Nitrogen detection in cotton leaves based on hyperspectral
author: cily
2022-12-23
In recent years, it has been increasingly noticed that detecting nitrogen in crops has a huge impact on crop growth, even destructive detection, and is costly, and these reasons make people eager for the emergence of a new technology that can replace such destructive detection methods. Hyperspectral imaging technology can solve these problems. Hyperspectral imaging technology has won people's favor with its non-destructive, fast and accurate characteristics, and gradually become an important means to detect crop content in recent years. Different fertility periods of maize have established more accurate
Maize nitrogen monitoring and diagnostic model for rational nitrogen application and efficient use of nitrogen fertilizer.
In this paper, hyperspectral data of cotton leaves in southern Xinjiang were collected, and the nitrogen content of cotton leaves was measured.
The nitrogen content of cotton leaves was measured, and the nitrogen data were analyzed using a combination of different algorithms such as PLS, normalization, SPA-PLS, and data centration. The nitrogen data and hyperspectral data of cotton leaves were simulated using a combination of different algorithms such as PLS, normalization, SPA-PLS, and data centering. The gradual regression of SPA-PLS was preferentially selected as the best method to reflect the nitrogen content of cotton leaves. It can reflect the hyperspectral response of cotton nitrogen more accurately, and provide the most accurate, fast and non-destructive detection of cotton nitrogen. It provides a theoretical method for accurate, rapid and nondestructive detection of cotton nitrogen.
1 Experimental part
1.1 Sample collection
The experiment was conducted in 2019 at the First Division Irrigation Experiment Station in Alar, Xinjiang.
The experimental field was designed with 18 plots and 3 treatments. Urea (N), calcium phosphate (P) and potassium sulfate (K) were applied to the experimental field.
(N), calcium dihydrogen phosphate (P), and potassium sulfate (K) were applied to cotton at regular intervals.
The fertilizer was applied on July 10, 2019 (bud stage), July 26, 2019 (flowering stage), and August 2019 (flowering stage). period), August 2, 2019 (boll stage), and August 26, 2019 (boll stage)
Nitrogen and hyperspectral data of cotton leaves were collected for four periods, each period 180 data were collected for each period, and a total of 720 samples were collected for the four periods.
1.2 Measurement of leaf nitrogen content
Cotton leaves were labeled on the plant with a plant nutrient meter to ensure that the measured nitrogen values of cotton leaves corresponded to the spectral values of the leaves to be measured, and then the nitrogen values were measured three times on each leaf. The nitrogen values were then measured three times on each leaf, and the three nitrogen values were averaged as the nitrogen value of the leaf.
1.3 Leaf hyperspectral data acquisition
The wavelength range was used in the experiment
325-1075 nm, and the number of sample spectra acquisition was set to 5 spectra per acquisition. The cotton leaves collected in the experiment were placed on a platform covered with a black baffle, and then the distance from the lens to the platform was adjusted to 20 cm to measure the spectral values of the desired leaves.
The 5 spectral data of each cotton leaf were averaged as the hyperspectral data of this leaf.
1.4 Modeling Method
In this study, PLS was used to model the correction and prediction sets for cotton leaf detection. The correlation coefficient R, root mean square erroe (RMSE), prediction accuracy (Prediction accuracy), and the number of leaves in the calibration and prediction sets were determined by the partial least squares method. (RMSE), prediction accuracy (Precision), root mean square error (RMSE), and root mean square erroe of prediction (RMSE). square erroe of prediction (RMSEP), cross-validation mean square erroe of prediction (RMSE), and cross-validation (RMSE), prediction accuracy (Precision), root mean square erroe of prediction (RMSEP), and root mean square erroe of cross validation (RMSECV).(RMSECV), which are the parameters for evaluating the model built by partial least squares method. The parameters R, R The parameters R, RMSE, Precision, RMSEP, and RMSECV for the four periods are good illustrations of the usefulness of modeling using partial least squares in the spectral detection of nitrogen in cotton leaves.
