Inversion Method for the Severity of Rice Bacterial Blight Based on UAV Hyperspectral

Abstract

The present invention relates to a method for inverting the severity of rice bacterial blight based on UAV hyperspectral, comprising the following steps: analyzing the spectral curves of the reflectance of different types of sample points varying with wavelength, determining the sensitive bands for different degrees of bacterial blight damage, and selecting the sensitive bands to construct the bacterial blight damaged rice inversion index NBLBRI; determining the NBLBRI thresholds for inverting four different degrees of bacterial blight damaged rice. The beneficial effects of the present invention are as follows: by analyzing the spectral curve differences of the training samples, determining the sensitive bands for different degrees of bacterial blight damage, selecting the sensitive bands to construct the bacterial blight damaged rice inversion index NBLBRI; analyzing the differences in the statistical characteristics of NBLBRI between the rice with different disease damages and the healthy rice, determining the threshold intervals of NBLBRI for estimating the degrees of different types of rice disease damages, so as to accurately extract the regions with different disease degrees; providing a flexible and broad application prospect for agricultural disease management.

Description

Technical Field

[0001] The present invention belongs to the field of remote sensing estimation of the damage degree of rice diseases, and particularly relates to a method for inverting the severity of rice bacterial blight based on unmanned aerial vehicle (UAV) hyperspectral data. Background Art

[0002] As one of the main food crops in the world, rice plays a crucial role in ensuring food security.

[0003] During the growth process of rice, it is often attacked by various diseases, especially bacterial blight, which is a serious disease caused by the bacterium Xanthomonas oryzae pv. oryzae. Bacterial blight not only destroys the leaf tissue of rice but also significantly reduces its photosynthetic capacity, thereby affecting yield and quality. Due to the complexity and prevalence of diseases, accurately and rapidly monitoring their occurrence and development is crucial for ensuring the healthy growth of crops and the stable supply of food. Traditional methods for monitoring rice diseases mainly rely on manual surveys and ground spectrometer observations, but these methods have limitations such as long time consumption, low efficiency, and insufficient spatial resolution, making it difficult to be widely applied in large-scale planting areas. Although satellite remote sensing has a wide coverage range, it still cannot meet the requirements of refined management and dynamic monitoring in terms of spatio-temporal resolution. In contrast, UAV hyperspectral technology, with its high spatial resolution, flexibility, and short revisit cycle, can quickly obtain high-precision spectral information on the ground, enabling accurate identification and quantification of the severity of diseases, providing strong support for the early warning, precise prevention and control, and scientific management of rice diseases.

[0004] However, for rice bacterial blight, there is still a lack of a detection method that can accurately extract the disease area and reflect the spatial distribution range of rice bacterial blight. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method for inverting the severity of rice bacterial blight based on UAV hyperspectral data.

[0006] This method for inverting the severity of rice bacterial blight based on UAV hyperspectral data quantitatively estimates the damage degree of rice bacterial blight and includes the following steps:

[0007] S1. Obtain UAV hyperspectral images;

[0008] S2. Combine field investigations and visual interpretation of hyperspectral images to select rice sample points with different degrees of damage by bacterial blight, including four types: healthy rice, rice with mild bacterial blight damage, rice with moderate bacterial blight damage, and rice with severe bacterial blight damage.

[0009] S3. Extract the hyperspectral band reflectance of the sample points in the training set, analyze the spectral curves of the reflectance of different types of sample points changing with wavelength, determine the sensitive bands for different degrees of bacterial blight damage, and select the sensitive bands to construct the bacterial blight damage rice inversion index NBLBRI;

[0010] S4. Obtain the bacterial blight damage rice inversion index NBLBRI of the four types of samples through band calculation, and determine the NBLBRI thresholds for the inversion of rice with four different degrees of bacterial blight damage;

[0011] S5. Use the sample points in the validation sample set to verify the constructed index NBLBRI and the corresponding threshold, and invert the severity of bacterial blight in the target rice.

[0012] Preferably, for quantitatively estimating the damage degree of bacterial blight in rice, in step S3, the calculation formula of the bacterial blight damage rice inversion index NBLBRI is:

[0013]

[0014] where NBLBRI is the inversion index of the damage degree of bacterial blight in rice, and ρ 904 represents the band reflectance with a central wavelength of 904 nm in the UAV hyperspectral image, and ρ 775 represents the band reflectance with a central wavelength of 775 nm, and ρ 675 represents the band reflectance with a central wavelength of 675 nm.

