CN120236192A - A method for constructing a high-throughput comprehensive scoring model for salt tolerance of pea based on drones - Google Patents
A method for constructing a high-throughput comprehensive scoring model for salt tolerance of pea based on drones Download PDFInfo
- Publication number
- CN120236192A CN120236192A CN202510239077.2A CN202510239077A CN120236192A CN 120236192 A CN120236192 A CN 120236192A CN 202510239077 A CN202510239077 A CN 202510239077A CN 120236192 A CN120236192 A CN 120236192A
- Authority
- CN
- China
- Prior art keywords
- index
- pea
- salt tolerance
- vegetation index
- spad
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/54—Extraction of image or video features relating to texture
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/58—Extraction of image or video features relating to hyperspectral data
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Software Systems (AREA)
- Remote Sensing (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Astronomy & Astrophysics (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
本发明涉及农业生长检测和无人机遥感技术领域，具体的涉及一种基于无人机的豌豆高通量耐盐性综合评分模型的构建方法。本发明通过无人机搭载传感器采集目标区域豌豆图像数据，根据传感器图像数据提取其株高和冠层覆盖度的结构信息，以及纹理信息和光谱信息，通过光谱信息组合得到植被指数，再利用机器学习算法通过纹理信息和植被指数估测豌豆生物量和SPAD，最后利用株高、冠层覆盖度、生物量和SPAD四种指标构建豌豆耐盐性综合评分模型PSTS。本发明通过传感器数据融合和合适的算法构建评分模型可以高通量并准确的筛选豌豆耐盐性品种。
The present invention relates to the fields of agricultural growth detection and unmanned aerial vehicle remote sensing technology, and specifically to a method for constructing a high-throughput salt tolerance comprehensive scoring model for peas based on unmanned aerial vehicles. The present invention collects pea image data in a target area by using a sensor mounted on an unmanned aerial vehicle, extracts structural information of plant height and canopy coverage, as well as texture information and spectral information according to the sensor image data, obtains a vegetation index by combining spectral information, and then estimates pea biomass and SPAD by using a machine learning algorithm through texture information and vegetation index, and finally constructs a pea salt tolerance comprehensive scoring model PSTS using four indicators of plant height, canopy coverage, biomass and SPAD. The present invention constructs a scoring model by using sensor data fusion and a suitable algorithm to screen pea salt-tolerant varieties with high throughput and accuracy.
Description
Technical Field
The invention belongs to the technical field of agricultural growth detection and unmanned aerial vehicle remote sensing, and particularly relates to a method for constructing a pea high-flux salt tolerance comprehensive scoring model based on an unmanned aerial vehicle.
Background
The saline-alkali soil is wide in area, has huge utilization space, and is an important means for improving and utilizing the saline-alkali soil by screening salt-tolerant crop varieties. Pea is one of important edible bean crops, has short growth period and certain salt tolerance, and timely and accurately screens pea salt tolerance crop varieties and plays an important role in improving saline-alkali soil. However, the traditional screening of salt-tolerant crop varieties is time-consuming, labor-consuming and susceptible to human factors. With the development of unmanned aerial vehicles and remote sensing technology, unmanned aerial vehicles can carry various sensors due to high portability, are not easily influenced by cloud cover, and are increasingly used for monitoring phenotypic traits of crops.
The sensors commonly carried by unmanned aerial vehicles include RGB sensors, MS sensors, TIR sensors, hyperspectral sensors and laser radar sensors. Lidar and hyperspectral sensors are generally capable of providing more information, but are costly and difficult to apply on a large scale. TIR sensors can provide canopy temperature information, often used to monitor soil moisture spatiotemporal changes. RGB sensors and MS sensors are favored by many researchers because of their low cost and simplicity of operation. However, the research based on unmanned aerial vehicles is remained in the estimation of biological phenotype traits obtained through observation or measurement, and a reliable identification method based on unmanned aerial vehicles for salt tolerance in physiological phenotypes is lacked.
Disclosure of Invention
Aiming at the technical problems, the invention provides a method for constructing a pea high-flux salt tolerance comprehensive scoring model based on an unmanned aerial vehicle, which is simple and feasible, has more reliable identification result, can realize high-flux and accurate evaluation of pea variety salt tolerance, and improves the efficiency of salt tolerance pea germplasm screening.
