Chilo suppressalis stres identification method based on unmanned aerial vehicle hyperspectral image

Abstract

The present application relates to a kind of rice Chilo suppressalis pest stress identification method based on unmanned aerial vehicle hyperspectral image, adopt unmanned aerial vehicle hyperspectral image, using continuous projection algorithm filters out the band sensitive to rice and pest rice, constructs rice index RI and rice Chilo suppressalis pest index RBPI, by analyzing the threshold of the box plot of RI index to determine rice and non-rice, analyze the threshold of the box plot of RBPI index to determine healthy rice and the rice stressed by Chilo suppressalis, and then provide effective basis for the distinction of sample type;Using decision tree algorithm effectively identifies the distribution of rice stressed by Chilo suppressalis, improves the efficiency and accuracy of Chilo suppressalis pest monitoring, realizes the rapid monitoring of rice Chilo suppressalis pest, provides important scientific basis for agricultural management and decision-making, and helps to promote the green and sustainable development of agriculture.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of remote sensing image classification and extraction, and particularly relates to a Chilo suppressalis threat identification method based on unmanned aerial vehicle hyperspectral images. BACKGROUND

[0002] Rice is the basis of food security in China and plays an extremely important role in China's grain production. Its stable yield and income increase are of great significance to the protection of national food security. However, rice is often threatened by Chilo suppressalis pest. In the seedling stage of rice, the invasion of Chilo suppressalis larvae into rice will cause dry heart; in the tillering stage, Chilo suppressalis larvae continue to invade the stems of rice, causing dry sheath; in the jointing stage to the booting stage, the ear part is dry and white, causing dry ear. Pest damage causes the growth and development of rice to be hindered, and can even lead to yield reduction or even no yield.

[0003] Traditional rice pest diagnosis methods observe the changes of physiological and morphological indicators such as rice leaf color, leaf wilting state, curling degree, and root stem health state in the field by producers and plant protection experts, and then determine the severity of rice pest stress. This labor-intensive field investigation method has problems such as low efficiency, high cost, and strong subjectivity.

[0004] Remote sensing technology can efficiently identify the spatial and temporal variation information of crop pests. Currently, the remote sensing data commonly used for pest identification mainly includes multispectral satellite data, hyperspectral satellite data (Landsat, Sentinel-2, Modis) and unmanned aerial vehicle hyperspectral data. The spatial resolution of multispectral and hyperspectral satellite data is low, and the observation period is long, which cannot capture small-area pests in the field in time. Unmanned aerial vehicles have high spatial resolution and flexible observation frequency, and can quickly cover large areas of farmland and accurately capture the occurrence of small-scale pests. In addition, unmanned aerial vehicle hyperspectral images can provide rich spectral information, each pixel has hundreds of band data, which can be used to identify and distinguish different types of vegetation pests, and help to discover pests earlier, providing scientific basis for the implementation of prevention and control measures.

[0005] Acquiring the planting range of rice is the basis for rice pest monitoring. Hyperspectral remote sensing is mostly used to monitor the planting area of rice by using spectral indices. Common spectral indices include NDVI, EVI, DVI, etc. NDVI is mainly used to evaluate the density and greenness of vegetation, but it is prone to saturation in high vegetation coverage areas (such as mature rice fields), and NDVI no longer increases significantly, making it difficult to effectively identify rice. EVI is not sensitive to mixed water areas, but there is a lot of water in the field during the key growth stages of rice (transplanting, tillering, and jointing), and EVI is easily affected by water in this case, leading to inaccurate extraction of rice. DVI has high sensitivity to soil background and atmospheric conditions, but its application range is limited, and it is mainly used to distinguish between vegetation and non-vegetation, making it difficult to further subdivide different vegetation. In addition, the above-mentioned vegetation index has a wide waveband range, and the spectral overlap makes rice and other vegetation types have similar spectral reflection characteristics in these wavebands, making it difficult to directly apply to rice identification in unmanned aerial vehicle hyperspectral images. Therefore, considering factors such as water mixing in rice fields, vegetation coverage, and similar spectral characteristics of different vegetation, it is necessary to develop a rice remote sensing index identification method for unmanned aerial vehicle hyperspectral images.