[0015] Preferably, in step S2, a certain proportion is randomly selected from the ground object sample points in the study area as the training set, and the remaining sample points are used as the validation set; steps S3 to S4 are implemented for the training samples; in step S5, through verification with the validation samples, use the sample points in the validation sample set to verify the constructed index NBLBRI and the estimation results of the damage degree of bacterial blight in rice corresponding to the threshold, and use the index with the highest verification accuracy as the final inversion index of the damage degree of bacterial blight, and invert the severity of bacterial blight in the target area.

[0016] Preferably, in step S4, the NBLBRI ranges corresponding to healthy rice, rice with mild bacterial blight damage, rice with moderate bacterial blight damage, and rice with severe bacterial blight damage do not overlap with each other.

[0017] Preferably, in step S4, by statistically analyzing the value range, mean, standard deviation, and median of the NBLBRI values corresponding to the four types of samples, determine the NBLBRI thresholds for the inversion of rice with four different degrees of bacterial blight damage.

[0018] The beneficial effects of the present invention are:

[0019] 1) The present invention utilizes unmanned aerial vehicle (UAV) hyperspectral technology, combines on-site investigations, selects sample points covering healthy rice and rice with different disease severities; by analyzing the spectral curve differences of the training samples, determines the sensitive bands for different degrees of bacterial blight damage, and selects the sensitive bands to construct the inversion index NBLBRI for rice damaged by bacterial blight; then, through indicators such as range, median, mean, and standard deviation, analyzes the differences in the statistical characteristics of NBLBRI between rice with different disease severities and healthy rice, determines the threshold interval of NBLBRI for estimating the severity of different types of rice diseases, so as to accurately extract areas with different disease severities; this method and device can be used in different environments and conditions, providing a flexible and wide application prospect for agricultural disease management.

[0020] 2) The present invention uses 70% of the collected data for training and 30% for verification, and uses the verification data to check the training results to ensure the accuracy of the inversion. Description of the Drawings

[0021] Figure 1 It is a flow chart of the inversion method for the severity of bacterial blight in rice based on UAV hyperspectral;

[0022] Figure 2 It is a schematic diagram of the mean value of the spectral curves of ground object sample points;

[0023] Figure 3 It is a box plot of different ground object sample points;

[0024] Figure 4a It is a schematic diagram of the inversion result of the severity of bacterial blight in rice;

[0025] Figure 4b For Figure 4a An enlarged view of a partial area in. Detailed Embodiment

[0026] The following further describes the present invention in combination with embodiments. The description of the following embodiments is only used to help understand the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several modifications can still be made to the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

[0027] Embodiment 1

[0028] As an embodiment, as Figure 1 shown, this inversion method for the severity of bacterial blight in rice based on UAV hyperspectral includes the following steps:

[0029] S1. Obtain the UAV hyperspectral image and preprocess the UAV hyperspectral image; the preprocessing includes: radiometric correction, geometric correction, and image mosaicking.

[0030] S2. Combine field surveys and visual interpretation of hyperspectral images to select rice sample points with different degrees of bacterial blight damage, including four types: healthy rice, rice with mild bacterial blight damage, rice with moderate bacterial blight damage, and rice with severe bacterial blight damage. Randomly select 70% of the sample points as training samples and 30% as validation samples;

[0031] S3. Calculate the mean of the spectral curves of the training samples and analyze the differences between the spectral curves of rice with bacterial blight damage of different severities and healthy rice. Determine the sensitive bands based on the differences in the spectral curves, select three key characteristic wavelengths, and construct the inversion index NBLBRI for rice with bacterial blight damage.

[0032] S4. By statistically analyzing the statistical indicators such as the value range, mean, standard deviation, and median of the NBLBRI values corresponding to the four types of samples, determine the optimal NBLBRI threshold for the inversion of rice with bacterial blight damage of four different severities, so as to maximize the discrimination between healthy rice and rice with bacterial blight damage of different severities. Rice satisfies T a <NBLBRI ≤ T b , and the NBLBRI thresholds for different severities of bacterial blight lesions are as follows:

[0033]

[0034] Among them, NBLBRI is the inversion index for rice with bacterial blight damage, and T1 to T5 respectively represent the NBLBRI thresholds for different degrees of bacterial blight damage. The NBLBRI ranges corresponding to healthy rice, rice with mild bacterial blight damage, rice with moderate bacterial blight damage, and rice with severe bacterial blight damage do not overlap with each other.

[0035] S5. Use 30% of the validation data to check the extraction results of S4, and use the extraction results that pass the check as the inversion of the severity of bacterial blight damage of rice with bacterial blight damage of different severities. Calculate NBLBRI for the entire UAV hyperspectral image, and realize the inversion of the severity of different rice bacterial blight according to the index and its threshold obtained in S4, and obtain the spatial distribution map of rice with different bacterial blight damages,

[0036] Example Two

[0037] As another example, this Example Two is proposed on the basis of Example One. Using this inversion method for the severity of rice bacterial blight based on UAV hyperspectral, the severity of bacterial blight in a certain rice farm was estimated.