The technical scheme of the invention is as follows:
The invention provides a pea high-flux salt tolerance comprehensive scoring model construction method based on an unmanned aerial vehicle, which comprises the steps of collecting pea images of a target area, extracting structural information, texture information and spectrum information in the pea images, estimating plant height and canopy coverage through the structural information, obtaining vegetation index information related to biomass and SPAD through the spectrum information, estimating the biomass and SPAD of the peas through the texture information and/or the vegetation index information by utilizing a machine learning algorithm, calculating salt tolerance coefficients of the plant height, the canopy coverage, the biomass and the SPAD, carrying out normalization treatment on the salt tolerance coefficients, carrying out principal component analysis to obtain principal components, determining principal component weights, and constructing a salt tolerance comprehensive scoring model according to the principal component weights and membership function values of the principal components, wherein the size of the salt tolerance factors is positively related to salt tolerance.
In one embodiment, the pea image acquisition period is a pea flowering period to a grouting period.
In one embodiment, the drone mount comprises an RGB sensor and/or a multispectral sensor to acquire the target area pea image, the pea image comprising an RGB image and/or an MS image.
In one embodiment, the texture information includes entropy, contrast, inverse variance, correlation, dominance, gray scale mean, and energy.
In one embodiment, the biomass-related vegetation index information in the RGB sensor includes WI (Wo Beike index), GLI (green leaf area index), VARI (visible light atmospheric impedance vegetation index), exR (hyper red index), exB (hyper blue index), IPCA (principal component analysis index), CIVE (color vegetation index), SPAD-related vegetation index information includes INT (color intensity index), WI (Wo Beike index), exG (hyper green index), exB (hyper blue index), IPCA (principal component analysis index), MRBVI (modified red blue vegetation index), VARI (visible light atmospheric impedance vegetation index).
In one embodiment, the biomass-related vegetation index information in the MS sensor comprises CVI (chlorophyll vegetation index), CI (chlorophyll index), GDVI (generalized differential vegetation index), EVI2 (dual-band enhanced vegetation index), EVI (enhanced vegetation index), GARI (green anti-atmospheric vegetation index), GLI (green leaf area index), BNDVI (blue normalized differential vegetation index), GRVI (green specific vegetation index), MCARI1 (improved chlorophyll absorption reflection index), MCARI2 (improved chlorophyll absorption reflection index 2), SAVI (soil-adjusted vegetation index), OSAVI (optimized soil-adjusted vegetation index), MCARI1/OSAVI, SPAD-related vegetation index information comprises CVI (chlorophyll vegetation index), CCCI (canopy content index), GNDVI (green normalized differential vegetation index), NDRE (normalized differential red edge index), PNDVI (projected normalized differential vegetation index), GSAVI (green soil-adjusted vegetation index), GOSAVI (green optimized soil-adjusted normalized red edge index), 3938 (normalized red edge), MCARI 25 (improved land-modified vegetation index), MCARI1/OSAVI, SPAD-related vegetation index information comprises CVI (chlorophyll vegetation index), CCCI (chlorophyll content index), GNDVI (crown content index), GDVI (generalized differential vegetation index).
In one embodiment, the machine learning algorithm uses Catboost algorithm to estimate pea biomass and Light GBM algorithm to estimate pea SPAD.
In one embodiment, R 2 and RMSE are used to assess the biomass and SPAD robustness of the peas.
In one embodiment, the salt tolerance comprehensive scoring model is psts= Σ (i=1, n) [ U (X ij)×wi ], where PSTS represents a salt tolerance factor, U (X ij) represents a membership function value, X ij represents an ith principal component value of a jth cell, and w i represents a weight of the ith principal component.
The invention also provides a screening method of the salt tolerance pea varieties, which uses the unmanned aerial vehicle-based pea high-throughput salt tolerance comprehensive scoring model to evaluate the salt tolerance of peas and selects the pea varieties with high comprehensive salt tolerance evaluation.
Compared with the prior art, the invention has the beneficial effects that:
1. The five salt-tolerant pea varieties are estimated by the pea salt-tolerant comprehensive scoring model, four salt-tolerant pea varieties are screened out by ground measurement data, and four varieties among the five varieties are consistent with 4 varieties screened out by the ground measurement data, so that the salt tolerance of the pea varieties can be evaluated with high accuracy by the scoring model constructed by sensor data fusion and a proper algorithm.
2. The method can realize the salt tolerance evaluation of pea varieties with high flux based on the model constructed by the unmanned aerial vehicle, and improves the efficiency of salt tolerance pea germplasm screening.
Drawings
FIG. 1 is a scatter plot of best results of estimating pea biomass (RM+ Catboost) and SPAD (RM+light GBM) for saline-alkali land and control land, wherein (a) is estimating pea biomass scatter plot for control land based on RGB+MS data and Catboost algorithm, (b) is estimating pea biomass scatter plot for saline-alkali land based on RGB+MS data and Catboost algorithm, (c) is estimating pea SPAD scatter plot for control land based on RGB+MS data and Light GBM algorithm, (d) is estimating pea SPAD scatter plot for saline-alkali land based on RGB+MS data and Light GBM algorithm;
Fig. 2 is a visual graph of ground measurement data and pea salt tolerance comprehensive score based on an unmanned plane-constructed model, wherein (a) is a pea salt tolerance comprehensive score display graph based on the ground measurement data, and (b) is a pea salt tolerance comprehensive score display graph based on the model estimation data.
