the surface reflectance data, perform a cubic fitting calculation to obtain a reflectance fitting coefficient, combine logical judgment to determine the region where the spectral response conforms to the second absorption feature, and generate a third mask;

[0012] Perform logical fusion based on the first mask, the second mask and the third mask to obtain the extraction result of the agricultural greenhouse spatial distribution information.

[0013] By adopting the above technical solution, comprehensively applying hyperspectral remote sensing data in multiple bands and combining quadratic and cubic fitting algorithms, the accuracy and efficiency of extracting agricultural greenhouse spatial distribution information are effectively improved. Through logical judgment combined with the reflectance fitting model, accurate capture of different band characteristics is ensured, and misjudgment and redundant processing are minimized. Finally, through the mask fusion technology, multiple mask information is integrated to obtain a more accurate and comprehensive agricultural greenhouse spatial distribution map, realizing the efficient and accurate extraction of agricultural greenhouse spatial distribution information, which is especially suitable for large-scale agricultural monitoring and real-time data processing, and provides strong support for fields such as agricultural resource management, disaster assessment and monitoring.

[0014] Optionally, the step of extracting a spectral subset in the first band interval from the surface reflectance data, performing a quadratic fitting to obtain a reflectance fitting coefficient, and combining logical judgment to determine the region where the spectral response conforms to the first absorption feature and generating a first mask includes:

[0015] Intercept the spectral subset of the pixel in the first band interval from the surface reflectance data to obtain a wavelength sequence and a reflectance sequence;

[0016] Construct a quadratic fitting coefficient matrix according to the wavelength sequence and the reflectance sequence, calculate the reflectance fitting coefficient by the least squares method to obtain the quadratic fitting coefficient;

[0017] Perform logical judgment based on the quadratic fitting coefficient to obtain a first logical judgment result;

[0018] Determine the region where the spectral response conforms to the first absorption feature according to the first logical judgment result and generate a first mask.

[0019] By adopting the above technical solution, spectral features in specific bands are extracted from aerospace hyperspectral remote sensing image data, and a mathematical model between wavelength and reflectivity is established through quadratic fitting. Combining logical judgment, the position of the first absorption feature is accurately determined. This technical solution not only improves the accuracy of absorption feature extraction, but also reduces noise and false judgment through mask generation and logical judgment, and can provide reliable remote sensing data support for applications such as greenhouse spatial distribution and crop health monitoring.

[0020] Optionally, the steps of extracting a spectral subset in a second band interval from the surface reflectivity data, calculating the reflectivity difference at both ends of the second band interval and performing a threshold comparison to generate a second mask include:

[0021] Intercepting a spectral subset in the second band interval from the surface reflectivity data;

[0022] Calculating the reflectivity difference at both ends of the second band interval;

[0023] Comparing the reflectivity difference with a preset threshold, and generating a second mask according to the region positions where the reflectivity difference is greater than the preset threshold.

[0024] By adopting the above technical solution, the second mask identifies regions with significant reflectivity changes, can effectively reflect the changes in surface features, especially the spectral response of plants, provides a clear regional calibration for subsequent image processing and spatial analysis, and improves the accuracy of feature extraction and analysis.

[0025] Optionally, the steps of extracting a spectral subset in a third band interval from the surface reflectivity data, performing cubic fitting calculation to obtain reflectivity fitting coefficients, and combining logical judgment to determine regions where the spectral response conforms to a second absorption feature and generating a third mask include:

[0026] Intercepting a spectral subset of pixels in the third band interval from the surface reflectivity data to obtain a wavelength sequence and a reflectivity sequence;

[0027] Constructing a cubic fitting coefficient matrix according to the wavelength sequence and the reflectivity sequence, calculating the reflectivity fitting coefficients, and obtaining the cubic fitting coefficients;

[0028] Performing logical judgment based on the cubic fitting coefficients to obtain a second logical judgment result;

[0029] Determining regions where the spectral response conforms to the second absorption feature according to the second logical judgment result, and generating a third mask.

[0030] By adopting the above technical solution, spectral data in the wavelength range of 1700 nm to 1745 nm is extracted from the hyperspectral remote sensing image. The relationship between the reflectance and the wavelength is accurately fitted using a cubic fitting model, and regions with spectral responses conforming to the absorption characteristics at 1730 nm are selected through complex logical judgments. These regions conforming to the characteristics are marked by the third mask, providing clear data support for subsequent spatial analysis and feature extraction.

[0031] Optionally, the spectral range of the spaceborne hyperspectral remote sensing image data is 400 nm to 2500 nm, the spectral resolution is less than 8 nm, and the spatial resolution is less than 30 m.

[0032] Optionally, the first band interval is selected as 2290 nm to 2325 nm; the second band interval is selected as 680 nm to 750 nm; the third band interval is selected as 1700 nm to 1745 nm.

[0033] Optionally, the first mask, the second mask, and the third mask are all logical matrices and are consistent with the spatial dimension of the spaceborne hyperspectral remote sensing image data.

[0034] By adopting the above technical solution, it is possible to effectively extract regions conforming to specific spectral characteristics on the basis of high precision and high efficiency. The logical matrix form of the mask facilitates subsequent logical fusion operations and ensures the precise correspondence between the mask and the image data, ultimately improving the accuracy and processing speed of extracting the spatial distribution information of agricultural greenhouses.