[0038] S1. Use a DJI Matrice 300 RTK drone equipped with a Nano-Hyperspec camera for aerial photography to obtain drone hyperspectral images. The image band range is 400 - 1000 nm, the spectral resolution is 2.2 nm, and the spatial resolution is 0.0275 m.

[0039] Perform preprocessing operations such as radiometric correction, geometric correction, and image mosaicking on the drone hyperspectral images. The data preprocessing is implemented in the SpectralView software of Headwall Company in the United States.

[0040] Image mosaicking: Use ENVI 5.3 software to perform hyperspectral image mosaicking on the 18 drone strips obtained to obtain the preprocessed drone hyperspectral images.

[0041] Visually interpret and determine the rice pixels, and use the Extract by Mask tool in ArcMap 10.7 to extract the rice area.

[0042] S2. Select a total of four types of sample data of healthy rice, rice damaged by mild bacterial blight, rice damaged by moderate bacterial blight, and rice damaged by severe bacterial blight on the drone hyperspectral images. Randomly select 80% of the training samples, calculate the average spectral reflectance of each band for the four categories, and draw a line graph of spectral reflectance, as Figure 2 shown. The specific process is as follows:

[0043] S201. Combine field survey data and visual interpretation to select regions of interest of different land cover types on the drone images. Evaluate the disease severity of the study area, obtain ground disease point data, and collect 26 sample points by GPS toolbox positioning. As the disease worsens, the color of diseased rice changes from yellowish-white to white on the true color image, and healthy rice appears green;

[0044] S202. Based on the field survey GPS sampling points and visual interpretation, a total of 1154 pixel sample points are selected on the drone images, including 242 of healthy rice, 313 of rice damaged by mild bacterial blight, 332 of rice damaged by moderate bacterial blight, and 239 of rice damaged by severe bacterial blight, and calculate the mean spectral curve of different land cover types of the training samples;

[0045] S3. In this embodiment, by analyzing the spectral curve graph, it can be found that near 904 nm and 775 nm, the reflectance of rice increases with the increase in the severity of bacterial blight damage. At 675 nm, there is an obvious difference in reflectance between healthy rice and diseased rice, and the reflectance of diseased rice is generally higher than that of healthy rice. This is because healthy rice has a higher reflectance in the near-infrared bands of 904 nm and 775 nm. Due to its intact leaf structure, it can strongly reflect near-infrared light. While for diseased rice, due to the damage of cell structure, the near-infrared reflectance gradually increases. In the red light band of 675 nm, healthy rice has a lower reflectance due to chlorophyll absorbing light, while diseased rice has an increased reflectance due to the reduction of chlorophyll. Therefore, 675 nm, 775 nm, and 904 nm are selected as sensitive bands to distinguish the disease severity, and the above three wavelength positions can represent the characteristics of the severity of bacterial blight damage to rice.

[0046] Based on the spectral curve graph and the method of experimental verification, and based on the spectral characteristics of vegetation, the inversion index NBLBRI for bacterial blight-damaged rice is proposed:

[0047]

[0048] In the above formula, NBLBRI is the inversion index for bacterial blight-damaged rice, and ρ 904 represents the reflectance of the band with a central wavelength of 904 nm, and ρ 775 represents the reflectance of the band with a central wavelength of 775 nm, and ρ 675 represents the reflectance of the band with a central wavelength of 675 nm.

[0049] S4. Based on the ENVI 5.3 software and by statistically analyzing the distribution ranges of the index values of four types of samples, namely healthy rice, rice damaged by mild bacterial blight, rice damaged by moderate bacterial blight, and rice damaged by severe bacterial blight, the threshold ranges for distinguishing the four types of samples are determined as shown below. The specific process is as follows: Figure 3 as follows:

[0050] S401. First, calculate the NBLBRI index through the Band Math tool. Superimpose the regions of interest of the four types of rice samples in S2 on the index result graph, calculate the value range of the selected sample pixels, and construct a sample box plot to determine the threshold range;

[0051] S402. Based on the index result and the sample box plot, extract the bacterial blight of rice, and obtain the threshold as:

[0052] T a <NBLBRI≤T b

[0053]

[0054] Among them, NBLBRI is the inversion index of rice damaged by bacterial blight, and T a and T b are the maximum and minimum values of the box plot of NBLBRI respectively. In this embodiment, T1, T2, T3, T4 and T5 are 0.94, 0.66, 0.60, 0.54 and 0.42 respectively.