[0006] The method for monitoring rice pest by unmanned aerial vehicle hyperspectral remote sensing also mostly uses vegetation index methods, such as NDVI, GNDVI, SAVI, DVI, etc., which can all reflect the physiological state and health status of vegetation and are effective for identifying vegetation pests. NDVI is a commonly used index for measuring vegetation greenness and can reflect the uneven distribution of greenness on pest leaves. GNDVI is sensitive to the reduction and degradation of chlorophyll content. SAVI is sensitive to the wilting, curling, or shedding of damaged plant leaves. DVI has good monitoring ability for yellowing, browning, and spotting of pest leaves.

[0007] However, the rice stem is mainly damaged by the diamondback moth, which mostly manifests as withered heart, withered sheath, and white head, and the above-mentioned vegetation indices cannot effectively reflect the characteristics of rice stressed by the diamondback moth. Therefore, it is necessary to construct a spectral index for rice diamondback moth pest stress to achieve effective monitoring of the diamondback moth pest. SUMMARY

[0008] The purpose of the present application is to overcome the shortcomings of the prior art and provide a method for identifying rice diamondback moth pest stress based on unmanned aerial vehicle hyperspectral images.

[0009] The method for identifying rice diamondback moth pest stress based on unmanned aerial vehicle hyperspectral images comprises the following steps:

[0010] Step one, conduct field investigation of rice in the experimental area to obtain ground cover categories and spatial location attributes; and take unmanned aerial vehicle hyperspectral images;

[0011] Step two, determine the sample points of rice and non-rice in the unmanned aerial vehicle hyperspectral image according to the field rice survey data, and analyze the spectral curve of the obtained sample points;

[0012] Step three, use the successive projection algorithm to filter and identify the sensitive waveband of rice and the sensitive waveband of rice stressed by Chilo spp. from the data obtained by field rice survey, use the reflectance ratio of the sensitive waveband to construct rice identification index RI and rice Chilo spp. pest stress identification index RBPI, and set threshold values for the two indexes respectively for the judgment of ground cover categories;

[0013] Step four, use the decision tree algorithm to extract the rice planting area from the unmanned aerial vehicle hyperspectral image, and use the decision tree algorithm to identify the rice area stressed by Chilo spp. within the rice planting area; use the confusion matrix to evaluate the accuracy.

[0014] As preferred, in step one, the categories of ground cover include rice and non-rice, and the rice includes healthy rice and rice stressed by Chilo spp.; in step two, the non-rice area and the rice area samples are distinguished according to the trend of reflectance changing with wavelength in the spectral curve, and the healthy rice and rice stressed by Chilo spp. samples are distinguished according to the value of reflectance in the spectral curve.

[0015] As preferred, in step three, the successive projection algorithm is used to realize the dimension reduction of the full waveband of 400-1000 nm, the projection of one wavelength on the remaining unselected wavelengths is calculated by loop, the wavelength containing the least amount of redundant information is found, the collinearity of the input data set is reduced through the wavelength, and the two most sensitive spectral wavelengths for rice are selected, the reflectance of the two spectral wavelengths from small to large is R S and R L , and the calculation formula of the rice identification index RI is:

[0016]

[0017] As preferred, in step three, the threshold value T1 of the rice identification index RI and the threshold value T2 of R L are determined; when the RI of the sample is greater than T1 and the R L is greater than T2, the ground cover category corresponding to the sample is rice; the determination method of T1 is as follows: calculate the pixel value of the pixel corresponding to RI in the unmanned aerial vehicle hyperspectral image of the rice and non-rice type sample points, and make a box plot, and determine the threshold value T1 from the average of the lower quartile of the RI of the rice sample points and the upper quartile of the RI of the non-rice sample points;

[0018] The determination method of T2 is as follows: calculate the R L of the rice and non-rice type sample points in the unmanned aerial vehicle hyperspectral image, and make a box plot, and determine the threshold value T2 from the average of the R LThe R of the lower quartile and the non-rice sample points L The mean of the upper quartile determines the threshold T2.

[0019] As preferred, in step three, the continuous projection algorithm is used to realize the dimension reduction of the full wave band of 400-1000 nm, the projection of one wavelength on the remaining unselected wavelengths is calculated through a loop, the wavelength containing the least amount of redundant information is found, the collinearity of the input data set is reduced through the wavelength, and the three spectral wavelengths most sensitive to the rice stressed by the rice borer are screened out, the reflectivity corresponding to the three spectral wavelengths from small to large is R1, R2 and R3 respectively, and the calculation formula of the rice borer damage identification index RBPI is:

[0020]

[0021] As preferred, in step three, the threshold T3 of the rice borer damage identification index RBPI is determined, when the ground cover class of the sample is rice, RBPI < T3, then the rice at the sample is healthy rice; the determination method of T3 is: the pixel value of the pixel corresponding to RBPI of the sample points of healthy rice and rice stressed by the rice borer in the UAV hyperspectral image is calculated, the threshold T3 is obtained from the lower quartile of the RBPI of the healthy rice sample points and the mean of the upper quartile of the RBPI of the rice sample points stressed by the rice borer.

