A precise monitoring method for nitrogen content in wheat plants by optimizing the image vegetation index

Abstract

The present invention provides a method for accurately monitoring nitrogen content of wheat plants by optimizing the image vegetation index. The method comprises: step 1: obtaining drone images of the study area, and preprocessing the drone images to obtain corresponding orthophotos; step 2: extracting vegetation indexes from the orthophotos; step 3: constructing natural logarithmic symmetric transformation rules, and optimizing the vegetation index using the natural logarithmic symmetric transformation rules; step 4: dividing the optimized vegetation index into a training set and a test set; step 5: constructing a prediction model, and using the vegetation index in the training set and its corresponding wheat plant nitrogen content measurement value to train the prediction model to obtain a nitrogen content accurate monitoring model; step 6: inputting the vegetation index in the test set into the nitrogen content accurate monitoring model to obtain a nitrogen content prediction result. The technical solution proposed by the present invention has the characteristics of fast speed and high precision.

Description

Technical Field

[0001] The present invention relates to the field of agricultural engineering technology, in particular to the field of wheat plant nitrogen content estimation, and specifically to a method for accurately monitoring the nitrogen content of wheat plants by optimizing an image vegetation index. Background Art

[0002] Nitrogen is an essential nutrient for plant growth and has a significant impact on plant growth and development. In the agricultural field, accurate monitoring of plant nitrogen content can help farmers assess the nutritional status of plants and the amount of fertilizer applied, thereby improving the yield and quality of crops. In the field of environmental science, accurate monitoring of plant nitrogen content can promptly detect nitrogen pollution problems in soil and water, promptly assess soil quality and water pollution, and take corresponding control measures. Wheat is an important food crop in my country. As an essential element to ensure wheat yield and quality, accurate monitoring and reasonable application of nitrogen are particularly important for the healthy growth of wheat, the improvement of yield and quality, and the reduction of farmland environmental pollution and resource waste.

[0003] At present, domestic and foreign scholars have carried out a lot of research on the rapid monitoring and accurate prediction of wheat nitrogen content. From sampling analysis to non-destructive monitoring of near-ground, UAV, and satellite remote sensing, certain results have been achieved in sensitive band screening, vegetation index construction, prediction inversion method optimization and accuracy improvement (Wang Yuna et al., 2020; Shi et al., 2021; Guo et al., 2022), but there are differences in different data selection and methods. With the rapid development of agricultural modernization and smart agriculture, low-cost UAV remote sensing image data plays an increasingly important role in the rapid monitoring of crop nitrogen. Domestic and foreign researchers have improved the accuracy of nitrogen content monitoring by constructing different vegetation indices or comprehensive applications (Jia Dan and Chen Pengfei, 2020; Usha and Vasuki, 2022), but the impact of data distribution on monitoring accuracy is crucial (Guo et al., 2022). Therefore, how to optimize the data processing is very critical. Summary of the invention

[0004] In response to the problem that the distribution of image vegetation index is non-normal, which affects the accuracy of the wheat plant nitrogen content monitoring model, the present invention proposes a method for accurately monitoring the nitrogen content of wheat plants by optimizing the image vegetation index. By optimizing the vegetation index extracted from drone images, a random forest model is constructed to accurately monitor the nitrogen content of wheat plants, providing technical support and theoretical basis for rapid and accurate monitoring of nitrogen in farmland.

[0005] The present invention provides a method for accurately monitoring nitrogen content in wheat plants by optimizing an image vegetation index, comprising:

[0006] Step 1: Obtain UAV images of the study area, and pre-process the UAV images to obtain corresponding orthophotos;

[0007] Step 2: extracting vegetation index from the orthophoto;

[0008] Step 3: construct a natural logarithm symmetric transformation rule, and optimize the vegetation index using the natural logarithm symmetric transformation rule;

[0009] Step 4: Divide the optimized vegetation index into training set and test set;

[0010] Step 5: construct a prediction model, and use the vegetation index in the training set and its corresponding wheat plant nitrogen content measurement value to train the prediction model to obtain a nitrogen content accurate monitoring model;

[0011] Step 6: Input the vegetation index in the test set into the nitrogen content precise monitoring model to obtain the nitrogen content prediction result.

