Cigar coating unmanned aerial vehicle monitoring method and device based on reflectivity mapping correction

Abstract

The invention relates to a cigar coating unmanned aerial vehicle monitoring method and device based on reflectivity mapping correction, and relates to the technical field of agricultural monitoring, and the method comprises the steps: obtaining first multispectral image data collected above a shading net in a cigar coating planting region; based on the first multispectral image data, extracting a eggplant feature data set; based on the SPAD measured value of the eggplant and a preset SPAD measured value of a reference sample, screening a paired sample set; and based on the first multispectral image data and the second multispectral image data corresponding to the paired sample set, establishing a reflectivity correction model and performing mapping correction on spectral features, and training a machine learning model based on the corrected spectral data and an SPAD measured value to obtain a monitoring model. According to the method, the real spectral characteristics of the eggplant under the shading net are effectively recovered, high-precision inversion of SPAD under the shading environment is realized, and a feasible path is provided for unmanned aerial vehicle monitoring of crops cultivated in the shading net.

Description

Technical Field

[0001] This invention relates to the technical field of agricultural monitoring, and in particular to a method and device for monitoring cigar wrappers using a drone based on reflectivity mapping correction. Background Technology

[0002] Cigar wrappers are one of the most important raw materials in cigar production, and their leaf color, thickness, and texture directly affect the appearance and quality of the cigar. Wrapper production typically employs shade cultivation to control light intensity, temperature, humidity, and airflow, resulting in high-quality raw materials with smooth leaf surfaces and uniform color. However, field monitoring under shade netting presents certain challenges. Traditional manual monitoring methods are time-consuming and labor-intensive, and struggle to achieve continuous and precise monitoring of large-scale wrapper growth. In contrast, unmanned aerial vehicle (UAV) remote sensing technology offers advantages in agricultural monitoring, including operational flexibility, high spatiotemporal resolution, and wide monitoring range. It has been widely applied in crop growth monitoring and nutrient diagnosis, providing a new potential direction for tobacco cultivation monitoring.

[0003] Currently, existing drone monitoring methods generally involve flying directly above crops and collecting image data, then analyzing the images to obtain crop growth information.

[0004] However, in cigar wrapper production, where plants are primarily grown under shade nets, conventional drone monitoring methods are difficult to apply directly. The limited space under the shade nets makes it impossible to meet the requirements for stable drone flight and image acquisition. If images of the wrapper plants are acquired above the shade nets, the spectral absorption and scattering characteristics of the nets distort the plant's reflectance, resulting in images that cannot directly reflect the plant's true reflectance. This, in turn, affects the accuracy of retrieving plant physiological parameters based on spectral information, failing to meet the need for accurate monitoring during cigar wrapper production. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and apparatus for monitoring and correcting cigar wrappers for drones based on reflectivity mapping correction, in order to solve at least one of the above-mentioned technical problems.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: Firstly, this application provides a method for monitoring cigar wrappers using a drone based on reflectivity mapping correction, employing the following technical solution: A method for monitoring cigar wrappers by drones based on reflectivity mapping correction includes: Acquire first multispectral image data of the cigar wrapper growing area, wherein the first multispectral image data represents images of cigar wrapper plants collected above the shading net; Feature extraction is performed on the first multispectral image data to obtain the feature dataset of the eggplant plant. The feature dataset includes spectral features, vegetation index, texture features and morphological features. Based on the SPAD measured values ​​of the eggplant plants and the SPAD measured values ​​of the preset reference samples, a paired sample set is selected. The reference samples represent the second multispectral image data of tobacco samples of the same variety as the eggplant plants and cultivated under no shading conditions, collected under a shading net. Based on the first and second multispectral image data corresponding to the paired sample set, a reflectance correction model is established, and the spectral features in the feature dataset are mapped and corrected based on the reflectance correction model to obtain the corrected spectral data. Based on the corrected spectral data and the corresponding SPAD measured values ​​of the eggplant leaves, a machine learning model is trained to obtain a monitoring model. The monitoring model is used to invert and monitor the SPAD measured values ​​of the eggplant leaves under the shading net.

[0007] The beneficial effects of this invention are as follows: By pairing spectral data of cigar wrappers collected under shading conditions with unshaded reference samples of similar physiological state, and using machine learning algorithms to establish a nonlinear correction model, the accuracy and stability of chlorophyll content (SPAD value) inversion are significantly improved. This effectively restores the true spectral characteristics of cigar wrapper plants under shading nets, improves the reliability and generalization ability of the correction model, achieves high-precision SPAD inversion under shading conditions, provides a feasible path for UAV monitoring of crops cultivated under shading nets, and enhances the informatization and intelligentization level of the cigar wrapper production process.

[0008] Based on the above technical solution, the present invention can be further improved as follows.