[0035] In a second aspect, the present application provides a system for extracting the spatial distribution information of agricultural greenhouses, adopting the following technical solution:

[0036] A system for extracting the spatial distribution information of agricultural greenhouses, the extraction system includes:

[0037] A data receiving module for receiving real-time acquired spaceborne hyperspectral remote sensing image data;

[0038] A preprocessing module for preprocessing the spaceborne hyperspectral remote sensing image data to obtain surface reflectance data;

[0039] A first mask generation module for extracting a spectral subset in the first band interval from the surface reflectance data, performing quadratic fitting to obtain a reflectance fitting coefficient, and determining a region with a spectral response conforming to the first absorption characteristic through logical judgment to generate a first mask;

[0040] A second mask generation module for extracting a spectral subset in the second band interval from the surface reflectance data, calculating the reflectance difference at both ends of the second band interval and performing a threshold comparison to generate a second mask;

[0041] The third mask generation module is used to extract the spectral subset in the third band interval from the surface reflectance data, perform three - time fitting calculations to obtain the reflectance fitting coefficients, determine the region where the spectral response conforms to the second absorption feature through logical judgment, and generate the third mask.

[0042] The logical fusion module is used to perform logical fusion based on the first mask, the second mask, and the third mask to obtain the extraction result of the spatial distribution information of the agricultural greenhouses.

[0043] In a third aspect, the present application provides a computer device, adopting the following technical solution:

[0044] A computer device includes a memory, a processor, and a computer program stored on the memory. The processor executes the computer program to implement the steps of the method as described in the first aspect.

[0045] In a fourth aspect, the present application provides a computer - readable storage medium, adopting the following technical solution:

[0046] A computer - readable storage medium stores a computer program that can be loaded and executed by a processor to perform any one of the methods in the first aspect.

[0047] In summary, the present application includes at least one of the following beneficial technical effects: Through the analysis of multi - band hyperspectral data, the characteristics of plastic greenhouses can be effectively extracted and identified. By extracting the reflectance data of different bands from space - borne hyperspectral remote sensing images and performing quadratic and cubic fittings, and combining logical judgments to generate multiple masks, the absorption and reflection characteristics of the greenhouses can be accurately captured. By logically fusing these masks, the spatial distribution information of agricultural greenhouses can be efficiently and accurately extracted, reducing the influence of natural surfaces and other interferences, and improving the accuracy and efficiency of agricultural greenhouse detection. This technical solution has important practical application value in the fields of precision agriculture, land use monitoring, and environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is the first flowchart of the method for extracting the spatial distribution information of agricultural greenhouses in one embodiment of the present application.

[0049] Figure 2 is the second flowchart of the method for extracting the spatial distribution information of agricultural greenhouses in one embodiment of the present application.

[0050] Figure 3 is the third flowchart of the method for extracting the spatial distribution information of agricultural greenhouses in one embodiment of the present application.

[0051] Figure 4 is the fourth flowchart of the method for extracting the spatial distribution information of agricultural greenhouses in one embodiment of the present application. Specific Embodiments

[0052] In order to make the objectives, technical solutions, and advantages of this application clearer and more understandable, the following further elaborates on this application in conjunction with the appended Figures 1-4 drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0053] Currently, there are significant challenges in identifying agricultural greenhouses from existing medium- and low-spectral-resolution remote sensing images. Since the materials of agricultural greenhouses (such as transparent or semi-transparent plastic films) often have similar spectral characteristics to surrounding ground objects (such as bare soil, crops, etc.), it is difficult to effectively distinguish them from other ground objects in the images. In addition, existing remote sensing technologies generally analyze remote sensing data based on a small number of bands, which results in poor accuracy and timeliness in extracting the spatial distribution information of agricultural greenhouses. Especially in spectral data processing and feature extraction, the deficiencies of traditional methods lead to the inability to fully utilize the rich spectral information in hyperspectral data.

[0054] Based on this, the embodiments of this application disclose a method for extracting the spatial distribution information of agricultural greenhouses.

[0055] Referring to Figure 1 , a method for extracting the spatial distribution information of agricultural greenhouses, the extraction method includes:

[0056] Step S101: Receive the real-time acquired aerospace hyperspectral remote sensing image data and perform preprocessing to obtain surface reflectance data;

[0057] Among them, aerospace hyperspectral remote sensing images are usually obtained through satellite sensors and cover multiple bands (from 400 nanometers to 2500 nanometers). These image data contain the reflected spectral information of surface objects. In practical applications, due to the influence of the atmosphere, sensor characteristics, and other environmental factors, the original remote sensing image data often has noise, radiation errors, and atmospheric interference. Therefore, before analyzing the remote sensing images, it is first necessary to preprocess these data.

[0058] Specifically, the preprocessing steps include bad line correction, radiometric calibration, and atmospheric correction processing. Bad line correction is to remove the bad pixels or bad lines in the image, and these abnormal pixels or lines may affect subsequent analysis; radiometric calibration is to convert the radiation amount received by the sensor into the surface reflectance value, and this step is crucial for ensuring data consistency and accuracy; atmospheric correction is used to remove the influence of the atmosphere on the spectral signal and obtain the true surface reflectance value.

[0059] Step S102: Extract the spectral subset in the first band interval from the surface reflectance data, perform quadratic fitting to obtain the reflectance fitting coefficients, combine logical judgment to determine the region where the spectral response conforms to the first absorption feature, and generate the first mask;

[0060] Among them, in hyperspectral remote sensing data, the reflectance values of different bands reflect the characteristics of surface objects. Agricultural greenhouse materials (such as plastic films) have obvious absorption features in specific bands (for example, 2310 nanometers). The first absorption feature is the absorption feature at 2310 nanometers.