2 Results and Analysis
2.1Hyperspectral characteristics of cotton leaves in different periods
It can be seen from Fig. 1 that the hyperspectral characteristics of cotton leaves in different important fertility periods show a general consistent trend.A clear reflectance peak was observed in the green band (about 550 nm) and the absorption valley in the red band (about 680 nm) and the maximum spectral value is reached in the near-infrared band (>760 nm), and the trend gradually tends to level off. From the hyperspectral characteristics of cotton leaves in the four important reproductive stages, This phenomenon reaches a maximum at the reflection peak, and in the near-infrared region, the spectral features show a tendency to level off and are more pronounced in the wavelength band (760-1075 nm).
3 Conclusion
In this study, hyperspectral techniques were used to detect the nitrogen content of cotton leaves.
A prediction model based on the SPA-PLS method was developed by pre-processing the spectra of four important fertility stages of cotton leaves.The results showed that the minimum R-value of the model for the four fertility stages of cotton was 0.8032, the prediction accuracy was more than 0.9647, the maximum RMSEP was 0.2604, and the maximum RMSECV was 0.1414, and the prediction parameters all achieved good results.It shows that it is feasible to use the SPA-PLS method to build a model to predict the nitrogen content of cotton leaves, and this study provides a reference for using hyperspectral techniques to detect important nutrients in crops.
Maize nitrogen monitoring and diagnostic model for rational nitrogen application and efficient use of nitrogen fertilizer.
In this paper, hyperspectral data of cotton leaves in southern Xinjiang were collected, and the nitrogen content of cotton leaves was measured.
The nitrogen content of cotton leaves was measured, and the nitrogen data were analyzed using a combination of different algorithms such as PLS, normalization, SPA-PLS, and data centration. The nitrogen data and hyperspectral data of cotton leaves were simulated using a combination of different algorithms such as PLS, normalization, SPA-PLS, and data centering. The gradual regression of SPA-PLS was preferentially selected as the best method to reflect the nitrogen content of cotton leaves. It can reflect the hyperspectral response of cotton nitrogen more accurately, and provide the most accurate, fast and non-destructive detection of cotton nitrogen. It provides a theoretical method for accurate, rapid and nondestructive detection of cotton nitrogen.
1 Experimental part
1.1 Sample collection
The experiment was conducted in 2019 at the First Division Irrigation Experiment Station in Alar, Xinjiang.
The experimental field was designed with 18 plots and 3 treatments. Urea (N), calcium phosphate (P) and potassium sulfate (K) were applied to the experimental field.
(N), calcium dihydrogen phosphate (P), and potassium sulfate (K) were applied to cotton at regular intervals.
The fertilizer was applied on July 10, 2019 (bud stage), July 26, 2019 (flowering stage), and August 2019 (flowering stage). period), August 2, 2019 (boll stage), and August 26, 2019 (boll stage)
Nitrogen and hyperspectral data of cotton leaves were collected for four periods, each period 180 data were collected for each period, and a total of 720 samples were collected for the four periods.
1.2 Measurement of leaf nitrogen content
Cotton leaves were labeled on the plant with a plant nutrient meter to ensure that the measured nitrogen values of cotton leaves corresponded to the spectral values of the leaves to be measured, and then the nitrogen values were measured three times on each leaf. The nitrogen values were then measured three times on each leaf, and the three nitrogen values were averaged as the nitrogen value of the leaf.
1.3 Leaf hyperspectral data acquisition
The wavelength range was used in the experiment
325-1075 nm, and the number of sample spectra acquisition was set to 5 spectra per acquisition. The cotton leaves collected in the experiment were placed on a platform covered with a black baffle, and then the distance from the lens to the platform was adjusted to 20 cm to measure the spectral values of the desired leaves.
The 5 spectral data of each cotton leaf were averaged as the hyperspectral data of this leaf.
1.4 Modeling Method
In this study, PLS was used to model the correction and prediction sets for cotton leaf detection. The correlation coefficient R, root mean square erroe (RMSE), prediction accuracy (Prediction accuracy), and the number of leaves in the calibration and prediction sets were determined by the partial least squares method. (RMSE), prediction accuracy (Precision), root mean square error (RMSE), and root mean square erroe of prediction (RMSE). square erroe of prediction (RMSEP), cross-validation mean square erroe of prediction (RMSE), and cross-validation (RMSE), prediction accuracy (Precision), root mean square erroe of prediction (RMSEP), and root mean square erroe of cross validation (RMSECV).(RMSECV), which are the parameters for evaluating the model built by partial least squares method. The parameters R, R The parameters R, RMSE, Precision, RMSEP, and RMSECV for the four periods are good illustrations of the usefulness of modeling using partial least squares in the spectral detection of nitrogen in cotton leaves.
2 Results and Analysis
2.1Hyperspectral characteristics of cotton leaves in different periods
It can be seen from Fig. 1 that the hyperspectral characteristics of cotton leaves in different important fertility periods show a general consistent trend.A clear reflectance peak was observed in the green band (about 550 nm) and the absorption valley in the red band (about 680 nm) and the maximum spectral value is reached in the near-infrared band (>760 nm), and the trend gradually tends to level off. From the hyperspectral characteristics of cotton leaves in the four important reproductive stages, This phenomenon reaches a maximum at the reflection peak, and in the near-infrared region, the spectral features show a tendency to level off and are more pronounced in the wavelength band (760-1075 nm).
3 Conclusion
In this study, hyperspectral techniques were used to detect the nitrogen content of cotton leaves.
A prediction model based on the SPA-PLS method was developed by pre-processing the spectra of four important fertility stages of cotton leaves.The results showed that the minimum R-value of the model for the four fertility stages of cotton was 0.8032, the prediction accuracy was more than 0.9647, the maximum RMSEP was 0.2604, and the maximum RMSECV was 0.1414, and the prediction parameters all achieved good results.It shows that it is feasible to use the SPA-PLS method to build a model to predict the nitrogen content of cotton leaves, and this study provides a reference for using hyperspectral techniques to detect important nutrients in crops.
Study on the degree of damage to the Asian carlocust with hyperspectral remote sensing model
FAQ_HYPERSPECTAL IMAGER FAQ
Related Article
Discover Optosky’s air–ground collaborative hyperspectral solution for mine ecological restoration monitoring. Combining UAV hyperspectral imaging, fixed-point rotating-scan systems, and field spectroradiometers, it supports vegetation, soil, mineral alteration, and water environment monitoring for green mine evaluation and restoration acceptance.
Applications | Hyperspectral Mine Restoration Monitoring
Compare single InGaAs and CMOS+InGaAs detector architectures in hyperspectral imaging. Learn why both can cover 400–1700 nm, how visible-light-extended InGaAs works, and the trade-offs in calibration, spectral registration, and system integration for applications in agriculture, food safety, and industrial inspection.
Technology | Single InGaAs vs. CMOS+InGaAs in Hyperspectral Imaging
Discover how hyperspectral imaging combined with multi-feature fusion (PCA + 2DCOS + CARS) enables non-destructive three-level grading of chili pepper sunburn with 97.2% accuracy. Learn about the ATH3000 system and Optosky's customized solutions for horticultural crop quality inspection and damage classification.
Can Sunburned Chili Peppers Be Non-Destructively Graded? Hyperspectral Imaging ATH3000 Offers a New Answer
Discover the Optosky ATH9600 multi-lens UAV-borne multispectral imaging system—featuring 67° ultra-wide FOV, synchronous snapshot, 3.75 μm large pixel, and native DJI X-Port integration. Ideal for large-area farmland monitoring, orchard management, pest early warning, water quality inversion, and complex crop classification. Born for efficient, all-weather low-altitude remote sensing.
Born for Large-Area Remote Sensing! Optosky Releases the ATH9600 Multi-Lens UAV-Borne Multispectral Imaging System