Detailed Description
The invention provides a method for constructing a pea high-flux salt tolerance comprehensive scoring model based on an unmanned aerial vehicle.
In the invention, different varieties of peas are sown in a target area, wherein the sowing depth is 4-6cm, and the sowing row spacing is 30-50cm. Providing more favorable growth conditions for pea growth, improving the emergence rate and seedling quality, and preventing inconsistent plant growth caused by sowing difference from affecting salt tolerance evaluation.
The growth process of peas includes a seedling stage, a flowering stage, a pod bearing stage, a filling stage and a maturation stage, the flowering stage to the filling stage being an important stage in the formation of pea yield. According to the method, pea images are collected from the flowering period to the grouting period of the peas, so that the accuracy of screening salt-tolerant varieties can be improved, and the evaluation result is objective and real. As one embodiment, the harvest time period is a flowering or pod stage.
According to the method, the pea image of the target area is acquired through unmanned aerial vehicle aerial photography, and the structural information, the texture information and the spectrum information in the pea image are extracted. As one implementation mode, the flying height of the unmanned aerial vehicle is set to be 10-25m, the flying speed is 1-3m/s, and the aerial photographing is photographing at equal time intervals. The growth heights of different pea varieties are also different, and the unmanned aerial vehicle can perform close-up shooting above the pea plants at the flight height, so that the detail capturing is facilitated. The speed and the time interval photographing can completely capture the structural information, the texture information and the spectrum information of the peas in the whole target area. As an implementation mode, the longitudinal overlapping rate of the aerial images of the unmanned aerial vehicle is not less than 60%, the transverse overlapping rate is not less than 50%, and information of pea varieties in a target area is prevented from being omitted or lost.
In the invention, an unmanned aerial vehicle-mounted sensor is adopted to collect pea images. As an embodiment, the unmanned aerial vehicle comprises an RGB sensor and/or a multispectral sensor for acquiring the pea image of the target area, and an RGB image and/or an MS image is obtained. Before the multispectral image is acquired, a standard reflector image is acquired for subsequent radiation calibration.
In the invention, unmanned aerial vehicle image processing is to splice acquired RGB and MS images, wherein DN values are converted into reflectivity values when the MS images are spliced. As one embodiment, the stitching process includes geometric correction, image alignment, construction of dense point clouds, generation of a digital surface model, generation of a digital terrain model, and acquisition of an orthographic image. In one embodiment, canopy coverage assessment is represented by the ratio of vegetation area to ROI area in the background-removed RGB image, and plant height assessment is represented by the maximum value of the crop surface model.
In the invention, plant height and canopy coverage are estimated through structural information, vegetation index information related to biomass and SPAD is obtained through spectral information, and then biomass and SPAD of peas are estimated through texture information and/or vegetation index information by utilizing a machine learning algorithm.
Wherein the texture information can reflect the canopy structure and structural features. As one embodiment, the texture information includes entropy, contrast, inverse variance, correlation, dominance, gray mean, and energy. The vegetation index is formed by combining spectral information of different wave bands, and can effectively reflect the growth condition of crops. As one embodiment, the biomass-related vegetation index information in the RGB sensor comprises WI, GLI, VARI, exR, exB, IPCA, CIVE and the SPAD-related vegetation index information comprises INT, WI, exG, exB, IPCA, MRBVI, VARI. As one embodiment, the biomass-related vegetation index information in the MS sensor comprises CVI, CI, GDVI, EVI, EVI, GARI, GLI, BNDVI, GRVI, MCARI1, MCARI2, SAVI, OSAVI, MCARI1/OSAVI, and the SPAD-related vegetation index information comprises CVI, CCCI, GNDVI, NDRE, PNDVI, GSAVI, GOSAVI, NREI, NNRI, MNDI, MEVI, MCARI1, MTCI, GDVI.
In the invention, the machine learning algorithm is one or more of Catboost, light GBM, GPR or SVM. The peas were evaluated for biomass and SPAD robustness using R 2 and RMSE. As one embodiment, pea biomass is estimated using rgb+ms data and Catboost algorithm, and pea SPAD is estimated using rgb+ms data and Light GBM algorithm.