[0061] In some embodiments, in order to accurately extract these features, a specific band interval (for example, 2290 nanometers to 2325 nanometers) is selected from the surface reflectance data for analysis. The reflectance fitting coefficients help to identify the absorption features within a specific band range. The region where the spectral response conforms to the first absorption feature can be determined through a logical judgment formula (such as judging whether the reflectance fitting coefficients meet specific conditions). If the conditions are met, it indicates that this region belongs to the agricultural greenhouse, that is, the region where the spectral response conforms to the first absorption feature is used to locate the agricultural greenhouse area.

[0062] In remote sensing image processing, a mask is a binary image or data matrix, which is used to mark or distinguish specific regions or pixels. Each pixel point in the mask is usually marked as 1 (indicating that the pixel belongs to the region of interest) or 0 (indicating that the pixel does not belong to the region of interest). In the embodiments of this application, the pixels marked as 1 in the first mask correspond to the region where the spectral response conforms to the 2310 - nanometer absorption feature.

[0063] It can be understood that by identifying the position of the specific absorption band through the mathematical model of reflectance, due to the absorption characteristics of the agricultural greenhouse in a specific band, the reflectance value of this band will change significantly. By logical judgment, the occurrence position of this feature is determined, so as to accurately identify the agricultural greenhouse, avoiding the situation in traditional methods where it is difficult to distinguish agricultural greenhouses from other ground objects and improving the accuracy of the extraction results.

[0064] Step S103: Extract the spectral subset in the second band interval from the surface reflectance data, calculate the reflectance difference between the two ends of the second band interval and perform threshold comparison to generate the second mask;

[0065] Among them, the reflectance difference between the agricultural greenhouse and the surrounding environment can often be more obvious in certain band intervals. For example, the second band interval of 680 nanometers to 750 nanometers can be selected. This band interval is usually related to the reflection characteristics of vegetation and soil. By analyzing the reflectance of this band, the distinguishability between the greenhouse and other ground objects (such as soil, crops, etc.) can be further improved.

[0066] Specifically, by calculating the difference in reflectance at both ends of the second wavelength band interval (680 nm and 750 nm), the change in the reflectance slope can be revealed. Generally, the reflectance of agricultural greenhouses changes smoothly in this wavelength band. Therefore, the reflectance difference can be used as a feature. By presetting an appropriate threshold for comparison, the location of the greenhouse can be determined. In this embodiment, the second mask is used to identify the location of the reflectance slope feature from 680 nm to 750 nm, mainly for identifying the boundary of the agricultural greenhouse area, further enhancing the distinguishability between the greenhouse and other crops.

[0067] Step S104: Extract the spectral subset of the third wavelength band interval from the surface reflectance data, perform cubic fitting calculation to obtain the reflectance fitting coefficient, combine logical judgment to determine the area where the spectral response conforms to the second absorption feature, and generate the third mask.

[0068] In some embodiments, the third wavelength band interval can be selected as the wavelength band interval near 1730 nm (such as 1700 nm to 1745 nm), which is usually related to the absorption characteristics of agricultural greenhouse materials.

[0069] Specifically, since the spectral response of agricultural greenhouses shows a relatively complex non-linear relationship in this wavelength band interval, cubic fitting (using coefficients a, b, c, d) is used to fit the reflectance data, which can more accurately reflect the spectral characteristics of this area. Through logical judgment of the fitting results, it is determined whether the spectral response conforms to the second absorption feature (for example, the peak position is near 1730 nm). If it conforms, the pixel position is marked as 1 to generate the third mask. Therefore, in the embodiments of the present application, the pixels marked as 1 in the third mask correspond to the area where the spectral response conforms to the 1730 nm absorption feature, and are used to locate the agricultural greenhouse area.

[0070] It should be noted that the first mask, the second mask, and the third mask are all logical matrices and are consistent with the spatial dimension of the aerospace hyperspectral remote sensing image data, so as to effectively extract the area that conforms to specific spectral characteristics on the basis of high precision and high efficiency. The logical matrix form of the mask facilitates subsequent logical fusion operations and ensures the precise correspondence between the mask and the image data, ultimately improving the accuracy and processing speed of extracting the spatial distribution information of agricultural greenhouses.

[0071] Step S105: Perform logical fusion according to the first mask, the second mask, and the third mask to obtain the extraction result of the spatial distribution information of agricultural greenhouses.

[0072] Among them, the first mask, the second mask, and the third mask are logically fused. The first mask identifies the area where the spectral response conforms to the absorption characteristics at 2310 nm, the second mask identifies the reflectance slope characteristics in the range of 680 - 750 nm, and the third mask identifies the area where the spectral response conforms to the absorption characteristics at 1730 nm. By performing logical operations (such as A AND (B OR C)), the multi-mask information is integrated to generate a spatial distribution map of the greenhouse, and finally the spatial distribution information of the agricultural greenhouse is extracted.

[0073] In one embodiment of the present application, the logical fusion of A AND (B OR C) represents "the first mask AND (the second mask OR the third mask)". By screening the results of the first mask, it is ensured that only those areas that meet the characteristics of the plastic greenhouse are retained. At the same time, through the joint verification of the second mask and the third mask, the accuracy and reliability of the classification are improved.