[0022] The beneficial effects of the present application are:

[0023] 1) The present application uses UAV hyperspectral image, uses the continuous projection algorithm to screen out the wave band sensitive to rice and rice damaged by pests, constructs the rice index RI and the rice borer damage index RBPI, determines the threshold of rice and non-rice by analyzing the box plot of RI index, determines the threshold of healthy rice and rice stressed by the rice borer by analyzing the box plot of RBPI index, and then provides an effective basis for distinguishing the sample type.

[0024] 2) The present application effectively identifies the distribution of rice stressed by the rice borer, improves the efficiency and accuracy of the rice borer damage monitoring, realizes the rapid monitoring of the rice borer damage, and provides an important scientific basis for agricultural management and decision-making, which helps to promote the green and sustainable development of agriculture. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The figure is a sample point diagram of rice and non-rice, healthy rice and rice stressed by the rice borer;

[0026] Figure 2 The figure is a flow chart of the rice borer damage identification method;

[0027] Figure 3 The figure is a ground object spectrum curve diagram of rice and non-rice.

[0028] Figure 4 Fig. 1 is a schematic diagram of the ground spectrum curve of rice infested by Chilo suppressalis;

[0029] Figure 5 Fig. 2 is a schematic diagram of screening of the characteristic waveband of rice;

[0030] Figure 6 Fig. 3 is a schematic diagram of screening of the characteristic waveband of Chilo suppressalis;

[0031] Figure 7 Fig. 4 is a diagram of separating rice from non-rice by the index RI;

[0032] Figure 8 Fig. 5 is a diagram of separating rice from non-rice by the reflectance at 770 nm of the spectral waveband;

[0033] Figure 9 Fig. 6 is a schematic diagram of the identification result of rice;

[0034] Figure 10 Fig. 7 is a diagram of separating healthy rice from Chilo suppressalis infested rice by the index RBPI;

[0035] Figure 11 Fig. 8 is a schematic diagram of the identification result of Chilo suppressalis stressed rice. DETAILED DESCRIPTION

[0036] The present application will be further described below in conjunction with examples. The following examples are only used to help understand the present application. It should be pointed out that for those skilled in the art, without departing from the principles of the present application, a number of modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of the present application.

[0037] Example 1

[0038] As an example, the rice Chilo suppressalis pest stress identification method based on unmanned aerial vehicle hyperspectral image is shown in Fig. 1. Figure 1

[0039] In a first aspect, a rice Chilo suppressalis pest stress index based on unmanned aerial vehicle hyperspectral image is provided, comprising:

[0040] Step 1: Conduct field rice investigation in the experimental area, and the specific investigation data includes the categories and spatial location attributes of healthy rice, Chilo suppressalis pest stressed rice and non-rice.

[0041] Delimit the flight experimental area and obtain the unmanned aerial vehicle hyperspectral image;

[0042] Pretreat the unmanned aerial vehicle hyperspectral image, including radiation brightness calculation, reflectance correction, geometric correction, background value removal and inlaying. ​

[0043] Step two, determine the sample points of rice and non-rice according to the field survey data of rice, wherein the rice includes healthy rice and rice under the stress of chilo suppressalis; perform spectral curve analysis on the obtained sample points to obtain the spectral curve rule of rice and non-rice, healthy rice and rice under the stress of chilo suppressalis; specifically:

[0044] Rice and non-rice have their own unique spectral characteristics. The chlorophyll in rice leaves has strong absorption capacity for blue light and red light, and less absorption for green light. As shown in Figure 3 , therefore, the spectral curve of rice has lower reflectivity at 400nm-510nm and 600nm-690nm, and a small reflection peak at 520nm-570nm. Due to the reflection of leaf cell structure, there is a significant fluctuation at 700nm-781nm. The spectral reflectivity of non-rice area changes relatively smoothly with wavelength, without obvious absorption or reflection peak.

