A method for detecting nitrogen nutrition in sugarcane canopy based on UAV hyperspectral images

Abstract

This invention belongs to the field of sugarcane nutrient diagnosis technology. A method for detecting nitrogen nutrition in the sugarcane canopy based on UAV hyperspectral images includes the following steps: selecting several sugarcane plants of different growth stages in the study area and marking them for location; simultaneously collecting leaf samples from these plants and sending them to a laboratory for analysis to obtain the nitrogen content of the sugarcane sample points; using a UAV equipped with a hyperspectral imaging camera to fly over the study area and acquire hyperspectral images of the sugarcane field; constructing a spectral index reflecting the nitrogen content of the sugarcane canopy using the reflectance of the sensitive wavelengths of the sugarcane sample points; achieving automatic and high-precision extraction of the nutritional status of the sugarcane canopy, providing timely and targeted guidance for improving sugarcane yield and quality.

Description

Technical Field

[0001] This invention belongs to the field of sugarcane nutrition diagnostic technology, specifically relating to a method for detecting nitrogen nutrition in the sugarcane canopy based on UAV hyperspectral images. Background Technology

[0002] Nitrogen is an essential nutrient element for crop growth and development, significantly affecting crop growth, yield, and quality. Nitrogen nutrient status is a crucial indicator for evaluating crop growth, increasing yield, and improving crop quality; therefore, real-time and rapid nitrogen diagnosis is the basis for scientific crop management.

[0003] Unmanned aerial vehicle (UAV) remote sensing features high resolution and rapid observation over a wide area. Real-time and precise monitoring of crop nitrogen nutrition status using UAV hyperspectral imagery is crucial for achieving high crop yields and efficiency. Canopy spectral analysis technology can effectively monitor crop growth characteristics and plant nutritional status dynamically. Spectral indices such as difference vegetation index, normalized difference vegetation index, and ratio vegetation index are widely used for plant leaf area index, dry matter production, leaf nitrogen content and accumulation, leaf carbon-nitrogen ratio, and yield estimation. Compared to traditional physical and biochemical methods for studying crop nutritional status, canopy spectral analysis technology offers significant advantages such as speed, portability, non-destructive testing, and freedom from spatial and temporal limitations, making it a urgently needed technological tool in the field of agricultural informatization.

[0004] Most existing studies extract vegetation indices such as NDVI from crop drone hyperspectral images to establish nitrogen estimation models, thereby evaluating the nitrogen nutrient abundance or deficiency in the crop canopy. Many studies have also reduced the indices used in the estimation models to a finer scale, improving the accuracy of extracting crop canopy nitrogen nutrient status. Existing technology CN112986158A discloses a method and system for detecting nitrogen nutrient in sugar beets based on drone multispectral data, which inverts ten vegetation indices across three categories and 19 vegetation indices; analyzes the nitrogen nutrient variation patterns of different sugar beet varieties under different nitrogen application rates; evaluates the correlation and correlation coefficients between drone multispectral indices in sugar beet experimental plots and sugar beet leaf nitrogen content, root nitrogen content, whole plant nitrogen content, leaf nitrogen accumulation, root nitrogen accumulation, and whole plant nitrogen accumulation; selects the vegetation indices with the highest correlation to sugar beet nitrogen nutrient indicators from the three categories for modeling; identifies the most accurate sugar beet growth monitoring indicators; formulates sugar beet nitrogen nutrient diagnostic criteria based on the optimal spectral vegetation indices; and establishes an evaluation system. Experiments show that drones can diagnose nitrogen accumulation per unit area (NWL), root nitrogen accumulation (NWT), and whole plant nitrogen accumulation (NWP), and the diagnostic results are statistically significant. The aforementioned studies can predict plant nitrogen nutrition status through spectral analysis; however, these methods require repeated data conversions, are influenced by numerous factors, and have low reliability. Furthermore, the spectral indices used to predict the carbon-to-nitrogen ratio are obtained through simple mathematical calculations using only the raw reflectance of two or three characteristic wavelengths, leading to low stability. In addition, these studies primarily focus on sugar beets or grain crops such as rice and wheat, lacking coverage of major economic crops like sugarcane. Sugarcane is a crucial crop in the modern sugar industry, providing approximately 80% of the world's sugar. China is one of the earliest sugarcane-growing countries in the world, with a planting area second only to Brazil and India. The role of sugarcane in my country's agricultural industry is undeniable; therefore, timely and accurate understanding of the nitrogen nutrition status and growth information of the sugarcane canopy is of great significance. This invention provides a method for detecting nitrogen nutrition in sugarcane canopy using hyperspectral images from unmanned aerial vehicles (UAVs). It utilizes the characteristics of hyperspectral curves to determine spectral ranges sensitive to different nitrogen contents in sugarcane. Then, based on these sensitive spectral ranges, a spectral index M reflecting the nitrogen content in the sugarcane canopy is constructed using a ratio method. This ratio method avoids systematic, linear reflectance errors caused by factors such as shooting time, weather conditions, and lighting conditions. Finally, using the calculated spectral index M and the nitrogen content of sugarcane samples obtained through laboratory analysis, a nitrogen content prediction model is constructed to obtain the nitrogen nutrition status of sugarcane in the study area, and the results have high reliability. Summary of the Invention