[0074] Specifically, during the screening process, the role of the first mask is to capture the absorption characteristics of the plastic greenhouse material in the short-wave infrared region. Only when the reflectance response of a certain pixel meets the requirements at the 2310 nm absorption peak position is it considered to meet the conditions of the first mask; the second mask analyzes the spectral reflectance difference in the range of 680 - 750 nm to exclude the interference between natural surfaces (such as plants) and plastics, further confirming the area of the plastic greenhouse and reducing misjudgment; the third mask further verifies the existence of the plastic greenhouse, especially excluding interference signals with other similar absorption characteristics. When a secondary absorption peak at 1730 nm is detected, this spectral band characteristic is the absorption signal of the plastic greenhouse material in the mid-infrared region.

[0075] It should be noted that the first mask (2290 - 2325 nm) and the third mask (1700 - 1745 nm) respectively target the absorption characteristics of plastic materials in different infrared sub-regions (short-wave infrared and mid-infrared). Through the logical operation "A and (B or C)", multi-feature cross-verification is achieved to reduce misjudgment in a single band. For example, the spectral response of bare soil or buildings in this band is usually flat, lacking sharp absorption peaks, while vegetation has a high reflectance in the short-wave infrared region (such as the band of the first mask), but has no obvious absorption characteristics in the mid-infrared region (the band of the third mask), thereby reducing the false detection rate.

[0076] It can be understood that through mask fusion, the data advantages of different bands can be fully utilized, eliminating possible misidentifications in a single band, and ensuring that the extraction results are more accurate and comprehensive. The result after mask fusion can more accurately identify the distribution area of the agricultural greenhouse, providing reliable data support for applications such as agricultural resource management and disaster assessment.

[0077] In the above embodiments, by comprehensively applying hyperspectral remote sensing data in multiple bands and combining quadratic and cubic fitting algorithms, the accuracy and efficiency of extracting the spatial distribution information of agricultural greenhouses are effectively improved. Through logical judgment combined with the reflectance fitting model, the accurate capture of different band characteristics is ensured, minimizing misjudgment and redundant processing. Finally, through the mask fusion technology, multiple mask information is integrated to obtain a more accurate and comprehensive spatial distribution map of agricultural greenhouses, realizing the efficient and accurate extraction of the spatial distribution information of agricultural greenhouses, which is especially suitable for large-scale agricultural monitoring and real-time data processing, providing strong support for fields such as agricultural resource management, disaster assessment and monitoring.

[0078] In the embodiment of the present application, the spectral range of the spaceborne hyperspectral remote sensing image data is from 400 nanometers to 2500 nanometers, the spectral resolution is less than 8 nanometers, and the spatial resolution is less than 30 meters.

[0079] Among them, the spectral range from 400 nanometers to 2500 nanometers covers the spectra from visible light to short-wave infrared (SWIR) and even part of the mid-infrared (MIR) region. This spectral range can comprehensively capture various spectral characteristics of surface substances and is suitable for remote sensing monitoring of various ground objects including greenhouse materials (such as plastics), vegetation, soil, irrigation water bodies, etc.

[0080] Moreover, a spectral resolution of less than 8 nanometers means that the differences between each spectral band can be carefully distinguished, and the spectral characteristics of objects can be accurately captured. Especially in regions where the spectral bands are relatively dense, such a resolution can effectively distinguish the spectral differences of objects at close wavelengths. For example, in the short-wave infrared and mid-infrared regions, the absorption peaks of materials are relatively narrow, and an 8-nanometer resolution is sufficient to capture these absorption characteristics, thereby improving the recognition accuracy. A spatial resolution of less than 30 meters means that the ground area represented by each pixel is small, and small ground object features can be clearly distinguished, which is very important for the extraction of small-scale spatial targets such as agricultural greenhouses. Agricultural greenhouses usually have a certain area, but in large remote sensing images, their spatial dimensions are small. Therefore, a higher spatial resolution enables the clear distinction of the boundaries and shapes of greenhouses.

[0081] Referring to Figure 2 , as an implementation manner of step S102, the steps of extracting the spectral subset in the first band interval from the surface reflectance data, performing quadratic fitting to obtain the reflectance fitting coefficient, and combining logical judgment to determine the region where the spectral response conforms to the first absorption characteristic and generating the first mask include:

[0082] Step S201, intercepting the spectral subset of the pixel in the first band interval from the surface reflectance data to obtain the wavelength sequence and the reflectance sequence;

[0083] Among them, in aerospace hyperspectral remote sensing images, the data of each pixel consists of reflectance values of multiple bands. For the spectral characteristics of target areas such as agricultural greenhouses, specific band intervals are usually selected for analysis. The goal of this step is to extract spectral data within the band range (such as 2300 - 2325 nanometers) related to the first absorption feature from the hyperspectral data.

[0084] Specifically, by specifying the band interval (such as 2300 - 2325 nanometers), the reflectance values of the corresponding band are extracted from the spectral data of the overall image. These reflectance values correspond one-to-one with the wavelength sequence of this band, forming a reflectance sequence Y = [y1, y2,..., y n T and a wavelength sequence X = [x1, x2,..., x n T . The reflectance sequence represents the surface reflectance values at each corresponding wavelength, and the wavelength sequence represents the wavelength values of each band in this band interval.

[0085] Step S202, construct a quadratic fitting coefficient matrix according to the wavelength sequence and the reflectance sequence, and calculate the reflectance fitting coefficients through the least squares method to obtain the quadratic fitting coefficients;

[0086] Among them, quadratic fitting is a common mathematical modeling method used to describe the relationship between data points. Since the spectral reflectance of agricultural greenhouses usually presents a specific curve form, the trend of the reflectance of this band changing with the wavelength can be better fitted through quadratic fitting (i.e., a quadratic polynomial).