[0045] As shown in Figure 4 , within the spectral range of 400nm-700nm, the reflectivity of healthy rice is lower than that of rice under the stress of chilo suppressalis. Specifically, the healthy rice plant has lower reflectivity in this visible light band due to the strong absorption of blue light and red light by chlorophyll. The rice affected by chilo suppressalis has lower chlorophyll content, resulting in weaker light absorption, so the reflectivity is relatively high. In the near-infrared band of 730nm-1000nm, the reflectivity of healthy rice is significantly higher than that of chilo suppressalis. This is because the healthy rice leaves have good cell structure and high water content, which can reflect the near-infrared light more effectively. The chilo suppressalis rice has damaged leaves, disordered structure and reduced water content, resulting in lower reflectivity in this band.

[0046] Step three, use the continuous projection algorithm to screen and identify the sensitive band of rice, and use the sensitive band to construct the rice identification index RI; draw a box plot to determine the threshold range of RI.

[0047] Use the continuous projection algorithm to screen the sensitive band of rice under the stress of chilo suppressalis, and construct the chilo suppressalis rice pest identification index RBPI; draw a box plot to determine the threshold range of RBPI;

[0048] Step four, use the decision tree algorithm to extract the rice planting area, and use the confusion matrix to evaluate the accuracy; use the decision tree algorithm to identify the rice area under the stress of chilo suppressalis within the rice planting area, and use the confusion matrix to evaluate the accuracy.

[0049] Example two

[0050] As another embodiment, this example two proposes a more specific method for identifying rice under the stress of chilo suppressalis based on unmanned aerial vehicle hyperspectral image on the basis of example one:

[0051] Based on the spectral differences between rice and non-rice, healthy rice, and pest-damaged rice obtained in step two, the types of ground features can be identified and distinguished. The spectral resolution of the UAV hyperspectral imagery used reaches 18.5 cm, which can capture these spectral differences across multiple bands. However, this high resolution introduces band redundancy, increases the complexity of data processing, and leads to low computational efficiency. Therefore, dimensionality reduction processing of the 272 bands of the UAV hyperspectral imagery is required.

[0052] Specifically, in step three:

[0053] The successive projections algorithm (SPA) can achieve dimensionality reduction across the entire 400–1000 nm wavelength range. The SPA algorithm iteratively calculates the projection of a wavelength onto the remaining unselected wavelengths to find the wavelength with the least redundant information, thereby reducing collinearity in the input data sets and selecting the spectral bands most sensitive to rice. For example… Figure 5 As shown, in the parameter settings, the number of characteristic wavelengths is set to 5, and 5 sensitive characteristic bands are selected, with wavelengths of 770nm, 510nm, 719nm, 541nm, and 784nm. According to the importance of the band characteristics, the first two most important bands are selected, which are 770nm and 510nm in this embodiment.

[0054] Based on the sensitive wavelengths and spectral characteristics of rice, an index RI is proposed to accurately identify rice regions. Influenced by the chlorophyll content, cell structure, and water content of rice plants, rice exhibits high reflectance at 770 nm and relatively low reflectance at 510 nm. In contrast, non-rice plants have lower reflectance at 770 nm and higher reflectance at 510 nm. Therefore, by calculating the reflectance ratio at 770 nm to 510 nm, the difference between rice and non-rice plants can be highlighted. Furthermore, to better remove the influence of weeds on rice extraction, reflectance data at 770 nm is incorporated to further distinguish between rice and non-rice plants. Figure 8 As shown, R 770 The inclusion of this index amplifies the difference between rice and non-rice, enabling better identification and monitoring of rice. The formula for calculating the RI index is:

[0055]

[0056] Among them, R 770 R is the reflectance value at 770 nm in the spectral band. 510 This represents the reflectance value at 510 nm in the spectral band. When the sample's RI > T1, and R... 770 If T2 is true, then the land cover category corresponding to this sample is rice; if T1 is true, then T2 is true. The threshold, T2 is R 770 The threshold.

[0057] The threshold T1 is determined by calculating the number of rice and non-rice type sample points in the UAV hyperspectral image. Pixel values ​​for corresponding pixels are calculated, and box plots are created. The threshold T1 is determined by the mean of the lower quartile of rice sample points and the upper quartile of non-rice sample points.

[0058] The threshold T2 is determined as follows: the reflectance of rice and non-rice sample points in the 770nm band is calculated, and a box plot is created. The threshold T2 is determined by the mean of the lower quartile of the rice sample points and the upper quartile of the non-rice sample points.

[0059] Using the same method, SPA feature selection was performed on the sensitive spectral data of rice under rice stem borer stress, such as... Figure 6 As shown, the number of characteristic wavelengths was set to 5, and 5 sensitive characteristic bands were selected, with wavelengths of 781nm, 724nm, 510nm, 464nm, and 688nm. Based on the importance of the band characteristics, the top 3 most important bands were selected, with wavelengths of 781nm, 724nm, and 510nm.