[0005] This invention overcomes the shortcomings of the above-mentioned technical problems and provides a method for detecting nitrogen nutrition in sugarcane canopy based on UAV hyperspectral images. By constructing a spectral index M that reflects the nitrogen content in sugarcane canopy, the method achieves automatic and accurate extraction of nitrogen nutrition status in sugarcane canopy, providing data support for adjusting agronomic management measures to further improve sugarcane yield and quality.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for detecting nitrogen nutrition in the canopy of sugarcane based on hyperspectral images from unmanned aerial vehicles (UAVs) includes the following steps:

[0008] S1. Select several sugarcane plants of different growth rates in the study area, mark and locate them, and collect leaf samples at the same time as marking. Send the samples to the laboratory for analysis to obtain the nitrogen content of the sugarcane sample points.

[0009] S2. A drone equipped with a hyperspectral imaging camera was used to conduct aerial flights over the study area to acquire hyperspectral images of the sugarcane fields. The hyperspectral images were preprocessed to extract hyperspectral curves from several sugarcane sample points. Then, the spectral intervals sensitive to different leaf nitrogen contents were determined through feature analysis of the spectral curves. Within the sensitive spectral intervals, characteristic bands were further determined. Finally, a spectral index M reflecting the nitrogen content of the sugarcane canopy was constructed based on the reflectance values ​​of the characteristic bands. The formula for calculating the spectral index M is as follows:

[0010]

[0011] In the formula, M is a spectral index reflecting the nitrogen content of the sugarcane canopy; B 829 The reflectance value is for a wavelength of 829 nm; B 841 The reflectance value is for a wavelength of 841nm; B 827 The reflectance value is for a wavelength of 827nm; B 838 The reflectance value is for a wavelength of 838nm;

[0012] S3. Based on the calculated spectral index M of several sugarcane sample points and the nitrogen content of the sugarcane sample points obtained through laboratory analysis in step S1, a nitrogen content prediction model is constructed. The calculation formula of the nitrogen content prediction model is as follows:

[0013] y = 4.8721x 3 -16.039x 2 +16.26x-4.7508 (2)

[0014] In the formula, y is the nitrogen content of a sugarcane plant at a certain location to be predicted; x is the spectral index M of the sugarcane canopy at that location;

[0015] S4. Develop diagnostic standards for sugarcane nitrogen nutrition based on nitrogen content prediction models and establish an evaluation system;

[0016] S5. Obtain UAV hyperspectral images of the sugarcane field in the area to be predicted, and calculate the sugarcane nitrogen content in the area using the nitrogen content prediction model to verify the effectiveness of the method of the present invention.

[0017] Furthermore, in S1, leaf samples from the canopy of at least 20-30 sugarcane plants with different growth conditions are collected in the study area. The leaf samples are evenly distributed in space. The growth conditions include three types: good growth, average growth, and abnormal growth.