[0009] Furthermore, acquiring the first multispectral image data of the cigar wrapper growing area includes: Obtain raw multispectral image patches of cigar wrapper growing areas collected by drones, wherein the raw multispectral image patches include blue light, green light, red light, red edge and near-infrared bands; Radiometric correction and image stitching are performed on the original multispectral image blocks to generate the first multispectral image data of the cigar wrapper growing area.

[0010] The beneficial effects of adopting the above-mentioned further scheme are: acquiring original multispectral image blocks containing blue light, green light, red light, red edge and near-infrared bands, and obtaining the first multispectral image data through radiometric correction and image stitching processing, which can ensure image quality, ensure comparability between different batches of data, and provide a reliable basis for subsequent accurate analysis of the characteristics of eggplant leaves.

[0011] Furthermore, the step of selecting a paired sample set based on the measured SPAD values ​​of the eggplant peel plants and the measured SPAD values ​​of a preset reference sample includes: Based on the SPAD measured value of the eggplant plant and the SPAD measured value of the preset reference sample, the error value between the eggplant plant and the reference sample is calculated. Based on the error value and the preset error threshold, the reference samples are filtered to obtain multiple filtered reference samples; Each selected reference sample is mapped to its corresponding eggplant seedling to obtain a paired sample set.

[0012] The beneficial effects of adopting the above-mentioned further scheme are as follows: Through error calculation and threshold screening mechanisms based on SPAD measured values, accurate physiological state matching between shaded and unshaded samples is achieved. This ensures that paired samples have highly consistent chlorophyll content and nitrogen nutrient levels, laying a reliable physiological foundation for subsequent spectral mapping correction. It effectively eliminates interference to the spectrum caused by differences in the plant's own physiological state, allowing the constructed reflectance correction model to more purely learn and correct spectral deviations caused by the shading net, thereby improving the accuracy of the correction model.

[0013] Furthermore, the step of establishing a reflectance correction model based on the first and second multispectral image data corresponding to the paired sample set includes: Based on the first multispectral image data of the paired sample set, the first multispectral reflectance of the eggplant plant is determined; Based on the second multispectral image data corresponding to the first multispectral image data in the paired sample set, the second multispectral reflectance of the reference sample corresponding to the eggplant plant is determined. Based on the first multispectral reflectance and the second multispectral reflectance, a random forest regression model is trained to obtain the reflectance correction model; wherein, the random forest regression model is configured to learn the spectral mapping relationship from shading conditions to unshading conditions in order to perform nonlinear correction of spectral differences.

[0014] The beneficial effects of adopting the above-mentioned further scheme are as follows: First, determine the first multispectral reflectance of the eggplant peel plant and the second multispectral reflectance of the reference sample, and then train a random forest regression model based on the two to obtain a reflectance correction model. This model can learn the spectral mapping relationship from shading conditions to no-shading conditions, perform nonlinear correction on spectral differences, effectively capture the spectral relationship between the eggplant peel under shading net conditions and the eggplant core without shading, improve the accuracy of reflecting the actual spectrum of the eggplant peel, and effectively correct the signal affected by the light intensity reduction or spectral deviation caused by the shading net.

[0015] Furthermore, the monitoring model is trained based on the corrected spectral data and the corresponding SPAD measured values ​​of the eggplant peel plants, including: Based on the corrected spectral data and the corresponding SPAD values ​​of the eggplant peel plants, an XGBoost regression model was trained to obtain the monitoring model.

[0016] The beneficial effects of adopting the above-mentioned further scheme are: by using the corrected spectral data and the corresponding measured SPAD values ​​of the eggplant leaves to train the XGBoost regression model to obtain the monitoring model, it is possible to realize the inversion monitoring of the measured SPAD values ​​of eggplant leaves under the shading net. Moreover, the gradient boosting and regularization mechanism of XGBoost can optimize the feature selection and nonlinear fitting ability of the model, and improve the inversion accuracy.

[0017] Furthermore, based on the corrected spectral data and the corresponding measured SPAD values ​​of the eggplant peel plants, an XGBoost regression model is trained to obtain the monitoring model, which includes: Calculate the target vegetation index based on the corrected spectral data; Based on the corrected spectral data of the eggplant plants, the target vegetation index, and SPAD measurements, the monitoring model was obtained by training the model and optimizing its features through the gradient boosting and regularization mechanism of the XGBoost algorithm.

[0018] The beneficial effects of adopting the above-mentioned further scheme are: by calculating the target vegetation index and combining it with the corrected spectral data and SPAD measured values, and by using the gradient boosting and regularization mechanism of the XGBoost algorithm for model training and feature optimization, the accuracy of the monitoring model in retrieving and monitoring the SPAD value of eggplant leaves under shading nets can be improved.

[0019] Furthermore, it also includes: The monitoring model is evaluated based on at least one of the indicators of coefficient of determination, root mean square error, and mean absolute error, and the evaluation results are obtained. in, ; ; ; Among them, R 2 As the coefficient of determination, The root mean square error, The mean absolute error, Let be the measured SPAD value of the i-th sample. Let n be the SPAD prediction value for the i-th sample, and n be the number of samples. This is the average of all measured values ​​for SPAD.