[0087] Specifically, the quadratic fitting method is used to establish a mathematical model between the wavelength sequence X and the reflectance sequence Y. The quadratic fitting coefficient matrix is , and the shape of matrix A is a matrix of (n, 3), where n is the length of the wavelength sequence.

[0088] In the above formula, represents the element obtained by multiplying the elements of the wavelength sequence X with its own elements according to the position, and X represents the wavelength sequence X itself; 1 n T is a column vector of all 1s, representing the constant term.

[0089] Furthermore, through the least squares method, the sum of the squared errors between the reflectance sequence Y and the fitting result is minimized to determine the best fitting coefficients. The specific calculation formula is: K = (A T A) -1 A T Y, and the solved K contains the three coefficients a, b, c of the quadratic fitting, that is, K = [a, b, c] T , and these coefficients define the quadratic relationship between the wavelength and the reflectance. ​​

[0090] In the above formula, A T is the transpose of matrix A, and (A T A) -1 is the inverse matrix of matrix A T A, and A T Y is the product of the transpose of matrix A and the reflectance sequence Y.

[0091] It can be understood that through quadratic fitting, the non-linear relationship between the wavelength sequence and the reflectance sequence can be accurately captured, obtaining more detailed reflectance characteristics, thereby providing a reliable mathematical model for further judging the absorption characteristics.

[0092] Step S203, perform a logical judgment based on the quadratic fitting coefficients to obtain a first logical judgment result;

[0093] Specifically, the formula for the logical judgment is:

[0094] sgn(-b / (2a)-2308)*sgn(2315+b / (2a))*a*k>threshold1;

[0095] In the above formula, sgn represents the sign function that returns the value of a number. If x is greater than zero, then sgn(x)=1; if x is less than zero, then sgn(x)=-1; if x is equal to zero, then sgn(x)=0; -b / (2a)-2308 represents the vertex position of the fitting, and through this position, it can be judged whether it is near 2310 nm (2310 nm is a typical wavelength of this absorption feature); 2315+b / (2a) is used to judge whether the fitting vertex is near 2315 nm; a and b are the quadratic fitting coefficients; k is the normalization factor, which represents the ratio of the reflectance of this pixel in this band interval to the reflectance of the entire scene image. For example , M 2290-2325 is the average value of the reflectance sequence Y of this pixel, and TM 2290-2325 is the average reflectance value of all pixels in the entire scene image in the 2290 nm - 2325 nm interval; threshold1 represents the final judgment of whether it meets the specified threshold. If the condition is met, it means that this area conforms to the first absorption feature.

[0096] It should be noted that the coefficients a and b play a decisive role in the shape and change of the signal curve, and the coefficient c is a constant term, mainly affecting the curve translation. Therefore, it is not directly used for logical judgment in the subsequent steps.

[0097] Step S204, determine the area where the spectral response conforms to the first absorption feature according to the first logical judgment result, and generate a first mask.

[0098] Among them, based on the logical judgment result, it is determined pixel by pixel whether it conforms to the first absorption feature. If the logical judgment result of a certain pixel is true (i.e., the condition is satisfied), the position of this pixel is marked as 1 in the first mask; otherwise, it is marked as 0. The finally generated first mask represents the positions of all pixels in the image that conform to the first absorption feature.

[0099] It can be understood that the generation of the first mask is a calibration of the spatial area of the agricultural greenhouse, marking the area in the image that conforms to the first absorption feature and providing accurate basic data for subsequent spatial distribution analysis.

[0100] In the above-mentioned embodiment, spectral features in specific bands are extracted from the aerospace hyperspectral remote sensing image data, and a mathematical model between wavelength and reflectance is established through quadratic fitting. The position of the first absorption feature is accurately determined by combining logical judgment. This technical solution not only improves the accuracy of absorption feature extraction, but also reduces noise and misjudgment through mask generation and logical judgment, and can provide reliable remote sensing data support for applications such as greenhouse spatial distribution and crop health monitoring.

[0101] Refer to Figure 3 , as an implementation manner of step S103, the steps of extracting a spectral subset in the second band interval from the surface reflectance data, calculating the reflectance difference between both ends of the second band interval and performing threshold comparison, and generating the second mask include:

[0102] Step S301, intercepting a spectral subset in the second band interval from the surface reflectance data;

[0103] Among them, the second band interval is from 680 nanometers to 750 nanometers. This band is usually related to the chlorophyll absorption characteristics of plants and can reflect plant health conditions or other surface characteristics to further improve the distinguishability between the greenhouse and other ground objects.

[0104] Step S302, calculating the reflectance difference between both ends of the second band interval;

[0105] Specifically, the reflectance difference T is the difference between the reflectance values at both ends of this band (i.e., the 680 - nanometer and 750 - nanometer bands). , by calculating the reflectance difference T, the change in reflectance at both ends of this band interval can be quantitatively reflected. If the reflectance difference is large, it may mean that the spectral response in this area has a significant change in this band, which may be related to specific ground object characteristics (such as vegetation type or health condition).

[0106] Step S303, comparing the reflectance difference with a preset threshold, and generating the second mask according to the positions of the areas where the reflectance difference is greater than the preset threshold.

[0107] Among them, the preset threshold is determined based on experience or previous research, aiming to distinguish regions with significant changes from those without significant changes, so as to help filter out regions with small spectral changes and focus on regions that may have special surface features or environmental characteristics.