[0018] Furthermore, in S2, the preprocessing of the hyperspectral image includes: firstly, radiometric calibration and radiometric correction of the hyperspectral images of each flight zone based on a photometer and a standard reflector, and then geometric correction and mosaicking are performed by combining ground control points (GCP) and polynomial methods to form a complete hyperspectral reflectance image of the region.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] This invention discloses a method for detecting nitrogen nutrition in sugarcane canopy using hyperspectral images from unmanned aerial vehicles (UAVs). It identifies spectral ranges sensitive to different nitrogen contents in sugarcane by utilizing the characteristics of the hyperspectral curves. Then, based on these sensitive spectral ranges, a spectral index M reflecting the nitrogen content in the sugarcane canopy is constructed using a ratio method. This ratio method avoids systematic and linear reflectance errors caused by factors such as shooting time, weather conditions, and lighting conditions. Finally, using the calculated spectral index M and the nitrogen content of sugarcane samples obtained through laboratory analysis, a nitrogen content prediction model is constructed to obtain the nitrogen nutrition status of sugarcane in the study area. This achieves automated and high-precision extraction of the nitrogen and other nutritional status of the sugarcane canopy, providing timely and targeted guidance for improving sugarcane yield and quality. Furthermore, it fills a gap in current research on the dynamic monitoring of nitrogen nutrition status in the sugarcane canopy.

[0021] This invention diagnoses nitrogen levels in sugarcane using hyperspectral images from drones. Data acquisition is simple and convenient, nitrogen diagnosis is low-cost and fast, and the nitrogen nutritional status of sugarcane can be obtained in real time under natural light conditions in sugarcane fields. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention;

[0023] Figure 2 These are the sugarcane canopy leaf sample markers of the present invention;

[0024] Figure 3 and Figure 4All are hyperspectral curves of sugarcane canopy leaf samples from the present invention; wherein, Figure 3 The hyperspectral curves of each sugarcane leaf sample are shown. Figure 4 The following are local hyperspectral curves for each sugarcane leaf sample;

[0025] Figure 5 This invention relates to the correlation between the spectral index M of sugarcane leaf samples and the true nitrogen content at each sample point.

[0026] Figure 6 This is an unmanned aerial vehicle (UAV) image of the experimental sugarcane field of this invention;

[0027] Figure 7 This is a schematic diagram of the grading of nitrogen content in the canopy of sugarcane in the experimental sugarcane field of this invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Please see Figure 1 A method for detecting nitrogen nutrition in the canopy of sugarcane based on hyperspectral images from unmanned aerial vehicles (UAVs) includes the following steps:

[0030] S1. Select several sugarcane plants of different growth stages in the study area, mark and locate them, and collect leaf samples from the corresponding plants. Specifically, collect leaf samples from the canopy of at least 20-30 sugarcane plants of different growth stages in the study area. The leaf samples are evenly distributed in space; the growth stage includes three types: good growth, average growth, and abnormal growth. In this embodiment, 27 sugarcane plants of different growth stages in the study area are selected and marked, obtaining 27 sample points. Leaf samples are collected simultaneously with marking and sent to the laboratory for analysis to obtain the nitrogen content of the sugarcane sample points; details are as follows:

[0031] ① Select 27 sugarcane plants with different growth conditions in the sugarcane field, mark them with red ribbons as sample points, and collect leaf samples from the corresponding plants (two old, two medium and two young leaves from each plant to form a sample group). Figure 2 This paper demonstrates the locations of 27 sugarcane sample points collected using this invention within the sugarcane field. Marking the locations of the sugarcane sample points facilitates the accurate acquisition of their spectral curves from subsequent UAV hyperspectral imagery.

[0032] ② The 27 collected leaf samples were promptly sent to the laboratory for analysis of leaf nitrogen content using a Kjeldahl nitrogen analyzer.