In the invention, the plant height, canopy coverage, biomass and Salt Tolerance Coefficient (STC) of SPAD are calculated, and in one implementation mode, the salt tolerance coefficient refers to the ratio of a property value after salt treatment to a control property value. And carrying out normalization processing on the salt tolerance coefficient, and carrying out principal component analysis on the salt tolerance coefficient after normalization processing to obtain two principal components. And (3) performing membership function analysis on the two main components to obtain membership function values, and constructing a salt tolerance comprehensive scoring model by the weights of the two main components and the membership function values. The salt tolerance of peas can be predicted by using the salt tolerance comprehensive scoring model constructed by the method to obtain the salt tolerance factor (PSTS value). Wherein, the size of the salt tolerance factor is positively correlated with the salt tolerance.
In the present invention, the membership function calculation formula is U (X ij)＝(Xi,j-Xi,min)/(Xi,max-Xj,min), i=1, 2, & n
Wherein U represents a membership function value, X i,j represents an ith principal component value of a jth cell, X i,min represents a minimum value of the ith principal component, and X i,max represents a maximum value of the ith principal component.
In the invention, the weight of the main component is calculated by the formula w i =pi/Σ (i=1, n) pi, i=1, 2
Where w i represents the weight of the ith principal component, and Pi represents the contribution rate of the ith principal component.
In the invention, the calculation formula of the salt tolerance factor of the salt tolerance comprehensive scoring model is as follows:
PSTS=∑(i=1,n)[U(Xij)×wi]
Wherein PSTS represents a salt tolerance factor, U (X ij) represents a membership function value, X ij represents an ith principal component value of a jth cell, and w i represents a weight of the ith principal component.
The invention also provides a screening method of the salt tolerance pea varieties, which uses the unmanned aerial vehicle-based pea high-throughput salt tolerance comprehensive scoring model to evaluate the salt tolerance of peas and selects the pea varieties with high comprehensive salt tolerance evaluation.
The present invention will be described in detail below with reference to examples for the purpose of making the objects, technical solutions and advantages of the present invention more apparent, but they should not be construed as limiting the scope of the present invention.
The materials, reagents and the like used in the following examples, unless otherwise specified, are all commercially available, and the reagents, consumables and the like according to the present invention are usually carried out under conventional conditions or under conditions recommended by the company, unless otherwise specified.
Example 1
1. Research field arrangement
The research sites are divided into saline-alkali land parcels and common land parcels (or referred to as control land parcels), the common land parcels are in the northwest area of Tangshan city in Hebei province (north latitude 39.36, east longitude 118.11), the saline-alkali land parcels are in the Qingtao mountain city and mustine county in Hebei province (north latitude 39.43 east longitude 118.9), the two lands belong to temperate and monsoon climates, and the two lands are uniform except for different soil, such as climate, field land setting, variety, cultivation management measures and the like.
Both plots were sown with 14 pea varieties from all over the country, with a random block design, with three replicates for each variety. The soil loosening and fertilization are carried out before sowing, and the length of each cell is 5m, the width of each cell is 2m, 5 rows are arranged in total, and the row spacing is 40cm. Sowing is carried out on 13 days of 3 months of 2023, 120 seeds are sown in each row, and the sowing depth is about 5 cm. Insecticide is sprayed every half month, sprinkling is carried out every 7-10 days, manual weed removal is carried out regularly, and peas are harvested at 2023, 6 and 12 days.
2. Unmanned aerial vehicle data acquisition
The flowering period to the filling period are important periods for pea yield formation, so that the unmanned aerial vehicle of the Dajiang eidolon 4pro ((DJITechnology, shenzhen, china) and the Dajiang eidolon 4 multispectral edition (DJI Technology, shenzhen, china) is used for carrying out flight tasks on the day of 28 (flowering period) and the day of 10 (pod bearing period) of 2023 respectively, in order to ensure the data quality, the unmanned aerial vehicle respectively acquires a visible light image (RGB) and a multispectral image (MS) at 10:00-14:00 noon, and acquires a standard reflecting plate image before acquiring the multispectral image so as to carry out subsequent radiation calibration, wherein the flying height of the unmanned aerial vehicle is set to be 15m, the flying speed is set to be 2m/s, the unmanned aerial vehicle is set to take photos at equal time intervals, and the longitudinal overlapping rate and the transverse overlapping rate of two plots are respectively set to be 85% and 80%.
3. Unmanned aerial vehicle image processing
And splicing the acquired RGB and MS images by using Pix4d software, generating a digital surface model by geometric correction, image alignment and construction of dense point clouds, and generating a digital terrain model to obtain an orthographic image. When stitching MS images, the DN values are converted to reflectivity values by radiometric scaling with images of known reflectivity. Since the resolution of the RGB image is higher than that of the MS image, structural information of canopy coverage and plant height and texture information are extracted by using the RGB image, and spectral information is extracted by using the MS image. The RGB images are spliced and then input to the Arc Map pro for cutting, a threshold segmentation method is used for removing the soil background, each cell is marked as a specific area, a unique identifier is assigned, and the corresponding relationship between pea varieties and identifiers of two land areas is shown in table 1.