[0020] The beneficial effects of adopting the above-mentioned further scheme are as follows: First multispectral image data of the cigar wrapper plants above the shading net are obtained; spectral features are extracted to obtain a feature dataset; a paired sample set is selected based on SPAD measured values; a reflectance correction model is established to correct the spectral feature mapping to obtain corrected spectral data; an XGBoost regression model is trained to obtain a monitoring model; finally, at least one of the indicators of coefficient of determination, root mean square error, and mean absolute error is used to evaluate the monitoring model. This effectively restores the true spectral characteristics of the cigar wrapper plants, improves the model's reliability and generalization ability, achieves high-precision SPAD inversion under shading conditions, and verifies the effect of multispectral data mapping correction on improving model stability and prediction accuracy. It provides a feasible path for UAV monitoring of crops cultivated under shading nets and enhances the informatization and intelligentization level of the cigar wrapper production process.

[0021] Secondly, this application provides a cigar wrapper drone monitoring device based on reflectivity mapping correction, employing the following technical solution: A cigar wrapper drone monitoring device based on reflectivity mapping correction includes: The acquisition module is used to acquire first multispectral image data of the cigar wrapper planting area, wherein the first multispectral image data represents the image of the cigar wrapper plant collected above the shading net. The feature extraction module is used to extract features from the first multispectral image data to obtain the feature dataset of the eggplant plant. The feature dataset includes spectral features, vegetation index, texture features and morphological features. The screening module is used to screen a paired sample set based on the SPAD measured value of the eggplant plant and the SPAD measured value of a preset reference sample. The reference sample represents the second multispectral image data of tobacco samples of the same variety as the eggplant plant and cultivated under no shading conditions, collected under a shading net. The mapping correction module is used to establish a reflectance correction model based on the first multispectral image data and the second multispectral image data corresponding to the paired sample set, and to map and correct the spectral features in the feature dataset based on the reflectance correction model to obtain the corrected spectral data. The training module is used to train a machine learning model based on the corrected spectral data and the corresponding SPAD measured values ​​of the eggplant leaves to obtain a monitoring model. The monitoring model is used to invert and monitor the SPAD measured values ​​of the eggplant leaves under the shading net.

[0022] Thirdly, this application provides an electronic device that adopts the following technical solution: An electronic device includes a memory and a processor, wherein the memory stores a computer program capable of being loaded by the processor and executing the cigar wrapper drone monitoring method based on reflectivity mapping correction as described in any of the first aspects.

[0023] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium storing a computer program capable of being loaded by a processor and executing the cigar wrapper drone monitoring method based on reflectivity mapping correction as described in any one of the first aspects.

[0024] Additional aspects and advantages of this application will be set forth in part in the description which follows, and will become apparent from the description or may be learned by practice of this application. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating a method for monitoring cigar wrappers by drones based on reflectivity mapping correction, provided in one embodiment of the present invention. Figure 2 This is a diagram showing the correction result of UAV multispectral reflectance mapping based on a random forest model, provided in one embodiment of the present invention. Figure 3 An uncorrected SPAD inversion result of cigar wrapper plants based on the XGBoost model, provided as an embodiment of the present invention; Figure 4 A modified SPAD inversion result of cigar wrapper plants based on the XGBoost model provided in one embodiment of the present invention; Figure 5 A schematic diagram of a cigar wrapper drone monitoring device based on reflectivity mapping correction provided in one embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of an electronic device provided in one embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0028] This application provides a method for monitoring cigar wrappers by drones based on reflectivity mapping correction. This method can be executed by an electronic device, which can be a server or a mobile terminal device. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. The mobile terminal device can be a laptop, a desktop computer, etc., but is not limited to these.

[0029] like Figure 1 As shown, a method for monitoring cigar wrappers by drones based on reflectivity mapping correction includes: S1, acquire the first multispectral image data of the cigar wrapper planting area, wherein the first multispectral image data represents the image of the cigar wrapper plant collected above the shading net. In this embodiment of the application, acquiring the first multispectral image data of the cigar wrapper growing area includes: Obtain raw multispectral image patches of cigar wrapper growing areas collected by drones, wherein the raw multispectral image patches include blue light, green light, red light, red edge and near-infrared bands; Radiometric correction and image stitching are performed on the original multispectral image blocks to generate the first multispectral image data of the cigar wrapper growing area.

[0030] In this embodiment of the application, field sampling is carried out during the cigar harvesting period, relevant information is recorded, and a drone flight mission is performed to acquire multispectral images containing blue light, green light, red light, red edge and near-infrared light bands. After radiometric correction and image stitching, field multispectral images of the cigar harvesting period are obtained.