[0033] ③ Each group of sugarcane leaf samples was tested three times, and the average of the three test results was taken as the nitrogen content of that group of leaf samples (which can also be regarded as the sugarcane canopy at the corresponding sampling point). Table 1 shows the nitrogen content of sugarcane leaf samples. The left column of Table 1 is the leaf sample number, and the right column is the nitrogen content test result of each sample.

[0034] Table 1

[0035]

[0036]

[0037] S2. A drone equipped with a hyperspectral imaging camera was used to conduct aerial flights over the study area to acquire hyperspectral images of the sugarcane fields. These images were then preprocessed. The preprocessing included: first, radiometric calibration and correction of the hyperspectral images for each flight zone were performed using a photometer and a standard reflector; then, geometric correction and mosaicking were performed using ground control points (GCPs) and polynomial methods to form a complete hyperspectral reflectance image of the region. Based on the preprocessed hyperspectral reflectance image, hyperspectral curves were extracted from several sugarcane sample points. Characteristic analysis of the spectral curves determined the spectral ranges sensitive to nitrogen content in different leaves, and characteristic bands were further identified within these sensitive spectral ranges. Finally, a spectral index M reflecting the nitrogen content of the sugarcane canopy was constructed based on the reflectance values ​​of the characteristic bands, as detailed below:

[0038] ①Use a drone equipped with a hyperspectral camera to conduct aerial photography and data preprocessing of the study area to obtain hyperspectral reflectance images;

[0039] ② Select the 27 sugarcane sample points marked with red bands in the hyperspectral reflectance image;

[0040] ③ Extract the hyperspectral reflectance curves from the selection boxes of the 27 sample points respectively. Since the selection box of each sample point contains multiple pixels, the average value of each pixel in the selection box is calculated according to the wavelength, thus obtaining the hyperspectral reflectance curves of the 27 sugarcane samples respectively. Figure 3 The hyperspectral reflectance curves for sugarcane samples are in the wavelength range of 406-1028 nm. However, due to the influence of weather, lighting, and imaging systems, there is significant noise in the wavelength range of 900-1028 nm. Therefore, in practice, only data in the wavelength range of 406-900 nm are used.

[0041] ④ Through characteristic analysis of the spectral curves, it was found that the spectral range of 824-845nm is sensitive to different nitrogen contents in sugarcane leaves. Figure 4 The magnified images of the sugarcane samples in the 824-845 nm spectral range further reveal significant changes in reflectance at four wavelengths: 827 nm, 829 nm, 838 nm, and 841 nm, in response to different nitrogen contents. Therefore, a spectral index M reflecting the nitrogen content of the sugarcane canopy is constructed based on the reflectance values ​​at these four wavelengths. The formula for calculating the spectral index M is as follows:

[0042]

[0043] In the formula, M is a spectral index reflecting the nitrogen content of the sugarcane canopy; B 829 The reflectance value is for a wavelength of 829 nm; B 841 The reflectance value is for a wavelength of 841nm; B 827 The reflectance value is for a wavelength of 827nm; B 838 The reflectance value is for a wavelength of 838nm.

[0044] Table 2 and S3 show the calculation results of the sugarcane sample index M, with values ​​ranging from 0.63 to 0.93. Based on the calculated spectral index M of the 27 sugarcane sample points and the nitrogen content of the sugarcane sample points obtained from laboratory monitoring in step S1, a nitrogen content prediction model was constructed. The calculation formula for the nitrogen content prediction model is as follows:

[0045] y = 4.8721x 3 -16.039x 2 +16.26x-4.7508 (2)

[0046] In the formula, y is the nitrogen content of a sugarcane plant at a certain location to be predicted; x is the spectral index M of the sugarcane canopy at that location.