TABLE 1 correspondence table of pea varieties and identifiers
3.1 Estimation of pea canopy coverage and plant height
Estimating Canopy coverage (Canopy coverage) by using the ratio of the vegetation area to the ROI area in the background-removed RGB image as shown in equation 1, and estimating plant height by using the maximum value of the crop surface model as shown in equation 2:
Canopy coverage=Pcanopy/Ptotal (1)
Where P canopy represents the number of pixels in the pea canopy and P total represents the total number of pixels in the sampled plot.
CSM=DSM-DTM (2)
Where CSM represents the crop surface model, DSM represents the digital surface model, and DTM represents the digital terrain model.
3.2 Estimation of pea biomass and SPAD
The texture information can reflect the structure and the structural characteristics of the canopy, the image after background removal is input into ENVI 5.3, the sliding window and the sliding wavelength are respectively set to 7 multiplied by 7, and eight texture information including entropy, contrast, inverse variance, relativity, dominance, gray average value and energy are extracted altogether. The vegetation index is formed by combining spectral information in different wave bands, and can effectively reflect the growth condition of crops. The invention mainly utilizes Catboost, light GBM, GPR and SVM four machine learning methods to measure texture information and vegetation index information for estimating pea biomass and SPAD values. For fair and complete comparison of the estimation of pea Above Ground Biomass (AGB) and yield by the different algorithms, the four algorithms were each randomly selected 80% of the samples as training dataset and the remaining 20% as test dataset. The accuracy of the estimates was tested using a five-fold cross-validation test, thereby ensuring that all samples were independently used for validation.
In the invention, R 2 and RMSE are adopted to evaluate the pea biomass and the robustness of SPAD, and the lower the R 2 value is, the better the estimated performance is represented by the lower RMSE. The calculation method of R 2 and RMSE is shown in the formula 3 and the formula 4:
Wherein n represents the total number of samples, X i and Respectively representing the measured value and the estimated AGB or SPAD of each sample; Mean of measured AGB or SPAD is shown.
The invention extracts eight kinds of texture information (Contrast, correlation, DISSIMILARITY, entropy, homogeneity, mean, variance, energy) and 7 kinds of vegetation index information (wherein the biomass is WI, GLI, VARI, exR, exB, IPCA, CIVE and the SPAD is INT, WI, exG, exB, IPCA, MRBVI, VARI) with high Correlation with biomass and SPAD based on RGB images acquired by unmanned aerial vehicle data, and respectively estimates pea biomass and SPAD based on four machine learning methods (Catboost, light GBM, GPR, SVM). The RGB estimation results are shown in tables 2 and 3 below:
table 2RGB sensor estimation of pea biomass results
| Catboost | LightGBM | GPR | SVM | ||
| Common land block | R2 | 0.68 | 0.50 | 0.58 | 0.50 |
| RMSE | 2.35 | 2.77 | 2.68 | 2.95 | |
| Saline-alkali soil block | R2 | 0.54 | 0.48 | 0.59 | 0.48 |
| RMSE | 1.01 | 0.88 | 0.92 | 1.07 |
Table 3RGB sensor estimation of pea SPAD results
| Catboost | LightGBM | GPR | SVM | ||
| Common land block | R2 | 0.31 | 0.20 | 0.29 | 0.22 |
| RMSE | 3.53 | 3.32 | 3.49 | 3.81 | |
| Saline-alkali soil block | R2 | 0.19 | 0.38 | 0.26 | 0.25 |
| RMSE | 2.97 | 2.41 | 2.78 | 2.87 |
According to the invention, 14 vegetation index information with higher correlation with biomass and SPAD (wherein the biomass is CVI, CI, GDVI, EVI, EVI, GARI, GLI, BNDVI, GRVI, MCARI1, MCARI2, SAVI, OSAVI, MCARI1/OSAVI, the SPAD is CVI, CCCI, GNDVI, NDRE, PNDVI, GSAVI, GOSAVI, NREI, NNRI, MNDI, MEVI, MCARI, MTCI, GDVI) is extracted based on MS images acquired by unmanned aerial vehicle data, and pea biomass and SPAD are estimated based on four machine learning methods (Catboost, lightGBM, GPR, SVM) respectively. The MS estimation results are shown in tables 4 and 5 below:
Table 4MS sensor estimation of pea biomass results
| Catboost | LightGBM | GPR | SVM | ||
| Common land block | R2 | 0.58 | 0.56 | 0.64 | 0.51 |
| RMSE | 2.64 | 2.62 | 2.48 | 2.92 | |
| Saline-alkali soil block | R2 | 0.72 | 0.62 | 0.63 | 0.59 |
| RMSE | 0.78 | 0.76 | 0.87 | 0.88 |
Table 5MS sensor estimation of pea SPAD results
| Catboost | LightGBM | GPR | SVM | ||
| Common land block | R2 | 0.50 | 0.49 | 0.51 | 0.40 |
| RMSE | 2.85 | 3.03 | 2.86 | 3.16 | |