[0060] Based on the spectral bands that rice pests are sensitive to, a rice stem borer indices (RBPI) are proposed. For example... Figure 10 As shown, the RBPI of healthy rice is higher than that of rice affected by pests, effectively distinguishing healthy rice from rice stressed by the rice stem borer. The formula for calculating the RBPI index is:

[0061]

[0062] Among them, R 781 R 724 R 510 T3 represents the reflectance values ​​at 781nm, 724nm, and 510nm in the spectral band. If the RBPI of a sample with a surface cover category of rice is greater than T3, then the rice in that sample is rice under stress from the rice stem borer. T3 is the threshold of RBPI.

[0063] Box plots were drawn to determine the RBPI threshold range. The pixel values ​​corresponding to the RBPI values ​​of healthy and rice stem borer infestation type sample points in the UAV hyperspectral imagery were calculated, and box plots were created to determine the threshold T3. The threshold T3 was determined by the mean of the lower quartile of the healthy sample points and the upper quartile of the rice stem borer infestation sample points.

[0064] It should be noted that the parts in this embodiment that are the same as or similar to those in Embodiment 1 can be referred to each other, and will not be repeated in this application.

[0065] Example 3

[0066] As another embodiment, embodiment three is proposed on the basis of embodiments one and two, the rice borer stress recognition method based on unmanned aerial vehicle hyperspectral image uses a rice borer extraction device, which comprises:

[0067] An acquisition module is configured to acquire unmanned aerial vehicle hyperspectral image data and pre-process the unmanned aerial vehicle hyperspectral data.

[0068] A screening module is configured to screen characteristic sensitive bands from 272 hyperspectral bands.

[0069] A construction module is configured to construct a rice index RI and a rice borer index RBPI based on unmanned aerial vehicle hyperspectral technology.

[0070] An index calculation module is configured to calculate index RI and RBPI images corresponding to the unmanned aerial vehicle hyperspectral image.

[0071] An extraction module is configured to determine a threshold range to extract a rice range according to the RI result of the sample, and then determine a threshold range to extract a rice borer range according to the RBPI result.

[0072] A computer storage medium is provided, and the computer storage medium stores a computer program; when the computer program runs on a computer, the computer program causes the computer to execute the rice borer stress recognition method based on the unmanned aerial vehicle hyperspectral image of any one of the first aspect.

[0073] The rice planting area is extracted by the decision tree algorithm, and the precision is evaluated by the confusion matrix. The quantitative evaluation indexes include producer accuracy (PA), user accuracy (UA), overall classification accuracy (OA) and Kappa. The calculation formulas of the evaluation indexes are as follows:

[0074]

[0075] In the formula, TP is true positive, and FN is false negative.

[0076]

[0077] In the formula, TP is true positive, and FP is false positive.

[0078]

[0079] In the formula, ∑TP represents the sum of true positives of all categories, and ∑N is the total number of samples.

[0080]

[0081] In the formula, OA is the overall accuracy, P eis the accuracy obtained by random classification, N represents the total number of verification samples, n represents the total number of columns of the confusion matrix, xii represents the number of samples in the ith row and the ith column of the confusion matrix, X i+ and X +i respectively represent the sum of each row and each column element of the confusion matrix.

[0082] In this embodiment, the decision tree algorithm is used, the RI index threshold is used for division, and the rice planting range is extracted. The producer accuracy (PA) of rice is 0.96, and the user accuracy (UA) is 0.97; the PA and UA of non-rice are both 0.98, the overall classification accuracy (OA) is 0.97, and the Kappa is 0.92. After evaluation, the recognition result shows high accuracy and good performance in identifying rice and non-rice.

[0083] Using the same index, in the rice planting area, through the RBPI threshold division, the rice range stressed by the diamondback moth is extracted, and the accuracy is evaluated by using the confusion matrix, wherein the PA of the rice stressed by the diamondback moth is 0.96, and the UA is 0.95; the PA of the healthy rice is 0.94, and the UA is 0.95, the overall classification accuracy (OA) is 0.95, and the Kappa is 0.9. The overall accuracy of the recognition is high, which shows that the index method used can effectively distinguish the rice stressed by the diamondback moth and the healthy rice, and has high accuracy.

[0084] It should be noted that the same or similar parts in this embodiment and embodiments one to two can be mutually referred to, and will not be described in detail in this application.