[0047] Table 2

[0048] A-Ⅰ-1 0.722 B-Ⅰ-1 0.761 C-Ⅰ-1 0.801 A-Ⅰ-2 0.772 B-Ⅰ-2 0.738 C-Ⅰ-2 0.760 A-Ⅰ-3 0.786 B-Ⅰ-3 0.806 C-Ⅰ-3 0.742 A-Ⅱ-1 0.717 B-Ⅱ-1 0.825 C-Ⅱ-1 0.679 A-Ⅱ-2 0.751 B-Ⅱ-2 0.721 C-Ⅱ-2 0.736 A-Ⅱ-3 0.759 B-Ⅱ-3 0.730 C-Ⅱ-3 0.707 A-Ⅲ-1 0.721 B-Ⅲ-1 0.763 C-Ⅲ-1 0.692 A-Ⅲ-2 0.785 B-Ⅲ-2 0.752 C-Ⅲ-2 0.632 A-Ⅲ-3 0.741 B-Ⅲ-3 0.813 C-Ⅲ-3 0.925

[0049] Figure 5 A scatter plot of the spectral index M and its nitrogen content for sugarcane samples reveals the relationship between the spectral index M constructed in this invention and the sugarcane nitrogen content, and the R-value. 2 The correlation coefficient reached 0.84, indicating a high correlation between the two factors. Subsequent use of this model to predict the nitrogen content in the sugarcane canopy showed high accuracy. (Correlation coefficient R0) 2 R is a statistical indicator that measures the degree of correlation between two variables; it reflects the strength of the linear relationship between the two variables. In this invention, R... 2 This represents the correlation between the sugarcane canopy nitrogen nutrient index M (x value) constructed in this invention and the sugarcane nitrogen content (Y value) predicted by the laboratory. When R2 The closer the correlation coefficient is to 1, the stronger the correlation between the two. The correlation coefficient R between the sugarcane canopy nitrogen nutrient index M (x value) and the predicted sugarcane nitrogen content (Y value) was calculated. 2 The result of 0.84 demonstrates a close relationship between the two. Therefore, it is reasonable and reliable to construct a correlation between the two and use the sugarcane canopy nitrogen nutrient index M (x value) to calculate the sugarcane nitrogen content (Y value).

[0050] S4. Develop diagnostic standards for sugarcane nitrogen nutrition based on a nitrogen content prediction model and establish an evaluation system. Details are as follows:

[0051] The nitrogen content of sugarcane canopy is obtained based on the nitrogen content prediction model. By setting certain thresholds, the nitrogen content of sugarcane is divided into three levels: low, medium, and high. This allows growers or relevant departments to adjust agronomic measures such as root fertilization and foliar fertilization according to the different nitrogen nutrition status of sugarcane, thereby improving sugarcane yield and quality.

[0052] S5. Obtain UAV hyperspectral images of the sugarcane field in the area to be predicted, and calculate the sugarcane nitrogen content in the area using the nitrogen content prediction model to verify the effectiveness of the method of the present invention. Specifically as follows:

[0053] Based on the sugarcane nitrogen content prediction model constructed in this invention, a UAV hyperspectral image of a sugarcane field was selected to calculate the sugarcane nitrogen content in that area. This experiment was conducted to further verify the reliability of the sugarcane canopy nitrogen nutrition detection method based on UAV hyperspectral images proposed in this invention.

[0054] Figure 6 This invention presents drone imagery of an experimental sugarcane field and a schematic diagram illustrating the nitrogen content grading of the sugarcane canopy. The experimental field was located in Ba Miao Village, Longzhou County, Chongzuo City, Guangxi Zhuang Autonomous Region. Drone imagery of the experimental sugarcane field was taken on June 29, 2023. Figure 6 As shown. From Figure 6The sugarcane seedlings in the field were observed to be neatly arranged, but varied in size and growth vigor, possibly due to differences in soil nutrient status. Therefore, to accurately understand the nitrogen nutrition status of sugarcane and adjust agronomic measures in a timely manner, we calculated the nitrogen content of the sugarcane in this field based on the nitrogen content prediction model constructed in this invention. First, the hyperspectral images of the experimental area were preprocessed, including radiometric correction, geometric correction, and image stitching, to obtain the reflectance image of the area. Second, the spectral index M was calculated using the reflectance values ​​at four wavelengths: 827nm, 829nm, 838nm, and 841nm. Finally, the index M was substituted into the nitrogen content prediction model constructed in this invention to obtain the nitrogen content distribution of the sugarcane canopy in the experimental field. The nitrogen distribution of the sugarcane in the field was then graded by defining three thresholds: low (0g-0.3g), medium (0.3g-0.4g), and high (>0.4g). The results are as follows: Figure 7 As shown in the figure. This experiment further confirms the practical applicability of the proposed method for detecting nitrogen nutrition in the sugarcane canopy based on UAV hyperspectral images.

[0055] This invention utilizes hyperspectral images from unmanned aerial vehicles (UAVs) to construct a spectral index that reflects the nitrogen content in the sugarcane canopy, thereby achieving automatic and high-precision extraction of the nutritional status of sugarcane canopy, including nitrogen. This research provides a good theoretical reference for real-time nitrogen nutrition diagnosis and optimized nitrogen management in sugarcane, and also fills the current research gap in the field of dynamic monitoring of nitrogen nutrition status in sugarcane canopy.

[0056] The above description is a detailed description of the preferred embodiments of the present invention. However, the embodiments are not intended to limit the scope of the patent application of the present invention. All equivalent changes or modifications made under the technical spirit of the present invention should fall within the patent scope covered by the present invention.

Claims

1. A method for detecting nitrogen nutrition in sugarcane canopy based on UAV hyperspectral images, characterized in that, Includes the following steps: S1. Select several sugarcane plants of different growth rates in the study area, mark and locate them, and collect leaf samples at the same time as marking. Send the samples to the laboratory for analysis to obtain the nitrogen content of the sugarcane sample points. S2. A drone equipped with a hyperspectral imaging camera was used to conduct aerial flights over the study area to acquire hyperspectral images of the sugarcane fields. The hyperspectral images were preprocessed to extract hyperspectral curves from several sugarcane sample points. Then, the spectral intervals sensitive to different leaf nitrogen contents were determined through feature analysis of the spectral curves. Within the sensitive spectral intervals, characteristic bands were further determined. Finally, a spectral index M reflecting the nitrogen content of the sugarcane canopy was constructed based on the reflectance values ​​of the characteristic bands. The formula for calculating the spectral index M is as follows: In the formula, M is a spectral index reflecting the nitrogen content of the sugarcane canopy; B 829 is the reflectance value at a wavelength of 829 nm;B 841 is the reflectance value at a wavelength of 841 nm;B 827 is the reflectance value at a wavelength of 827 nm;B 838 is the reflectance value at a wavelength of 838 nm; S3. Based on the calculated spectral index M of several sugarcane sample points and the nitrogen content of the sugarcane sample points obtained through laboratory analysis in step S1, a nitrogen content prediction model is constructed. The calculation formula of the nitrogen content prediction model is as follows: y=4.8721x 3 -16.039x 2 +16.26x-4.7508 (2) In the formula, y is the nitrogen content of a sugarcane plant at a certain location to be predicted; x is the spectral index M of the sugarcane canopy at that location; S4. Develop diagnostic standards for sugarcane nitrogen nutrition based on nitrogen content prediction models and establish an evaluation system; S5. Obtain UAV hyperspectral images of the sugarcane field in the area to be predicted, and calculate the sugarcane nitrogen content in the area using the nitrogen content prediction model to verify the effectiveness of the method of the present invention.

2. The method for detecting nitrogen nutrition in sugarcane canopy based on UAV hyperspectral images according to claim 1, characterized in that: In step S1, leaf samples from the canopy of at least 20-30 sugarcane plants with different growth conditions are collected in the study area. The leaf samples are evenly distributed in space. The growth conditions include three types: good growth, average growth, and abnormal growth.

3. The method for detecting nitrogen nutrition in sugarcane canopy based on UAV hyperspectral images according to claim 1, characterized in that: In step S2, the preprocessing of the hyperspectral image includes: firstly, radiometric calibration and radiometric correction of the hyperspectral image of each flight zone based on a photometer and a standard reflector, and then geometric correction and mosaicking combined with ground control points (GCP) and polynomial methods to form a complete hyperspectral reflectance image of the region.

Images