| Saline-alkali soil block | R2 | 0.40 | 0.55 | 0.42 | 0.26 |
| RMSE | 2.47 | 2.11 | 2.37 | 2.68 |
The biomass and SPAD values of peas were estimated based on four machine learning methods (Catboost, lightGBM, GPR, SVM) using fusion data of RGB and MS sensors, respectively, resulting in the estimation results shown in tables 6 and 7 below:
Table 6rgb+ms sensor estimation of pea biomass results
| Catboost | LightGBM | GPR | SVM | ||
| Common land block | R2 | 0.67 | 0.57 | 0.68 | 0.50 |
| RMSE | 2.41 | 2.62 | 2.33 | 2.95 | |
| Saline-alkali soil block | R2 | 0.73 | 0.63 | 0.65 | 0.57 |
| RMSE | 0.78 | 0.83 | 0.84 | 0.97 |
Table 7RGB+MS sensor estimation pea SPAD results
| Catboost | LightGBM | GPR | SVM | ||
| Common land block | R2 | 0.59 | 0.64 | 0.53 | 0.48 |
| RMSE | 2.79 | 2.59 | 2.73 | 2.97 | |
| Saline-alkali soil block | R2 | 0.43 | 0.57 | 0.41 | 0.35 |
| RMSE | 2.44 | 2.07 | 2.39 | 2.62 |
From the above study, the estimation results of two calculation methods, catboost and Light GBM, are better. The evaluation results of the RGB+MS data and Catboost algorithm for evaluating the pea biomass of the saline-alkali land block and the pea biomass of the control land block and the evaluation results of the RGB+MS data and the Light GBM algorithm for evaluating the pea SPAD of the saline-alkali land block and the pea SPAD of the control land block are the best, and the scatter diagram is shown in figure 1.
3.3 Establishment of PSTS
In order to realize the evaluation of the pea salt tolerance, a pea salt tolerance comprehensive scoring model is constructed based on four indexes of plant height, biomass, canopy coverage and SPAD, and a salt tolerance factor (PSTS value) is calculated, wherein the larger the PSTS value is, the higher the salt tolerance is represented.
Firstly, calculating a Salt Tolerance Coefficient (STC) of each property, wherein the salt tolerance coefficient refers to the ratio of a property value after salt treatment to a control property value, and a calculation formula is shown in a formula 5. And carrying out normalization processing on the salt tolerance coefficient, carrying out principal component analysis on the salt tolerance coefficient after normalization processing to obtain principal components FAC1 and FAC2 with two characteristic values larger than 1, wherein the contribution rates of the two principal components in each cell are shown in a table 8, and calculating the weight of the cell according to the contribution rate of the principal component by using a formula 6. And then, the two main components are subjected to membership function (see formula 7) analysis to obtain corresponding membership function values of the two main components, wherein the membership function values are shown in a table 10, finally, a salt tolerance comprehensive scoring model is constructed according to the weights and the membership function values, and the salt tolerance factors (PSTS values) of peas are calculated and obtained as shown in a table 11.
STC=salt treatment trait value/normal trait value (5)
wi=pi/∑(i=1,n)pi,i=1,2,...,n (6)
Where w i represents the weight of the ith principal component, and Pi represents the contribution rate of the ith principal component.
U(Xij)＝(Xi,j-Xi,min)/(Xi,max-Xj,min),i=1,2,...,n (7)
U represents a membership function value, X i,j represents an ith principal component value of a jth cell, X i,min represents a minimum value of the ith principal component, and X i,max represents a maximum value of the ith principal component.
PSTS=∑(i=1,n)[U(Xij)×wi] (8)
Wherein U represents a membership function value, and X ij represents an ith main component value of the jth cell.
Table 8 contribution ratio of two main components in each cell
TABLE 9 weights of two principal Components
| Main component | wi |
| FAC1 | 0.59 |
| FAC2 | 0.41 |
Table 10 membership function values of two principal components for each cell
Comparative example 1 ground measurement data acquisition
And measuring the plant height, SPAD and biomass data of the two land parcels on the same day acquired by the unmanned aerial vehicle. Each cell randomly selects 5 plants, and the average value of the distances from the ground to the top ends of the plants is measured as the plant height data. 10 plants were randomly selected and chlorophyll content of the upper and middle leaves was measured using a SPAD-502plus chlorophyll meter. And finally, three plants are randomly selected, roots are removed, and the overground biomass is measured by weighing.