[0108] Specifically, the calculated reflectivity difference T is compared with a preset threshold threshold2. The second mask is a logical matrix, and the value of each pixel can be 0 or 1. If the reflectivity difference T is greater than the preset threshold threshold2, the corresponding position of the pixel in the mask is assigned a value of 1, indicating that the spectral response at this position meets specific conditions; otherwise, it is assigned a value of 0.

[0109] In the above embodiment, the second mask identifies regions with significant reflectivity changes, can effectively reflect changes in surface features, especially the spectral response of plants, provides a clear regional calibration for subsequent image processing and spatial analysis, and improves the accuracy of feature extraction and analysis.

[0110] Refer to Figure 4 , as an embodiment of step S104, the steps of extracting the spectral subset in the third band interval from the surface reflectivity data, performing cubic fitting calculations to obtain the reflectivity fitting coefficients, and determining the region where the spectral response conforms to the second absorption feature through logical judgment to generate the third mask include:

[0111] Step S401, intercept the spectral subset of the pixel in the third band interval from the surface reflectivity data to obtain the wavelength sequence and the reflectivity sequence;

[0112] Among them, since plastic greenhouse covering materials (such as polyethylene, polyvinyl chloride, etc.) exhibit significant spectral absorption characteristics near the wavelength of 1730 nanometers, the spectral data in the band interval near 1730 nanometers (1700 nanometers to 1745 nanometers) is selected for the third band interval to form a high contrast for accurate detection.

[0113] Specifically, the reflectivity values in the corresponding band (1700 nanometers to 1745 nanometers) are extracted from the spectral data of the overall image. These reflectivity values correspond one by one to the wavelength sequence of this band to form the reflectivity sequence Y = [y1, y2,..., y n T and the wavelength sequence X = [x1, x2,..., x n T . The reflectivity sequence represents the surface reflectivity values at each corresponding wavelength, and the wavelength sequence represents the wavelength values of each band in this band interval.

[0114] Step S402, construct a cubic fitting coefficient matrix according to the wavelength sequence and the reflectivity sequence, calculate the reflectivity fitting coefficients, and obtain the cubic fitting coefficients;​​

[0115] Among them, cubic fitting is used to accurately describe the trend of reflectance changing with wavelength within this wavelength range. By performing cubic polynomial fitting on the extracted wavelength sequence and reflectance sequence, the fitting curve of spectral data is obtained.

[0116] Specifically, when performing fitting, the cubic fitting coefficient matrix A is constructed as follows: ;

[0117] In the above formula, represents the element obtained by multiplying the elements of the wavelength sequence X with themselves according to the position. X represents the wavelength sequence X itself, and 1 n T is a column vector of all 1s, representing the constant term.

[0118] Furthermore, the fitting coefficients are solved by the least squares method. The specific calculation formula is: K = (A T A) -1 A T Y. The solved K contains the three coefficients a, b, c, and d of the quadratic fitting, that is, K = [a, b, c, d] T , and these coefficients define the cubic relationship between wavelength and reflectance.

[0119] In the above formula, A T is the transpose of matrix A, (A T A) -1 is the inverse matrix of matrix A T A, and A T Y is the product of the transpose of matrix A and the reflectance sequence Y.

[0120] It can be understood that cubic fitting can accurately capture the complex non-linear relationship between reflectance and wavelength within the wavelength band interval, and the fitting coefficients a, b, c, and d provide a basis for subsequent spectral response judgment.

[0121] Step S403: Based on the cubic fitting coefficients, perform a logical judgment to obtain the second logical judgment result;

[0122] Specifically, the formula for logical judgment is:

[0123] sgn(b / a > 100) * sgn( - 1726) * sgn(1734 - ) * b * k > threshold3;

[0124] In the above formula, sgn(b / a > 100) is used to determine whether the fitting coefficient b / a is greater than 100 to ensure the significance of the change in reflectance; sgn( -1726) Determine whether the inflection point position of the reflectance curve (i.e., the point where the second derivative is zero) is close to 1726 nm, which is related to the wavelength position of the second absorption feature; sgn(1734 - ) is used to ensure that the inflection point of the reflectance curve is near 1734 nm, which is also the wavelength range of the second absorption feature; a, b, and c are cubic fitting coefficients, and k is the ratio of the reflectance of this pixel in the 1700 - 1745 nm interval to the average reflectance value of all pixels in the same band in the entire image, such as , M 1700-1745 is the average value of the reflectance sequence Y of this pixel, and TM 1700-1745 is the average reflectance value of all pixels in the entire image in the 1700 nm - 1745 nm interval. This ratio reflects the relative reflectance level of this pixel, thereby helping to further screen out areas with significant absorption features; threshold3 represents the final judgment of whether the specified threshold is met. If the condition is met, it indicates that this area conforms to the second absorption feature.

[0125] It should be noted that the coefficients a, b, and c have important effects on the shape, inflection point, and other characteristics of the signal curve, while the constant term d is mainly used to adjust the translation of the curve. Therefore, it is not directly used in subsequent logical judgments.

[0126] Step S404, determine the area where the spectral response conforms to the second absorption feature according to the second logical judgment result, and generate the third mask.

[0127] Among them, combining the result of the second logical judgment, determine the area where the spectral response conforms to the 1730 nm absorption feature, and mark these areas as 1 to generate the third mask. The pixels marked as 1 in the third mask indicate that the spectral responses of these pixels conform to the second absorption feature (1730 nm absorption feature).

[0128] In the above embodiments, spectral data in the 1700 nm to 1745 nm band is extracted from the hyperspectral remote sensing image, the relationship between the reflectance and the wavelength is accurately fitted using a cubic fitting model, and areas where the spectral response conforms to the 1730 nm absorption feature are screened out through complex logical judgments. These areas that conform to the feature are marked by the third mask, providing clear data support for subsequent spatial analysis and feature extraction.