[0055] S5. Use 20% of the verification data to check the extraction results of S4, and use the extraction results that pass the check as the inversion results of the rice area damaged by bacterial blight with different severities, as Figure 4a and Figure 4b shown.

[0056] To sum up, the present invention uses unmanned aerial vehicle hyperspectral technology to effectively realize the inversion of the severity of rice bacterial blight. Through hyperspectral image processing and spectral difference analysis, the disease area is accurately extracted, reflecting the spatial distribution of the damage degree of rice bacterial blight, which has more practical application significance in the early disease prevention and control.

[0057] It should be noted that the parts that are the same or similar to those in Embodiment 1 in this embodiment can be referred to each other and will not be elaborated in this application.

[0058] Embodiment 3

[0059] As another embodiment, this Embodiment 3 is proposed on the basis of Embodiments 1 and 2, and is used for an inversion system for the inversion method of the severity of rice bacterial blight based on unmanned aerial vehicle hyperspectral, including:

[0060] Acquisition module: Acquire unmanned aerial vehicle hyperspectral images and preprocess the images;

[0061] Construction module: Analyze the spectral curve differences between healthy rice and rice damaged by bacterial blight with different severities, and construct the inversion index NBLBRI of rice damaged by bacterial blight based on the spectral differences;

[0062] Extraction module: Use the NBLBRI index, combine with training samples to construct a sample index box plot, determine the threshold range of different disease severities, and accurately extract the spatial distribution of rice damaged by bacterial blight with different severities;

[0063] Verification module: Use verification samples to check the extraction results to ensure the classification accuracy. The extraction results that pass the check are used as the final inversion results of different degrees of bacterial blight damage.

[0064] At the same time, a computer storage medium and a computer program product are provided. The computer storage medium stores a computer program; when the computer program runs on a computer, the computer is enabled to execute any one of the above-mentioned rice bacterial blight disease severity inversion methods.

[0065] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other.

Claims

1. An inversion method for the severity of rice bacterial blight based on UAV hyperspectral, characterized in that Quantitatively estimating the damage degree of bacterial blight in rice, including the following steps: S1. Obtain unmanned aerial vehicle hyperspectral images; S2. Combine field surveys and hyperspectral image visual interpretation to select rice sample points with different damage degrees of bacterial blight, including four types: healthy rice, rice with mild bacterial blight damage, rice with moderate bacterial blight damage, and rice with severe bacterial blight damage; S3. Extract the hyperspectral band reflectance of the sample points in the training set, analyze the spectral curves of the reflectance of different types of sample points changing with wavelength, determine the sensitive bands with different damage degrees of bacterial blight, and select the sensitive bands to construct the inversion index NBLBRI of rice damaged by bacterial blight; When quantitatively estimating the damage degree of bacterial blight in rice, in step S3, the calculation formula of the inversion index NBLBRI of rice damaged by bacterial blight is: Among them, NBLBRI is the inversion index of the damage degree of rice bacterial blight, and ρ 904 represents the reflectance of the band with a central wavelength of 904 nm in the UAV hyperspectral image, and ρ 775 represents the reflectance of the band with a central wavelength of 775 nm, and ρ 675 represents the reflectance of the band with a central wavelength of 675 nm; S4. Obtain the inversion index NVLBRI of rice damaged by bacterial blight for the four types of samples through band calculation, and determine the NVLBRI thresholds for the inversion of the four different damage degrees of bacterial blight in rice; In step S4, the NBLBRI ranges corresponding to healthy rice, rice with mild bacterial blight damage, rice with moderate bacterial blight damage, and rice with severe bacterial blight damage do not overlap with each other; By statistically analyzing the value ranges, means, standard deviations, and medians of the NBLBRI values corresponding to the four types of samples, determine the NBLBRI thresholds for the inversion of the four different damage degrees of bacterial blight in rice; S5. Use the sample points in the validation sample set to verify the constructed index NBLBRI and the corresponding thresholds, and invert the severity of bacterial blight in the target rice.

2. The method for inverting the severity of rice bacterial blight based on UAV hyperspectral according to claim 1, wherein In step S2, randomly select a certain proportion of the ground object sample points in the study area as the training set, and the remaining sample points as the validation set; Implement steps S3 to S4 for the training samples; In step S5, conduct a check through the validation samples, use the sample points in the validation sample set to verify the estimation results of the damage degree of bacterial blight in rice of the constructed index NBLBRI and the corresponding thresholds, and use the index with the highest validation accuracy as the final inversion index of the damage degree of bacterial blight, and invert the severity of bacterial blight in the target area.

Images