Test example 1
The salt tolerance factors (PSTS values) of different pea varieties in each cell of the saline-alkali land block are estimated based on the pea salt tolerance comprehensive scoring model constructed by unmanned aerial vehicle measurement data, the salt tolerance factors of different pea varieties in each cell of the saline-alkali land block are estimated based on ground measurement data in comparative example 1 by adopting the PSTS construction method of the invention, and the calculation results of the salt tolerance factors of the invention and comparative example 1 are shown in Table 11:
table 11 comparative example 1 ground measurement and salt tolerance factor for unmanned aerial vehicle evaluation of the present invention
An illustrative graph of the results of the calculations according to Table 11 is shown in FIG. 2. As can be seen from fig. 2, the square frame marked by the unmanned aerial vehicle evaluation of the invention shows five varieties with the best salt tolerance, namely 20231-2, 0804-5, middle pea 03 (repeated twice) and Tang pea 3, respectively, and the square frame marked by the ground measurement data of comparative example 1 shows five varieties with the best salt tolerance, namely 0804-5, 20231-2, long pea 15, middle pea 03 and Tang pea 3, respectively. The observation shows that four varieties of varieties which are screened out based on unmanned aerial vehicle estimation and ground measurement data of comparative example 1 and have higher salt tolerance are consistent, and the accuracy is high, so that the model obtained through the sensor data fusion and the proper algorithm of the invention can accurately realize high-flux salt tolerance evaluation on pea varieties, and the efficiency of salt tolerance pea germplasm screening is improved.
The foregoing description is only illustrative of the present invention and is not intended to limit the scope of the invention, and all equivalent structures or equivalent processes or direct or indirect application in other related arts are included in the scope of the present invention.
Claims (10)
1. The method for constructing the pea high-flux salt tolerance comprehensive scoring model based on the unmanned aerial vehicle is characterized by comprising the following steps of:
Acquiring pea images of a target area, and extracting structural information, texture information and spectrum information in the pea images;
Obtaining vegetation index information related to biomass and SPAD through spectrum information, and estimating biomass and SPAD of peas through texture information and/or vegetation index information by using a machine learning algorithm;
Calculating the plant height, canopy coverage, biomass and salt tolerance coefficient of SPAD, carrying out normalization treatment on the salt tolerance coefficient, carrying out principal component analysis to obtain principal components, determining principal component weights, and constructing a salt tolerance comprehensive scoring model according to the principal component weights and membership function values of the principal components, wherein the size of the salt tolerance factor is positively correlated with the salt tolerance.
2. The method for constructing a salt tolerance composite scoring model according to claim 1, wherein the pea image acquisition period is a pea flowering period to a grouting period.
3. The method for constructing a salt tolerance comprehensive scoring model according to claim 1, wherein the unmanned aerial vehicle comprises an RGB sensor and/or a multispectral sensor to acquire a pea image of the target area, and the pea image comprises an RGB image and/or an MS image.
4. The method of claim 1, wherein the texture information comprises entropy, contrast, inverse variance, correlation, dominance, gray scale mean, and energy.
5. The method of claim 3, wherein the vegetation index information related to the biomass in the RGB sensor comprises wobbe index, green leaf area index, visible light atmospheric resistance vegetation index, hyper-red index, hyper-blue index, principal component analysis index, color vegetation index, SPAD related vegetation index information comprises color intensity index, wobbe index, hyper-green index, hyper-blue index, principal component analysis index, modified red-blue vegetation index, visible light atmospheric resistance vegetation index.
6. The method of claim 3, wherein the biomass-related vegetation index information in the MS sensor comprises a chlorophyll vegetation index, a chlorophyll index, a generalized differential vegetation index, a dual-band enhanced vegetation index, an enhanced vegetation index, a green anti-atmospheric vegetation index, a green leaf area index, a blue normalized differential vegetation index, a green specific vegetation index, a modified chlorophyll absorption reflection index 2, a soil-adjusted vegetation index, an optimized soil-adjusted vegetation index, a modified chlorophyll absorption reflection index 1/optimized soil-adjusted vegetation index, a SPAD-related vegetation index information comprises a chlorophyll vegetation index, a canopy chlorophyll content index, a green normalized differential vegetation index, a normalized differential infrared index, a projected normalized differential vegetation index, a green soil-adjusted vegetation index, a green optimized vegetation index, a normalized infrared red index, a modified normalized infrared differential infrared index, a modified enhanced vegetation index, a modified chlorophyll absorption reflection index, a land chlorophyll, a generalized differential vegetation index.
7. The method for constructing a salt tolerance comprehensive scoring model according to claim 1, wherein the machine learning algorithm is used for estimating pea biomass by using Catboost algorithm and pea SPAD by using Light GBM algorithm.