[0129] As the basis for the selection of the first band interval in the embodiments of the present application, plastic greenhouse covering materials (such as polyethylene and polyvinyl chloride) have significant spectral absorption characteristics in the short-wave infrared region (about 2000-2500 nm). The absorption peak near 2310 nm is the typical position caused by the stretching vibration of C-H bonds in plastic materials. Therefore, a band interval of 2290-2325 nm is selected to cover the complete profile of this absorption peak, ensuring that the central position (2310 nm) of the absorption signal and the changes on both sides can be accurately captured.

[0130] As the basis for the selection of the third band interval in the embodiments of the present application, there are also secondary absorption peaks in the mid-infrared region (about 1700-1800 nm) of plastic materials, mainly caused by the overtone vibration of C-H bonds or material additives (such as stabilizers). The absorption near 1730 nm is the typical position of such absorption. Therefore, a band range of 1700-1745 nm is selected to cover the complete range of the secondary absorption peak, avoiding signal omission caused by spectral drift or noise.

[0131] In summary, the first mask (2310 nm) and the third mask (1730 nm) respectively target the absorption characteristics of plastic materials in different infrared regions, forming a double verification. By combining multi-band analysis, the specificity is enhanced, and the omission or misdetection caused by misjudgment in a single band (such as interference from similar materials) is reduced. As the basis for the selection of the second band interval in the embodiments of the present application, the second band interval (680-750 nm) utilizes the reflectance difference between plastic and vegetation in the red edge interval to further exclude natural surface interference, realizing the extraction of the spatial distribution of agricultural greenhouses with high precision and high efficiency.

[0132] In addition, in order to address the problem of misdetection that may be caused by weak spectral response in specific bands, the spectral images of specific bands can also be enhanced. Smoothing filtering is used to reduce noise interference. At the same time, according to the actual reflection characteristics of agricultural greenhouses in the target band, the above-mentioned preset threshold is dynamically adjusted, so as to ensure that the mask generation is more accurate and effectively improve the overall recognition effect. Moreover, a parallel computing architecture and an efficient algorithm library can be introduced. In each step of preprocessing and mask generation, the resources of multi-core processors are fully utilized to perform piecewise processing of large-scale remote sensing image data, so as to significantly shorten the processing time and improve the speed and efficiency of data processing.

[0133] The embodiments of the present application also disclose a system for extracting the spatial distribution information of agricultural greenhouses.

[0134] A system for extracting the spatial distribution information of agricultural greenhouses includes:

[0135] A data receiving module for receiving real-time collected aerospace hyperspectral remote sensing image data;

[0136] A preprocessing module for preprocessing hyperspectral remote sensing image data of aerospace to obtain surface reflectance data;

[0137] A first mask generation module for extracting a spectral subset in a first band interval from the surface reflectance data, performing quadratic fitting to obtain reflectance fitting coefficients, and determining regions where the spectral response conforms to the first absorption feature through logical judgment to generate a first mask;

[0138] A second mask generation module for extracting a spectral subset in a second band interval from the surface reflectance data, calculating the reflectance difference between both ends of the second band interval and performing threshold comparison to generate a second mask;

[0139] A third mask generation module for extracting a spectral subset in a third band interval from the surface reflectance data, performing cubic fitting calculation to obtain reflectance fitting coefficients, and determining regions where the spectral response conforms to the second absorption feature through logical judgment to generate a third mask;

[0140] A logical fusion module for performing logical fusion based on the first mask, the second mask, and the third mask to obtain the extraction result of the spatial distribution information of agricultural greenhouses.

[0141] The agricultural greenhouse spatial distribution information extraction system according to the embodiments of the present application can implement any of the above - mentioned methods for extracting the spatial distribution information of agricultural greenhouses, and the specific working processes of each module in the distribution information extraction system can refer to the corresponding processes in the above - mentioned method embodiments.

[0142] In several embodiments provided by the present application, it should be understood that the provided methods and systems can be implemented in other ways. For example, the system embodiments described above are only illustrative; for example, the division of a certain module is only a logical function division, and there may be other division methods in actual implementation. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0143] The embodiments of the present application also disclose a computer device.

[0144] A computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements a method for extracting the spatial distribution information of agricultural greenhouses as described above.

[0145] The embodiments of the present application also disclose a computer - readable storage medium.

[0146] A computer - readable storage medium stores a computer program that can be loaded and executed by a processor to implement any of the methods for extracting the spatial distribution information of agricultural greenhouses as described above.

[0147] Among them, the computer-readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, apparatus, or device; the program code contained on the computer-readable medium can be transmitted by any suitable medium, including but not limited to wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0148] It should be noted that in the above embodiments, the descriptions of each embodiment have their own emphases. For parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0149] The above are all preferred embodiments of the present application. The protection scope of the present application is not limited hereby. Any feature disclosed in this specification (including the abstract and drawings), unless specifically described, can be replaced by other equivalent or similar-purpose alternative features. That is, unless specifically described, each feature is only an example of a series of equivalent or similar features.

Claims

1. A method for extracting spatial distribution information of agricultural greenhouses, characterized in that: The extraction method comprises: Receive real-time aerospace hyperspectral remote sensing image data and pre-process it to obtain surface reflectance data; Extracting a spectral subset of a first band interval from the surface reflectance data, performing quadratic fitting to obtain a reflectance fitting coefficient, combining logical judgment to determine an area where the spectral response meets the first absorption feature, and generating a first mask; Extracting a spectral subset of a second band interval from the surface reflectance data, calculating a reflectance difference between two ends of the second band interval and performing a threshold comparison to generate a second mask; Extracting a spectral subset of a third band interval from the surface reflectance data, performing three fitting calculations to obtain a reflectance fitting coefficient, combining logical judgment to determine an area where the spectral response meets the second absorption feature, and generating a third mask; The first mask, the second mask and the third mask are logically integrated to obtain the extraction result of the spatial distribution information of the agricultural greenhouse; the spectral subset of the first band interval is extracted from the surface reflectance data, and a quadratic fitting is performed to obtain the reflectance fitting coefficient, and the area where the spectral response meets the first absorption feature is determined by combining logical judgment. The step of generating the first mask includes: Extracting a spectral subset of pixels in a first band interval from the surface reflectance data to obtain a wavelength sequence and a reflectance sequence; Constructing a quadratic fitting coefficient matrix according to the wavelength sequence and the reflectivity sequence, calculating the reflectivity fitting coefficients by the least squares method, and obtaining the quadratic fitting coefficients; Performing logical judgment based on the quadratic fitting coefficient to obtain a first logical judgment result; A region whose spectral response conforms to a first absorption characteristic is determined according to the first logical judgment result, and a first mask is generated.

2. The method for extracting spatial distribution information of agricultural greenhouses according to claim 1, characterized in that: The steps of extracting a spectral subset of a second band interval from the surface reflectance data, calculating the reflectance difference between the two ends of the second band interval and performing threshold comparison, and generating a second mask include: Extracting a spectral subset of a second wavelength band from the surface reflectance data; Calculating the reflectivity difference between the two ends of the second waveband interval; The reflectivity difference is compared with a preset threshold, and a second mask is generated according to the position of the area where the reflectivity difference is greater than the preset threshold.

3. The method for extracting spatial distribution information of agricultural greenhouses according to claim 1, characterized in that: The steps of extracting a spectral subset of the third band interval from the surface reflectance data, performing three fitting calculations to obtain reflectance fitting coefficients, and combining logical judgment to determine the area where the spectral response meets the second absorption feature, and generating a third mask include: Extracting a spectral subset of pixels in a third band from the surface reflectance data to obtain a wavelength sequence and a reflectance sequence; Constructing a cubic fitting coefficient matrix according to the wavelength sequence and the reflectivity sequence, calculating the reflectivity fitting coefficients, and obtaining the cubic fitting coefficients; Performing logical judgment based on the cubic fitting coefficients to obtain a second logical judgment result; The region whose spectral response conforms to the second absorption characteristic is determined according to the second logic judgment result, and a third mask is generated.

4. The method for extracting spatial distribution information of agricultural greenhouses according to claim 1, characterized in that: The spectral range of the aerospace hyperspectral remote sensing image data is 400 nanometers to 2500 nanometers, the spectral resolution is less than 8 nanometers, and the spatial resolution is less than 30 meters.

5. A method for extracting spatial distribution information of agricultural greenhouses according to claim 4, characterized in that: The first waveband range is 2290 nanometers to 2325 nanometers; the second waveband range is 680 nanometers to 750 nanometers; and the third waveband range is 1700 nanometers to 1745 nanometers.

6. A method for extracting spatial distribution information of agricultural greenhouses according to any one of claims 1 to 5, characterized in that: The first mask, the second mask and the third mask are all logical matrices and are consistent with the spatial dimensions of the aerospace hyperspectral remote sensing image data.

7. An agricultural greenhouse spatial distribution information extraction system, characterized in that: The extraction system comprises: A data receiving module is used to receive aerospace hyperspectral remote sensing image data collected in real time; A preprocessing module, used for preprocessing the aerospace hyperspectral remote sensing image data to obtain surface reflectance data; A first mask generation module is used to extract a spectral subset of a first band interval from the surface reflectance data, perform quadratic fitting to obtain a reflectance fitting coefficient, determine an area where the spectral response meets the first absorption feature by combining logical judgment, and generate a first mask; A second mask generating module, used for extracting a spectral subset of a second band interval from the surface reflectance data, calculating a reflectance difference between two ends of the second band interval and performing a threshold comparison to generate a second mask; A third mask generation module is used to extract a spectral subset of a third band interval from the surface reflectance data, perform three fitting calculations to obtain a reflectance fitting coefficient, and combine logical judgment to determine an area where the spectral response meets the second absorption feature to generate a third mask; A logic fusion module, used for performing logic fusion according to the first mask, the second mask and the third mask to obtain an extraction result of the spatial distribution information of the agricultural greenhouse; The first mask generation module is configured to: The step of extracting a spectral subset of the first band interval from the surface reflectance data, performing quadratic fitting to obtain a reflectance fitting coefficient, and combining logical judgment to determine an area where the spectral response meets the first absorption feature, and generating a first mask includes: Extracting a spectral subset of pixels in a first band interval from the surface reflectance data to obtain a wavelength sequence and a reflectance sequence; Constructing a quadratic fitting coefficient matrix according to the wavelength sequence and the reflectivity sequence, calculating the reflectivity fitting coefficients by the least squares method, and obtaining the quadratic fitting coefficients; Performing logical judgment based on the quadratic fitting coefficient to obtain a first logical judgment result; A region whose spectral response conforms to a first absorption characteristic is determined according to the first logical judgment result, and a first mask is generated.

8. A computer device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and execute the method according to any one of claims 1 to 6.

Images