[0031] For agricultural monitoring, drones such as the DJI P4M multispectral drone can be selected. This image patch contains blue, green, red, red-edge, and near-infrared bands, providing rich spectral information for subsequent analysis. When acquiring images, certain requirements exist for factors such as flight altitude, solar altitude angle, lateral overlap, and longitudinal overlap. For example, setting the flight altitude appropriately ensures surface resolution, the solar altitude angle is greater than a certain angle to reduce the influence of illumination, and the lateral and longitudinal overlap rates are set to appropriate ratios to ensure image integrity and accuracy. After acquiring the original multispectral image patch, radiometric correction and image stitching are performed. Radiometric correction can be done using specialized software and calibration plates, such as grayscale calibration plates with reflectances of 10%, 25%, 35%, 50%, and 70%, to calibrate reflectance and eliminate radiometric errors, ensuring comparability between different batches of data. Image stitching combines multiple image patches into a complete first multispectral image dataset, which can be accomplished using software such as Agisoft Metashape.

[0032] S2, extract features from the first multispectral image data to obtain the feature dataset of the eggplant plant, which includes spectral features, vegetation index, texture features and morphological features; In this embodiment, the feature dataset includes spectral features, vegetation indices, texture features, and morphological features. Spectral features reflect the reflectance characteristics of plants in different spectral bands and can be extracted through band analysis of image data. Vegetation indices are calculated based on spectral data; common vegetation indices include the Differential Vegetation Index (DVI), Chlorophyll Index (LCI), Standardized Differential Vegetation Index (NDVI), and Green Standardized Differential Vegetation Index (GNDVI). The calculation formulas for vegetation indices are shown in Table 1.

[0033] Table 1: Among them, B, G, R, RE, and NIR are the reflectances of blue, green, red, red-edge, and near-infrared in their respective spectral bands.

[0034] Texture features, such as roughness and contrast, reflect the texture information of plant images and can be extracted using image processing algorithms. Morphological features describe the shape and size of the plant and can be obtained through image recognition and measurement techniques. The extraction of these features provides comprehensive data support for subsequent analysis and model training.

[0035] S3. Based on the SPAD measured value of the eggplant plant and the SPAD measured value of the preset reference sample, a paired sample set is selected. The reference sample represents the second multispectral image data of tobacco samples of the same variety as the eggplant plant and cultivated under no shading conditions, collected under the shading net. In this embodiment of the application, the step of selecting a paired sample set based on the measured SPAD value of the eggplant peel plant and the measured SPAD value of a preset reference sample includes: Based on the SPAD measured value of the eggplant plant and the SPAD measured value of the preset reference sample, the error value between the eggplant plant and the reference sample is calculated. Based on the error value and the preset error threshold, the reference samples are filtered to obtain multiple filtered reference samples; Each selected reference sample is mapped to its corresponding eggplant seedling to obtain a paired sample set.

[0036] In the above implementation, the SPAD values ​​of the eggplant-coated plants and reference samples are obtained. SPAD values ​​can be measured using a professional instrument, such as a SPAD-502 Plus chlorophyll meter. During measurement, normally growing plants free from pests and diseases are selected. Measurements are taken on functional leaves, avoiding the midrib. For each leaf, a five-point sampling method is used to select five measurement points, and the average value is taken as the SPAD value for that leaf. Three leaves are measured from each plant, and the average value is taken as the representative value for the plant. Measurements are conducted during a suitable time period in sunny weather to minimize the influence of ambient light variations. Then, the error value between the eggplant-coated plants and the reference samples is calculated based on these measured values. The error value can be calculated using a relative error method, i.e., the ratio of the difference between the two SPAD values ​​to the SPAD value of the reference sample. Next, reference samples are screened based on the error value and a preset error threshold. The preset error threshold can be set according to actual conditions, for example, 5%. When the error value is within this threshold, the reference sample is considered to have similar chlorophyll and nitrogen states to the corresponding eggplant-coated plant and can be used as a paired sample. Finally, a mapping relationship is established between each selected reference sample and the corresponding eggplant seedling to obtain a paired sample set.

[0037] S4. Based on the first multispectral image data and the second multispectral image data corresponding to the paired sample set, a reflectance correction model is established, and the spectral features in the feature dataset are mapped and corrected based on the reflectance correction model to obtain the corrected spectral data. In this embodiment of the application, the step of establishing a reflectance correction model based on the first multispectral image data and the second multispectral image data corresponding to the paired sample set includes: Based on the first multispectral image data of the paired sample set, the first multispectral reflectance of the eggplant plant is determined; Based on the second multispectral image data corresponding to the first multispectral image data in the paired sample set, the second multispectral reflectance of the reference sample corresponding to the eggplant plant is determined. Based on the first multispectral reflectance and the second multispectral reflectance, a random forest regression model is trained to obtain the reflectance correction model; wherein, the random forest regression model is configured to learn the spectral mapping relationship from shading conditions to unshading conditions in order to perform nonlinear correction of spectral differences.

[0038] In the above implementation, based on a paired sample set, multispectral reflectance data of the eggplant peel and core are extracted and used as input and target variables, respectively. A random forest (RF) regression model is constructed to achieve spectral mapping correction between the two types of image data. Through the nonlinear feature learning and multi-decision tree ensemble mechanism of the RF algorithm, spectral differences under different lighting conditions or shading environments are corrected, thereby obtaining an eggplant peel-corrected image consistent with the reflectance characteristics of the eggplant core, providing a consistent spectral basis for subsequent vegetation index calculation and physiological parameter inversion.

[0039] During training, the model is configured to learn the spectral mapping from shading conditions to unshading conditions in order to perform nonlinear correction of spectral differences. Through training, the model is able to capture the influence of the shading net on the spectrum, thereby accurately mapping and correcting the spectral features in the feature dataset to obtain corrected spectral data.

[0040] S5. Based on the corrected spectral data and the corresponding measured SPAD values ​​of the eggplant leaves, a machine learning model is trained to obtain a monitoring model. The monitoring model is used to invert and monitor the measured SPAD values ​​of the eggplant leaves under the shading net.

[0041] In this embodiment of the application, the XGBoost regression model can be used for training.

[0042] Furthermore, the monitoring model is obtained by training a machine learning model based on the corrected spectral data and the corresponding SPAD measured values ​​of the eggplant peel plants, including: Based on the corrected spectral data and the corresponding SPAD values ​​of the eggplant peel plants, an XGBoost regression model was trained to obtain the monitoring model.

[0043] Furthermore, based on the corrected spectral data and the corresponding measured SPAD values ​​of the eggplant peel plants, an XGBoost regression model is trained to obtain the monitoring model, including: Calculate the target vegetation index based on the corrected spectral data; Based on the corrected spectral data of the eggplant plants, the target vegetation index, and SPAD measurements, the monitoring model was obtained by training the model and optimizing its features through the gradient boosting and regularization mechanism of the XGBoost algorithm.

[0044] In this embodiment, firstly, a target vegetation index is calculated based on the corrected spectral data. The target vegetation index can be one of the common vegetation indices mentioned earlier, calculated using the band reflectance in the corrected spectral data. Then, based on the corrected spectral data of the eggplant leaves, the target vegetation index, and the measured SPAD values, model training and feature optimization are performed using the gradient boosting and regularization mechanism of the XGBoost algorithm. The XGBoost algorithm is an efficient gradient boosting algorithm that iteratively trains multiple weak classifiers to gradually improve the model's prediction accuracy. During training, the algorithm filters and optimizes features to find those that have the greatest impact on the measured SPAD values. During training, the model training set and prediction set can be divided in a certain ratio, such as 8:2, to evaluate the model's performance. The final monitoring model can be used to invert and monitor the measured SPAD values ​​of eggplant leaves under shading nets.

[0045] Optional, also includes: The monitoring model is evaluated based on at least one of the indicators of coefficient of determination, root mean square error, and mean absolute error, and the evaluation results are obtained. in, ; ; ; Among them, R 2 As the coefficient of determination, The root mean square error, The mean absolute error, Let be the measured SPAD value of the i-th sample. Let n be the SPAD prediction value for the i-th sample, and n be the number of samples. This is the average of all measured values ​​for SPAD.

[0046] In the embodiments of this application, the coefficient of determination (R) 2 The value of 1 reflects the model's fit to the data; the closer it is to 1, the better the model fit. The root mean square error (RMSE) and mean absolute error (MAE) measure the magnitude of the error between the model's predicted values ​​and the measured values; the smaller their values, the higher the model's prediction accuracy. By calculating and analyzing these indicators, we can understand and monitor the model's performance, allowing for further optimization and improvement.

[0047] The following example illustrates the entire process based on a field experiment.

[0048] The field experiments included nitrogen fertilizer trials and density trials: Density Experiment: Under the same nitrogen application rate (10 kg / mu), four planting density treatments were set up, namely D1: 36,000 plants / hm.2 D2: 31,500 plants / hm 2 D3: 24,000 plants / hm 2 , D4: 19500 plants / hm 2 The row and plant spacing is R1: 0.90 m × 0.30 m, R2: 0.90 m × 0.35 m, R3: 1.00 m × 0.40 m, and R4: 1.10 m × 0.45 m.

[0049] Nitrogen fertilizer experiment: Under the same planting density (density: 31,500 plants / hm2, row spacing 95cm×35cm), three nitrogen fertilizer application treatments were set up: N1: 6-8 kg / mu, N2: 8-10 kg / mu, and N3: 10-12 kg / mu.

[0050] Figure 2 The paper demonstrates the prediction results of using a random forest (RF) model to map and correct multispectral data of eggplant peel and core under shading netting. The predicted values ​​for the five bands (B, G, R, RE, NIR) are highly consistent with the measured values, and the R values ​​for all bands are... 2 All values ​​were between 0.79 and 0.87, indicating that the model can effectively capture the spectral relationship between the eggplant wrapper and the unshaded eggplant core under shading conditions. RMSE and MAE indices show that the prediction error for each band is small, and the error between the predicted and measured values ​​is basically within an acceptable range. Overall, after mapping correction, the signals that were originally affected by the light intensity reduction or spectral deviation caused by the shading net are effectively corrected, improving the accuracy of reflecting the actual spectrum of the eggplant wrapper.

[0051] Figure 3 This paper presents the uncorrected SPAD inversion results for cigar wrapper plants based on the XGBoost model. Under shading conditions, the model's inversion performance is significantly limited. As shown in the figure, overfitting occurred during model training, but the model's prediction set determination coefficient R0 remains high. 2 The value was only 0.048, indicating a significant discrepancy between the predicted and measured values ​​(RMSE=5.118, MAE=3.721). The points in the scatter plot deviated significantly from the 1:1 reference line, suggesting that in shading environments, the correlation between the vegetation index calculated based on multispectral reflectance and the actual SPAD value of cigar wrappers is weak. The shading netting causes changes in the light environment, leading to spectral curve distortion, making it difficult for the directly established XGBoost model to accurately reflect chlorophyll content.

[0052] Figure 4 This shows the SPAD inversion results based on the corrected XGBoost model. After correction using multispectral data matching mapping, the model performance is significantly improved, R... 2The prediction error was significantly reduced to 0.570 (RMSE=2.188, MAE=1.743). The scatter plot shows that the predicted points are more closely distributed around the 1:1 line, and the fitting trend is significantly closer to the ideal state. This indicates that by matching and mapping the spectral data of the eggplant peel under shading conditions with the data of the eggplant core without shading, the interference of the shading net on the spectral characteristics is effectively reduced.

[0053] Overall, the correction method successfully restored the true physiological state of the plants. The significant difference in model performance before and after correction further demonstrates that shading conditions distort the relationship between vegetation index and SPAD value. Through cross-environmental data matching and mapping correction, spectral bias caused by shading can be effectively eliminated, restoring the intrinsic link between vegetation index and chlorophyll content. This provides reliable technical support for the accurate inversion and remote sensing monitoring of crop physiological parameters under special light conditions.

[0054] This method pairs spectral data of cigar wrappers collected under shaded conditions with unshaded reference samples of similar physiological state, and uses machine learning algorithms to establish a nonlinear correction model, significantly improving the accuracy and stability of chlorophyll content (SPAD value) retrieval. It effectively restores the true spectral characteristics of cigar wrapper plants under shaded netting, improves the reliability and generalization ability of the correction model, achieves high-precision SPAD retrieval under shaded conditions, provides a feasible path for UAV monitoring of crops cultivated under shaded netting, and enhances the informatization and intelligentization level of cigar wrapper production.

[0055] Figure 5 A schematic diagram of a cigar wrapper drone monitoring device 200 based on reflectivity mapping correction is shown.

[0056] like Figure 5 As shown, a cigar wrapper drone monitoring device 200 based on reflectivity mapping correction mainly includes: The acquisition module 201 is used to acquire first multispectral image data of the cigar wrapper planting area, wherein the first multispectral image data represents the image of the cigar wrapper plant collected above the shading net. Feature extraction module 202 is used to extract features from the first multispectral image data to obtain a feature dataset of the eggplant plant, the feature dataset including spectral features, vegetation index, texture features and morphological features; The screening module 203 is used to screen a paired sample set based on the SPAD measured value of the eggplant plant and the SPAD measured value of a preset reference sample. The reference sample represents the second multispectral image data of tobacco samples of the same variety as the eggplant plant and cultivated under no shading conditions, collected under a shading net. The mapping correction module 204 is used to establish a reflectance correction model based on the first multispectral image data and the second multispectral image data corresponding to the paired sample set, and to map and correct the spectral features in the feature dataset based on the reflectance correction model to obtain the corrected spectral data. Training module 205 is used to train a machine learning model based on the corrected spectral data and the measured SPAD values ​​of the corresponding eggplant leaves to obtain a monitoring model. The monitoring model is used to invert and monitor the measured SPAD values ​​of eggplant leaves under the shading net.

[0057] Optionally, module 201 is used specifically for: Obtain raw multispectral image patches of cigar wrapper growing areas collected by drones, wherein the raw multispectral image patches include blue light, green light, red light, red edge and near-infrared bands; Radiometric correction and image stitching are performed on the original multispectral image blocks to generate the first multispectral image data of the cigar wrapper growing area.

[0058] Optional, the filtering module 203 is specifically used for: Based on the SPAD measured value of the eggplant plant and the SPAD measured value of the preset reference sample, the error value between the eggplant plant and the reference sample is calculated. Based on the error value and the preset error threshold, the reference samples are filtered to obtain multiple filtered reference samples; Each selected reference sample is mapped to its corresponding eggplant seedling to obtain a paired sample set.

[0059] Optionally, the mapping correction module 204 includes: The first determining submodule is used to determine the first multispectral reflectance of the eggplant plant based on the first multispectral image data of the paired sample set. The second determining submodule determines the second multispectral reflectance of the reference sample corresponding to the eggplant plant based on the second multispectral image data corresponding to the first multispectral image data in the paired sample set. The training submodule is used to train a random forest regression model based on the first multispectral reflectance and the second multispectral reflectance to obtain the reflectance correction model; wherein the random forest regression model is configured to learn the spectral mapping relationship from shading conditions to unshading conditions in order to perform nonlinear correction of spectral differences.

[0060] Optional, training submodule, specifically used for: Based on the corrected spectral data and the corresponding SPAD values ​​of the eggplant peel plants, an XGBoost regression model was trained to obtain the monitoring model.

[0061] Optionally, the training submodule is used to train an XGBoost regression model based on the corrected spectral data and the corresponding measured SPAD values ​​of the eggplant peel plants. When obtaining the monitoring model, it is specifically used for: Based on the corrected spectral data and the corresponding SPAD values ​​of the eggplant peel plants, an XGBoost regression model was trained to obtain the monitoring model.

[0062] Optionally, the cigar wrapper drone monitoring device 200 based on reflectivity mapping correction also includes an evaluation module, which is used for: The monitoring model is evaluated based on at least one of the indicators of coefficient of determination, root mean square error, and mean absolute error, and the evaluation results are obtained. in, ; ; ; Among them, R 2 As the coefficient of determination, The root mean square error, The mean absolute error, Let be the measured SPAD value of the i-th sample. Let n be the SPAD prediction value for the i-th sample, and n be the number of samples. This is the average of all measured values ​​for SPAD.

[0063] In one example, the module in any of the above devices may be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), or one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0064] For example, when modules in a device can be implemented via a processing element scheduler, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor capable of calling programs. Alternatively, these modules can be integrated together as a system-on-a-chip (SOC).

[0065] In this application, various objects such as messages / information / devices / network elements / systems / apparatus / actions / operations / processes / concepts may be named. It is understood that these specific names do not constitute a limitation on the relevant objects. The names may be changed depending on the scenario, context, or usage habits. The understanding of the technical meaning of the technical terms in this application should be mainly determined from their functions and technical effects embodied / performed in the technical solution.

[0066] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0067] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0068] Figure 6 This is a structural block diagram of an electronic device 300 according to an embodiment of this application.

[0069] like Figure 6 As shown, the electronic device 300 includes a processor 301 and a memory 302, and may further include one or more of an information input / output (I / O) interface 303, a communication component 304, and a communication bus 305.

[0070] The processor 301 controls the overall operation of the electronic device 300 to complete all or part of the steps in the aforementioned method for monitoring cigar wrappers by drones based on reflectivity mapping correction. The memory 302 stores various types of data to support the operation of the electronic device 300. This data may include, for example, instructions for any application or method operating on the electronic device 300, as well as application-related data. The memory 302 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as one or more of Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0071] I / O interface 303 provides an interface between processor 301 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 304 is used to test wired or wireless communication between electronic device 300 and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 304 may include a Wi-Fi component, a Bluetooth component, and an NFC component.

[0072] The communication bus 305 may include a path for transmitting information between the aforementioned components. The communication bus 305 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 305 may be divided into an address bus, a data bus, a control bus, etc.

[0073] The electronic device 300 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the cigar wrapper drone monitoring method based on reflectivity mapping correction given in the above embodiments.

[0074] The following describes the computer-readable storage medium provided in the embodiments of this application. The computer-readable storage medium described below can be referred to in correspondence with the above-described method for monitoring cigar wrappers by drones based on reflectivity mapping correction.

[0075] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for monitoring cigar wrappers by drones based on reflectivity mapping correction.

[0076] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0077] The terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0078] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions claimed in this application.

Claims

1. A method for monitoring cigar wrappers using a drone based on reflectivity mapping correction, characterized in that, include: Acquire first multispectral image data of the cigar wrapper growing area, wherein the first multispectral image data represents images of cigar wrapper plants collected above the shading net; Feature extraction is performed on the first multispectral image data to obtain the feature dataset of the eggplant plant. The feature dataset includes spectral features, vegetation index, texture features and morphological features. Based on the SPAD measured values ​​of the eggplant plants and the SPAD measured values ​​of the preset reference samples, a paired sample set is selected. The reference samples represent the second multispectral image data of tobacco samples of the same variety as the eggplant plants and cultivated under no shading conditions, collected under a shading net. Based on the first and second multispectral image data corresponding to the paired sample set, a reflectance correction model is established, and the spectral features in the feature dataset are mapped and corrected based on the reflectance correction model to obtain the corrected spectral data. Based on the corrected spectral data and the corresponding SPAD measured values ​​of the eggplant leaves, a machine learning model is trained to obtain a monitoring model. The monitoring model is used to invert and monitor the SPAD measured values ​​of the eggplant leaves under the shading net.