[0012] Furthermore, in step 1, the drone image is preprocessed to obtain a corresponding orthophoto, specifically including:

[0013] The acquired drone images of the study area are stitched together to obtain the image of the study area;

[0014] Orthorectification of the image of the study area based on ground control points;

[0015] Based on the grayscale value of the radiation calibration plate and combined with the reflection characteristics of the radiation calibration plate, a linear conversion formula is established. The grayscale value of the orthorectified study area image is converted into reflectivity using the linear conversion formula to obtain the corresponding orthophoto image.

[0016] Furthermore, in step 2, the extracted vegetation indices include the Normalized Difference Vegetation Index NDVI, the Enhanced Vegetation Index EVI, the Modified Soil Adjusted Vegetation Index MSAVI, the Chlorophyll Vegetation Index MTCI and the Red-Blue Vegetation Index RBI.

[0017] Furthermore, step 3 specifically includes:

[0018]

[0019] Among them, ω ij (x) is the vegetation index at the i-th row and j-th column of the image grid TIF data, ω i ' j (x) is the natural logarithm symmetric ratio conversion value of the vegetation index at the i-th row and j-th column of the image grid TIF data, k is the number of vegetation index categories; ε is a constant, which is half of the number of vegetation index categories k.

[0020] Furthermore, in step 5, the prediction model adopts a random forest model.

[0021] Beneficial effects of the present invention:

[0022] Based on the vegetation index extracted from low-altitude drone images, the vegetation index is optimized by using the natural logarithm symmetric transformation rule to construct a wheat plant nitrogen content monitoring model, thereby improving the monitoring accuracy. Compared with the prior art, the technical solution proposed in the present invention has the characteristics of fast speed and high precision. On the basis of existing image acquisition, by analyzing and converting the data, the characteristics of the data itself are fully exploited to achieve fast and high-precision monitoring of plant nitrogen content. Compared with the method of directly using the vegetation index to build the model, the determination coefficient R 2 It was improved by 14.10%, and the relative error E was reduced by 48.63%, achieving high-precision monitoring of nitrogen content in wheat plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic diagram of a flow chart of a method for accurately monitoring nitrogen content in wheat plants by optimizing the image vegetation index provided by an embodiment of the present invention;

[0024] Figure 2 The nitrogen content measurement value y of the wheat plant after optimizing the vegetation index provided by the embodiment of the present invention is m The predicted value of nitrogen content in wheat plants obtained by the monitoring model p Schematic diagram of the relationship between;

[0025] Figure 3 The measured value y of nitrogen content in wheat plants when the vegetation index is not optimized provided by the embodiment of the present invention is m Compared with the predicted value y' of nitrogen content in wheat plants obtained by the monitoring model p Schematic diagram of the relationship between them. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] like Figure 1 As shown, an embodiment of the present invention provides a method for accurately monitoring nitrogen content in wheat plants by optimizing the image vegetation index, comprising the following steps:

[0028] S101: Use the drone multispectral sensor to obtain drone images of the study area, and pre-process the drone images to obtain orthophoto data in raster TIF format.

[0029] Specifically, before the sowing of winter wheat in 2022 and during the jointing, heading and flowering stages of winter wheat in 2023, a DJI UAV M600 carrying a MicaSense RedEdge M multispectral sensor was used to obtain drone images of the experimental area from 10:00 to 14:00 in clear and cloudless weather. The data format was TIF and saved on an SD card.

[0030] Among them, the MicaSense RedEdge M multispectral sensor mainly includes five bands: blue light band (central wavelength 475nm, B), green light band (central wavelength 560nm, G), red light band (central wavelength 668nm, R), red edge band (central wavelength 717nm, Rededge), and near infrared band (central band 840, Nir). The UAV flew at an altitude of 30m, and the spatial resolution of the acquired image was 0.02m. Ground control points (GCP) were set at the four vertices and the middle position of the test area to accurately calibrate the flight position and altitude to reduce position errors.