8. The method of claim 1, wherein R 2 and RMSE are used to evaluate the biomass of peas and the robustness of SPAD.
9. The method for constructing a salt tolerance composite score model according to claim 1, wherein the salt tolerance composite score model is psts= Σ (i=1, n) [ U (X ij)×wi ], wherein PSTS represents a salt tolerance factor, U (X ij) represents a membership function value, X ij represents an i-th principal component value of a j-th cell, and w i represents a weight of the i-th principal component.
10. A method for screening pea varieties with salt tolerance, characterized in that the pea varieties with high comprehensive salt tolerance evaluation are evaluated by using the unmanned aerial vehicle-based pea high-throughput salt tolerance comprehensive scoring model according to any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510239077.2A CN120236192A (en) | 2025-03-03 | 2025-03-03 | A method for constructing a high-throughput comprehensive scoring model for salt tolerance of pea based on drones |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510239077.2A CN120236192A (en) | 2025-03-03 | 2025-03-03 | A method for constructing a high-throughput comprehensive scoring model for salt tolerance of pea based on drones |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN120236192A true CN120236192A (en) | 2025-07-01 |
Family
ID=96160974
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202510239077.2A Pending CN120236192A (en) | 2025-03-03 | 2025-03-03 | A method for constructing a high-throughput comprehensive scoring model for salt tolerance of pea based on drones |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN120236192A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN120992519A (en) * | 2025-10-23 | 2025-11-21 | 四川农业大学 | A method for evaluating the cold resistance of Napier grass based on UAV multispectral imagery |
-
2025
- 2025-03-03 CN CN202510239077.2A patent/CN120236192A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN120992519A (en) * | 2025-10-23 | 2025-11-21 | 四川农业大学 | A method for evaluating the cold resistance of Napier grass based on UAV multispectral imagery |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Combining texture, color, and vegetation indices from fixed-wing UAS imagery to estimate wheat growth parameters using multivariate regression methods | |
| Duan et al. | Remote estimation of grain yield based on UAV data in different rice cultivars under contrasting climatic zone | |
| Liu et al. | Estimating leaf area index using unmanned aerial vehicle data: shallow vs. deep machine learning algorithms | |
| CN109459392B (en) | A method for estimating aboveground biomass of rice based on UAV multispectral imagery | |
| CN114694020B (en) | Construction method of cotton aphid remote sensing prediction model | |
| CN106372592B (en) | A calculation method of winter wheat planting area based on winter wheat area index | |
| Dobrowski et al. | Grapevine dormant pruning weight prediction using remotely sensed data | |
| US20240404045A1 (en) | A Monitoring And Evaluation Method For Comprehensive Evaluation Index Of Machine-Harvested Cotton Defoliation Effect And System Thereof | |
| CN100394212C (en) | A method for remote sensing detection and estimation of planting area and yield of large-scale crops | |
| Sharifi | Estimation of biophysical parameters in wheat crops in Golestan province using ultra-high resolution images | |
| Chen et al. | Mapping rice planting areas in southern China using the China Environment Satellite data | |
| Jeong et al. | Application of an unmanned aerial system for monitoring paddy productivity using the GRAMI-rice model | |
| Yang et al. | Cotton hail disaster classification based on drone multispectral images at the flowering and boll stage | |
| Ma et al. | Monitoring defoliation rate and boll-opening rate of machine-harvested cotton based on UAV RGB images | |
| Liu et al. | Application of UAV-retrieved canopy spectra for remote evaluation of rice full heading date | |
| CN109001125A (en) | A kind of growth of cereal crop seedlings detection method and system based on high score No.1 satellite image | |
| Liang et al. | CanopyCAM–an edge-computing sensing unit for continuous measurement of canopy cover percentage of dry edible beans | |
| CN120236192A (en) | A method for constructing a high-throughput comprehensive scoring model for salt tolerance of pea based on drones | |
| CN118243634A (en) | Method, device, equipment and storage medium for predicting nitrogen content of field tobacco leaves | |
| Waine et al. | Towards improving the accuracy of opium yield estimates with remote sensing | |
| Zhang et al. | Retrieving the chlorophyll content of individual apple trees by reducing canopy shadow impact via a 3D radiative transfer model and UAV multispectral imagery | |
| Bao et al. | UAV remote sensing technology: A review of Chinese agricultural patents | |
| CN120339836A (en) | A method for planning crop pest and disease image collection points based on UAV | |
| CN112345467A (en) | A Model for Estimating Rice Physiological Parameters Using Remote Sensing Technology and Its Application | |
| Andrade et al. | Monitoring of Corn Growth Stages by UAV Platform Sensors |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |