Rice bacterial leaf blight severity inversion method based on unmanned aerial vehicle hyperspectrum

Abstract

The invention relates to an unmanned aerial vehicle hyperspectrum-based rice bacterial blight severity inversion method, which comprises the following steps of: analyzing a spectrum curve of different types of sample point reflectivity along with wavelength change, determining sensitive wavebands of different bacterial blight harm degrees, and selecting the sensitive wavebands to construct a bacterial blight harm rice inversion index NBLBRI; and determining NBLBRI thresholds of inversion of the rice harmed by the four different bacterial blight diseases. The method has the beneficial effects that the sensitive wavebands with different bacterial blight damage degrees are determined by analyzing the spectral curve difference of the training samples, and the sensitive wavebands are selected to construct the bacterial blight damage rice inversion index NBLBRI; analyzing differences between statistical characteristics of NBLBRI between rice with different disease damage degrees and healthy rice, and determining threshold intervals of NBLBRI for estimating the damage degrees of different types of rice diseases so as to accurately extract areas with different disease degrees; and a flexible and wide application prospect is provided for agricultural disease management.

Description

Technical Field

[0001] The invention belongs to the field of remote sensing estimation of rice disease damage degree, and in particular relates to an inversion method for the severity of rice bacterial blight based on unmanned aerial vehicle hyperspectral. Background Art

[0002] As one of the world's major food crops, rice plays a vital role in ensuring food security. As one of my country's important crops, rice production plays an important role in ensuring national food supply. However, due to the invasion of bacterial blight pathogens, rice plant physiological functions are damaged, which in turn affects yield and quality. Therefore, monitoring of rice diseases has become the key to disease prevention and control.

[0003] Rice is often attacked by various diseases during its growth, especially bacterial blight, a serious disease caused by the bacterium Xanthomonas oryzae pv.oryzae. Bacterial blight not only destroys the leaf tissue of rice, but also causes a significant decrease in its photosynthesis capacity, which in turn affects yield and quality. Due to the complexity and widespread nature of the disease, accurate and rapid monitoring of its occurrence and development is essential to ensure the healthy growth of crops and a stable supply of food. Traditional rice disease monitoring methods mainly rely on manual surveys and ground-based spectrometer observations, but these methods have limitations such as long time consumption, low efficiency, and insufficient spatial resolution, making them difficult to be widely used in large-scale planting areas. Although satellite remote sensing has a wide coverage range, its temporal and spatial resolution still cannot meet the needs of refined management and dynamic monitoring. In contrast, drone hyperspectral technology, with its high spatial resolution, flexibility, and short revisit cycle, can quickly obtain high-precision spectral information on the ground, accurately identify and quantify the severity of the disease, and provide strong support for early warning, precise prevention and control, and scientific management of rice diseases.

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

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art and provide an inversion method for the severity of rice bacterial blight based on unmanned aerial vehicle hyperspectral images.

[0006] This method of inversion of the severity of rice bacterial blight based on UAV hyperspectral data quantitatively estimates the damage degree of rice bacterial blight, including the following steps:

[0007] S1, obtain UAV hyperspectral images;

[0008] 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 of rice: healthy rice, rice with mild bacterial blight damage, rice with moderate bacterial blight damage, and rice with severe bacterial blight damage;

[0009] S3, extracting the reflectance of the high-spectral bands of the sample points in the training set, and analyzing the spectral curves of the reflectance of different types of sample points changing with wavelength, determining the sensitive bands of different degrees of bacterial leaf blight damage, and selecting sensitive bands to construct the NBLBRI of rice inversion index of bacterial leaf blight damage;

[0010] S4. Obtain the NBLBRI of rice damaged by bacterial leaf blight of four types of samples through band calculation, and determine the NBLBRI thresholds of four different rice damaged by bacterial leaf blight inversions;

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

[0012] Preferably, the damage degree of rice bacterial blight is quantitatively estimated. In step S3, the calculation formula of the rice bacterial blight damage inversion index NBLBRI is:

[0013]

[0014] Among them, NBLBRI is the rice bacterial blight damage inversion index, ρ 904 represents the reflectance of the band with a central wavelength of 904 nm in the UAV hyperspectral image, ρ 775 It represents the reflectivity of the band with a central wavelength of 775nm, ρ 675 Indicates the reflectivity of the band with a central wavelength of 675nm.

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

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

[0017] Preferably, in step S4, the NBLBRI thresholds for inversion of four different rice species damaged by bacterial blight are determined by counting the range, mean, standard deviation and median of the NBLBRI corresponding to the four types of samples.

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

[0019] 1) The present invention utilizes the UAV hyperspectral technology, combined with field research, to select sample points, covering healthy rice and rice with different disease degrees; by analyzing the spectral curve differences of the training samples, the sensitive bands of different degrees of bacterial leaf blight damage are determined, and the sensitive bands are selected to construct the rice inversion index NBLBRI of bacterial leaf blight damage; then, by using the range, median, mean, standard deviation and other indicators, the differences between the statistical characteristics of the NBLBRI between rice with different degrees of disease damage and healthy rice are analyzed, and the threshold interval of the NBLBRI for estimating the degree of damage of different types of rice diseases is determined, so as to accurately extract areas with different disease degrees; the method and the device can be used in different environments and conditions, providing a flexible and broad 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. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The flowchart of the inversion method of the severity of rice bacterial blight based on UAV hyperspectral data;

[0022] Figure 2 This is a schematic diagram of the mean value of the spectral curve of the ground feature sample points;

[0023] Figure 3 Box plots of statistics for different ground feature sample points;

[0024] Figure 4a This is a schematic diagram of the inversion results of the severity of rice bacterial blight;

[0025] Figure 4b for Figure 4a A magnified view of the middle area. DETAILED DESCRIPTION

[0026] The present invention is further described below in conjunction with embodiments. The description of the following embodiments is only used to help understand the present invention. It should be noted that for ordinary persons in the art, without departing from the principle of the present invention, the present invention can also be modified in some ways, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

[0027] Embodiment 1

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

[0029] S1. Acquire UAV hyperspectral images and preprocess the UAV hyperspectral images; the preprocessing includes: radiation correction, geometric correction and image stitching.

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

[0031] S3. Calculate the mean of the spectral curve of the training samples and analyze the differences in the spectral curves of rice damaged by white leaf blight of different severity and healthy rice. Determine the sensitive bands based on the differences in the spectral curves, select three key characteristic wavelengths, and construct the NBLBRI of rice damaged by white leaf blight.

[0032] S4. By calculating the range, mean, standard deviation, median and other statistical indicators of NBLBRI corresponding to the four types of samples, the optimal NBLBRI threshold for inversion of four different types of rice damaged by bacterial blight is determined, so that the distinction between healthy rice and rice with different degrees of bacterial blight damage is maximized. Rice satisfies T a <NBLBRI≤T b The NBLBRI thresholds for different severity levels of bacterial blight are as follows:

[0033]

[0034] Among them, NBLBRI is the rice inversion index of bacterial leaf blight damage, and T1 to T5 represent the NBLBRI thresholds of different degrees of bacterial leaf blight damage. The NBLBRI ranges corresponding to healthy rice, rice with mild bacterial leaf blight damage, rice with moderate bacterial leaf blight damage, and rice with severe bacterial leaf 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 verified extraction results as the inversion of the severity of rice bacterial leaf blight with different severity. Calculate the NBLBRI for the entire UAV hyperspectral image, and invert the severity of rice bacterial leaf blight according to the index and threshold obtained in S4 to obtain the spatial distribution map of rice with different severity of bacterial leaf blight.

[0036] Embodiment 2

[0037] As another embodiment, based on the embodiment one, the second embodiment uses the inversion method of the severity of rice bacterial blight disease based on the drone hyperspectral to estimate the severity of rice bacterial blight disease in a rice farm.

[0038] S1. Use the DJI M300 RTK drone equipped with a Nano-Hyperspec camera to take aerial photos and obtain drone hyperspectral images. The image band range is 400-1000nm, the spectral resolution is 2.2nm, and the spatial resolution is 0.0275m.

[0039] The UAV hyperspectral images were preprocessed by radiation correction, geometric correction and image stitching. Data preprocessing was implemented in the SpectralView software of Headwall, USA.

[0040] Image stitching: ENVI 5.3 software was used to stitch the 18 acquired UAV strips into hyperspectral images to obtain the preprocessed UAV hyperspectral images.

[0041] The rice pixels were determined by visual interpretation, and the rice areas were extracted using the Extract by Mask tool in ArcMap 10.7.

[0042] S2. Select four types of sample data, including healthy rice, rice with mild bacterial blight damage, rice with moderate bacterial blight damage, and rice with severe bacterial blight damage, from the drone hyperspectral image. Randomly select 80% of the training samples, calculate the average spectral reflectance of each band of the four categories, and draw a spectral reflectance line graph, as shown in Figure 2. Figure 2 As shown, the specific process is:

[0043] S201. Combine field survey data and visual interpretation to select areas of interest of different types of land objects on drone images. Evaluate the degree of disease in the study area, obtain ground disease point data, and collect 26 sample points through GPS toolbox. As the disease worsens, the color of diseased rice on true color images changes from yellow-white to white, and healthy rice appears green;

[0044] S202. Based on the field survey of GPS sampling points and visual interpretation, a total of 1154 pixel sample points were selected on the UAV images, including 242 healthy rice, 313 rice with mild bacterial blight damage, 332 rice with moderate bacterial blight damage, and 239 rice with severe bacterial blight damage, and the mean of the spectral curves of different ground object types of the training samples were calculated;

[0045] S3. In the present embodiment, by analyzing the spectrum curve graph, it can be found that the reflectivity of rice near 904nm and 775nm increases with the increase in the severity of the damage caused by white leaf blight. At 675nm, there is a significant difference in the reflectivity between healthy rice and diseased rice, and the reflectivity of diseased rice is generally higher than that of healthy rice; this is because healthy rice has a higher reflectivity in the near-infrared band of 904nm and 775nm, because its leaf structure is intact and can strongly reflect near-infrared light; and the diseased rice has a damaged cell structure, which causes the near-infrared reflectivity to gradually increase. In the red light band of 675nm, healthy rice has a lower reflectivity due to the absorption of light by chlorophyll, while the reflectivity of diseased rice increases due to the reduction of chlorophyll. Therefore, 675nm, 775nm and 904nm are selected as sensitive bands for distinguishing the degree of disease, and the above three wavelength positions can represent the characteristics of the severity of the damage caused by rice white leaf blight.

[0046] Based on the spectral curve and experimental verification, the NBLBRI of rice damage inversion index was proposed based on the spectral characteristics of vegetation:

[0047]

[0048] In the above formula, NBLBRI is the rice bacterial blight damage inversion index, ρ 904 represents the reflectivity of the band with a central wavelength of 904nm, ρ 775 It represents the reflectivity of the band with a central wavelength of 775nm, ρ 675 Indicates the reflectivity of the band with a central wavelength of 675nm.

[0049] S4. Based on ENVI 5.3 software, the distribution range of index values ​​of four types of samples, namely healthy rice, rice with mild bacterial blight damage, rice with moderate bacterial blight damage and rice with severe bacterial blight damage, was calculated to determine the threshold range for distinguishing the four types of samples. Figure 3 As shown, the specific process is:

[0050] S401. First, the NBLBRI index is calculated using the Band Math tool, the regions of interest of the four rice samples in S2 are superimposed on the index result map, the value range of the selected sample pixels is calculated, and a sample box plot is constructed to determine the threshold range;

[0051] S402, extracting rice bacterial blight based on the index result and the sample box plot, and obtaining the threshold value:

[0052] T a <NBLBRI≤T b

[0053]

[0054] Among them, NBLBRI is the rice bacterial blight damage inversion index, T a , 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 validation data to check the extraction results of S4, and use the extracted results that pass the check as the inversion results of rice areas damaged by white leaf blight of different severity, such as Figure 4a and Figure 4b shown.

[0056] In summary, the present invention uses drone hyperspectral technology to effectively invert the severity of rice bacterial blight. Through hyperspectral image processing and spectral difference analysis, the diseased area is accurately extracted to reflect the spatial distribution of the degree of damage of rice bacterial blight, which has more practical application significance in early disease prevention and control.

[0057] It should be noted that the parts in this embodiment that are the same or similar to those in the first embodiment can be referenced to each other and will not be described in detail in this application.

[0058] Embodiment 3

[0059] As another embodiment, this embodiment 3 proposes, based on the embodiments 1 and 2, an inversion system for performing the inversion method of the severity of rice bacterial blight based on the UAV hyperspectral, including:

[0060] Acquisition module: acquires UAV hyperspectral images and preprocesses them;

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

[0062] Extraction module: Using the NBLBRI index and combining it with training samples to construct a sample index box plot, the threshold range of different disease severity is determined, and the spatial distribution of rice damaged by white leaf blight with different severity is accurately extracted;

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

[0064] A computer storage medium and a computer program product are also provided. The computer storage medium stores a computer program. When the computer program is run on a computer, the computer executes any of the above-mentioned rice bacterial blight disease severity inversion methods.

[0065] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

Claims

1. A method for inverting the severity of rice bacterial blight based on UAV hyperspectral images, characterized in that Quantitative estimation of the damage degree of rice bacterial blight includes the following steps: S1, obtain UAV hyperspectral images; 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 of rice: healthy rice, rice with mild bacterial blight damage, rice with moderate bacterial blight damage, and rice with severe bacterial blight damage; S3, extracting the reflectance of the high-spectral bands of the sample points in the training set, and analyzing the spectral curves of the reflectance of different types of sample points changing with wavelength, determining the sensitive bands of different degrees of bacterial leaf blight damage, and selecting sensitive bands to construct the NBLBRI of rice inversion index of bacterial leaf blight damage; S4. Obtain the NBLBRI of rice damaged by bacterial leaf blight of four types of samples through band calculation, and determine the NBLBRI thresholds of four different rice damaged by bacterial leaf blight inversions; S5. Use the sample points in the validation sample set to validate the constructed index NBLBRI and the corresponding threshold value, and invert the severity of the target rice bacterial leaf blight.

2. The method for inverting the severity of rice bacterial blight based on UAV hyperspectral according to claim 1 is characterized in that: To quantitatively estimate the degree of damage caused by rice bacterial leaf blight, in step S3, the calculation formula of the rice bacterial leaf blight damage inversion index NBLBRI is: Among them, NBLBRI is the rice bacterial blight damage inversion index, ρ 904 represents the reflectance of the band with a central wavelength of 904 nm in the UAV hyperspectral image, ρ 775 It represents the reflectivity of the band with a central wavelength of 775nm, ρ 675 Indicates the reflectivity of the band with a central wavelength of 675nm.

3. The method for inverting the severity of rice bacterial blight based on UAV hyperspectral according to claim 1 is characterized in that: In step S2, a certain proportion of the sample points in the study area are randomly selected as the training set, and the remaining sample points are used as the verification set; steps S3 to S4 are implemented on the training samples; in step S5, the verification samples are used for verification, and the constructed index NBLBRI and the rice bacterial blight damage degree estimation results of the corresponding threshold are verified using the sample points in the verification sample set, and the index with the highest verification accuracy is used as the final rice bacterial blight damage degree inversion index, and the rice bacterial blight severity is inverted in the target area.

4. The method for inverting the severity of rice bacterial blight based on UAV hyperspectral according to claim 1, characterized in that: In step S4, the NBLBRI ranges corresponding to the healthy rice, the rice with mild bacterial blight damage, the rice with moderate bacterial blight damage, and the rice with severe bacterial blight damage do not overlap with each other.

5. The method for inverting the severity of rice bacterial blight based on UAV hyperspectral according to claim 1, characterized in that: In step S4, the NBLBRI thresholds for inversion of four different rice species damaged by bacterial blight are determined by counting the range, mean, standard deviation and median of the NBLBRI corresponding to the four types of samples.

Images