[0031] The preprocessing of drone images includes:

[0032] S1011. Image stitching: First, copy the image data stored on the SD card to the computer, use pix4D software, and stitch the images based on the 3D model mode to obtain the image of the study area.

[0033] S1012, orthorectification: Based on the ground control points (GCP), the image of the study area obtained in step S1011 is accurately corrected to eliminate the position deviation caused by high flight speed, unstable posture, and image distortion. This step can make the position of the drone images of the study area obtained at different times accurately match.

[0034] S1013, radiation calibration: Based on the grayscale value of the radiation calibration plate and combined with the reflection characteristics of the radiation calibration plate, a linear conversion formula is established to convert the grayscale value of the orthorectified image into reflectivity to eliminate the influence of factors such as different time, solar altitude angle, and weather, thereby obtaining a precisely corrected orthophoto of the study area.

[0035] S102: extracting vegetation index from the orthophoto;

[0036] Specifically, based on the orthophoto in step S1013, a raster calculation method is used to extract five types of vegetation indices, and the data format is raster TIF. The vegetation indices are normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), modified soil adjusted vegetation index (MSAVI), chlorophyll vegetation index (MTCI) and red-blue vegetation index (RBI), and the calculation method is as follows:

[0037]

[0038]

[0039]

[0040]

[0041]

[0042] Where: BLUE , RED , NIR , REDedge Represent the spectral reflectance of the blue band, red band, near-infrared band and red-edge band of the orthophoto respectively.

[0043] S103: construct a natural logarithmic symmetric transformation rule to optimize the vegetation index in step S102.

[0044] Specifically, the conversion formula of the natural logarithm symmetry conversion rule is:

[0045]

[0046] Among them, ω ij (x) is the vegetation index at the i-th row and j-th column of the image grid TIF data, ω i ' j (x) is the natural logarithm symmetric ratio conversion value of the vegetation index at the i-th row and j-th column of the image grid TIF data, k is the number of vegetation index categories, and ε is a constant whose value is half of the number of vegetation index categories k. In this embodiment, k=5, ε=2.5.

[0047] Formula (6) is used to transform the five vegetation indices, namely NDVI, EVI, MSAVI, MTCI and RBI, to obtain the optimized vegetation indices, which are denoted as NDVI', EVI', MSAVI', MTCI' and RBI' respectively.

[0048] S104: dividing the optimized vegetation index into a training set and a test set;

[0049] Specifically, first, based on the measured nitrogen content of wheat plants, the corresponding NDVI', EVI', MSAVI', MTCI', and RBI' were extracted according to the above steps; then, the measured nitrogen content of wheat plants was set as the dependent variable y, with the unit of g / m 2 The corresponding vegetation indices NDVI', EVI', MSAVI', MTCI', and RBI' are the independent variables x, and the data are divided into training and test sets according to the ratio of 7:3.

[0050] S105: constructing a prediction model, and using the vegetation index in the training set and the corresponding wheat plant nitrogen content measurement value to train the prediction model to obtain a nitrogen content accurate monitoring model;

[0051] Specifically, the random forest model is used as the prediction model in the embodiment of the present invention.

[0052] S106: Input the vegetation index in the test set into the nitrogen content precise monitoring model to obtain the nitrogen content prediction result.

[0053] Specifically, the predicted value of nitrogen content in wheat plants output by the nitrogen content precise monitoring model based on the optimized vegetation index is recorded as y p , the measured value of nitrogen content in wheat plants is recorded as y m , Figure 2 for y m and p The relationship between the two is: p =0.84y m +2.42.

[0054] Furthermore, in order to compare with the prediction results of the unoptimized vegetation index, according to the method of steps S104 and S105, the corresponding NDVI, EVI, MSAVI, MTCI, and RBI are extracted based on the measured nitrogen content of wheat plants; the random forest model is still used as the prediction model, and the extracted vegetation index data such as NDVI, EVI, MSAVI, MTCI, and RBI are divided into a training set and a test set, and the random forest model is trained to obtain a nitrogen content monitoring model. The predicted value of the nitrogen content of wheat plants output by the nitrogen content monitoring model based on the unoptimized vegetation index is recorded as y' p , the measured value of nitrogen content in wheat plants is recorded as y m , Figure 3 for y m and y' p The relationship between the two is: y' p =0.78y m +3.96.

[0055] S107: Evaluation of the accuracy of the wheat plant nitrogen content monitoring model. The accuracy evaluation indicators are the mean absolute error (E) and the determination coefficient (R 2 ).

[0056]

[0057]

[0058] Where n is the number of samples of nitrogen content in wheat plants measured, ymi is the measured value of nitrogen content in wheat plants, y pi is the predicted value of nitrogen content in wheat plants, is the average value of the measured nitrogen content of wheat plants.

[0059] In this embodiment, after the image vegetation index is converted, y m With y p Coefficient of determination R 2 is 0.89, E is 1.50g / m 2 When the image vegetation index is not converted, y m With y' p Coefficient of determination R' 2 is 0.78, E' is 2.92g / m 2 .

[0060] S108: Compare and analyze the monitoring results of the nitrogen content of wheat plants obtained by the conversion method proposed in the present invention with the monitoring results of the nitrogen content of wheat plants obtained without conversion.

[0061] Specifically, the magnitude of the reduction in the mean absolute error E, ΔE, and the coefficient of determination R are used. 2 Increase range △R 2 To express.

[0062]

[0063] The nitrogen content monitoring results of wheat plants obtained by the conversion method proposed in the present invention are compared with those of wheat plants without conversion. 2 =14.10%, △E=-48.63%, that is, R 2 It increased by 14.10% and E decreased by 48.63%.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for accurately monitoring nitrogen content in wheat plants by optimizing the image vegetation index, characterized in that: include: Step 1: Obtain drone images of the study area and pre-process the drone images to obtain corresponding orthophotos; Step 2: extracting vegetation index from the orthophoto; Step 3: Constructing a natural logarithm symmetric transformation rule, and optimizing the vegetation index using the natural logarithm symmetric transformation rule; specifically including: Among them, ω ij (x) is the vegetation index at the i-th row and j-th column of the image grid TIF data, ω i ' j (x) is the natural logarithm symmetric ratio conversion value of the vegetation index at the i-th row and j-th column of the image grid TIF data, k is the number of vegetation index categories; ε is a constant, which is half of the number of vegetation index categories k; Step 4: Divide the optimized vegetation index into training set and test set; Step 5: Construct a prediction model, and use the vegetation index in the training set and its corresponding wheat plant nitrogen content measurement value to train the prediction model to obtain an accurate nitrogen content monitoring model; Step 6: Input the vegetation index in the test set into the nitrogen content precise monitoring model to obtain the nitrogen content prediction result.

2. The method for accurately monitoring nitrogen content in wheat plants by optimizing the image vegetation index according to claim 1, characterized in that: In step 1, the drone image is preprocessed to obtain the corresponding orthophoto, specifically including: The acquired drone images of the study area were stitched together to obtain the image of the study area; Orthorectification of the image of the study area based on ground control points; Based on the grayscale value of the radiation calibration plate and combined with the reflection characteristics of the radiation calibration plate, a linear conversion formula is established. The grayscale value of the orthorectified study area image is converted into reflectivity using the linear conversion formula to obtain the corresponding orthophoto image.

3. The method for accurately monitoring nitrogen content in wheat plants by optimizing the image vegetation index according to claim 1, characterized in that: In step 2, the extracted vegetation indices include the Normalized Difference Vegetation Index (NDVI), the Enhanced Vegetation Index (EVI), the Modified Soil Adjusted Vegetation Index (MSAVI), the Chlorophyll Vegetation Index (MTCI), and the Red-Blue Vegetation Index (RBI).

4. The method for accurately monitoring nitrogen content in wheat plants by optimizing the image vegetation index according to claim 1, characterized in that: In step 5, the prediction model adopts a random forest model.

Images