2. The method for monitoring cigar wrappers by drones based on reflectivity mapping correction according to claim 1, characterized in that, The acquisition of the first multispectral image data of the cigar wrapper growing area includes: The original multispectral image patch of the cigar wrapper growing area was acquired by the drone. The original multispectral image patch includes blue light, green light, red light, red edge and near-infrared bands. Radiometric correction and image stitching are performed on the original multispectral image blocks to generate the first multispectral image data of the cigar wrapper growing area.

3. The method for monitoring cigar wrappers by drones based on reflectivity mapping correction according to claim 1, characterized in that, The process of selecting a paired sample set based on the measured SPAD values ​​of the eggplant peel plants and the measured SPAD values ​​of a preset reference sample includes: Based on the SPAD measured value of the eggplant plant and the SPAD measured value of the preset reference sample, the error value between the eggplant plant and the reference sample is calculated. Based on the error value and the preset error threshold, the reference samples are filtered to obtain multiple filtered reference samples; Each selected reference sample is mapped to its corresponding eggplant seedling to obtain a paired sample set.

4. The method for monitoring cigar wrappers by drones based on reflectivity mapping correction according to claim 3, characterized in that, The step of establishing a reflectance correction model based on the first and second multispectral image data corresponding to the paired sample set includes: Based on the first multispectral image data of the paired sample set, the first multispectral reflectance of the eggplant plant is determined; Based on the second multispectral image data corresponding to the first multispectral image data in the paired sample set, the second multispectral reflectance of the reference sample corresponding to the eggplant plant is determined. Based on the first multispectral reflectance and the second multispectral reflectance, a random forest regression model is trained to obtain the reflectance correction model; wherein, the random forest regression model is configured to learn the spectral mapping relationship from shading conditions to unshading conditions in order to perform nonlinear correction of spectral differences.

5. The method for monitoring cigar wrappers by drones based on reflectivity mapping correction according to claim 1, characterized in that, The monitoring model is obtained by training a machine learning model based on the corrected spectral data and the corresponding measured SPAD values ​​of the eggplant peel plants, including: Based on the corrected spectral data and the corresponding SPAD values ​​of the eggplant peel plants, an XGBoost regression model was trained to obtain the monitoring model.

6. The method for monitoring cigar wrappers by drones based on reflectivity mapping correction according to claim 5, characterized in that, The monitoring model is obtained by training an XGBoost regression model based on the corrected spectral data and the corresponding measured SPAD values ​​of the eggplant peel plants, including: Calculate the target vegetation index based on the corrected spectral data; Based on the corrected spectral data of the eggplant plants, the target vegetation index, and SPAD measurements, the monitoring model was obtained by training the model and optimizing its features through the gradient boosting and regularization mechanism of the XGBoost algorithm.

7. The method for monitoring cigar wrappers by drones based on reflectivity mapping correction according to claim 6, characterized in that, Also includes: The monitoring model is evaluated based on at least one of the indicators of coefficient of determination, root mean square error, and mean absolute error, and the evaluation results are obtained. in ; ； ； Among them, R 2 As the coefficient of determination, The root mean square error, The mean absolute error, Let be the measured SPAD value of the i-th sample. Let n be the SPAD prediction value for the i-th sample, and n be the number of samples. This is the average of all measured values ​​for SPAD.

8. A cigar wrapper drone monitoring device based on reflectivity mapping correction, characterized in that, include: The acquisition module is used to acquire first multispectral image data of the cigar wrapper planting area, wherein the first multispectral image data represents the image of the cigar wrapper plant collected above the shading net. The feature extraction module is used to extract features from the first multispectral image data to obtain the feature dataset of the eggplant plant. The feature dataset includes spectral features, vegetation index, texture features and morphological features. The screening module is used to screen a paired sample set based on the SPAD measured value of the eggplant plant and the SPAD measured value of a preset reference sample. The reference sample represents the second multispectral image data of tobacco samples of the same variety as the eggplant plant and cultivated under no shading conditions, collected under a shading net. The mapping correction module is used to establish a reflectance correction model based on the first multispectral image data and the second multispectral image data corresponding to the paired sample set, and to map and correct the spectral features in the feature dataset based on the reflectance correction model to obtain the corrected spectral data. The training module is used to train a machine learning model based on the corrected spectral data and the corresponding SPAD measured values ​​of the eggplant leaves to obtain a monitoring model. The monitoring model is used to invert and monitor the SPAD measured values ​​of the eggplant leaves under the shading net.

9. An electronic device, characterized in that, Includes a processor, which is coupled to a memory; The processor is configured to execute a computer program stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Includes a computer program or instructions that, when run on a computer, cause the computer to perform the method as described in any one of claims 1-7.

Images