[0085] Embodiment four

[0086] As another embodiment, this embodiment four proposes a more specific rice diamondback moth pest stress recognition method based on unmanned aerial vehicle hyperspectral image on the basis of embodiments one to three:

[0087] In step one, the polygon tool is used to demarcate the collection area and the flight area.

[0088] The unmanned aerial vehicle collection area parameters are set, including the color of the reflectance calibration cloth, the flight height and the speed. The white cloth is selected as the reflectance calibration cloth, which ensures that the area of the photographed calibration cloth cannot be overexposed, that is, the pixel brightness value (DN value) curve peak value <4000, and at the same time ensures that the frame staring time > exposure time. The flight height is set to 200m, and the flight speed is calculated through the calculator of the software.

[0089] The flight area plans the flight route, the flight height is matched with the height setting of the collection area, which is 200m, the lateral overlap rate is 30%, the heading overlap rate is 10%, the side margin is 0, and the main flight line angle is adjusted to be parallel to the long side of the collection area.

[0090] In step one, the process of pre-processing the unmanned aerial vehicle hyperspectral image is specifically as follows:

[0091] Radiance calculation is to convert the DN data collected by the hyperspectral imager into radiance data through a standard radiance calibration configuration file. The specific conversion formula is as follows:

[0092]

[0093] wherein DN is the original light intensity value recorded by the hyperspectral imager, L1 is the radiance value, G is the camera gain, texp is the integration time of the photosensitive element, and DF is the dark current noise.

[0094] Reflectivity correction first generates a white reference file by referring to the pixel reflectivity (0.56) of the white cloth, and then calculates the reflectivity of all strip data in the collection area.

[0095] Geometric correction automatically matches each frame of data of the hyperspectral imager and the corresponding attitude information by using a geometric correction algorithm, and completes geometric correction after successful matching.

[0096] Background value removal and mosaicking. In each geometrically corrected txt file, the values with zero radiance are removed to complete background value removal, and then the batch fast mosaicking tool is used to complete image mosaicking. After the pre-processing is completed, the unmanned aerial vehicle hyperspectral data is output, the number of bands is 272, and the spatial resolution is 18.5 cm.

[0097] It should be noted that the parts same as or similar to the embodiments one to three in the present embodiment can be mutually referenced, and will not be described herein again.

[0098] The embodiments in the present specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between the embodiments can be mutually referenced.

Claims

1. A method for identifying rice stem borer pest stress based on UAV hyperspectral imagery, characterized in that, Includes the following steps: Step 1: Conduct field rice surveys in the experimental area to obtain land cover type and spatial location attributes; capture hyperspectral images using drones. Step 2: Based on the field rice survey data, determine the sample points of rice and non-rice in the UAV hyperspectral imagery, and perform spectral curve analysis on the obtained sample points; Step 3: Using a continuous projection algorithm, the data obtained from the field rice survey are used to screen and identify the sensitive wavelengths of rice and the sensitive wavelengths of rice under the stress of rice stem borer. The reflectance ratio of the sensitive wavelengths is used to construct a rice identification index. Rice stem borer identification index Thresholds were set for the two indices to determine the land cover category; In step three, a continuous projection algorithm is used to achieve dimensionality reduction across the entire 400–1000 nm wavelength range. By iteratively calculating the projection of a wavelength onto the remaining unselected wavelengths, the wavelength containing the least redundant information is found. This wavelength is used to reduce the collinearity of the input data sets, thereby selecting the two spectral wavelengths most sensitive to rice and the three spectral wavelengths most sensitive to rice under rice stem borer stress. The reflectances of the two most sensitive spectral wavelengths to rice, from smallest to largest, are as follows: and Rice identification index The calculation formula is: The reflectances of the three most sensitive spectral wavelengths for rice under rice stem borer stress, from smallest to largest, are as follows: , and Rice stem borer identification index The calculation formula is: Step 4: Use the decision tree algorithm to extract rice planting areas from the UAV hyperspectral imagery, and use the decision tree algorithm to identify rice areas under the stress of rice stem borer within the rice planting area. Accuracy was evaluated using a confusion matrix; in step one, the land cover categories included rice and non-rice, and rice included healthy rice and rice under the stress of rice stem borer. In step two, samples from non-rice and rice regions are distinguished based on the trend of reflectance variation with wavelength in the spectral curve, and healthy rice samples and rice samples under the stress of rice stem borer are distinguished based on the reflectance value in the spectral curve.

2. The method for identifying rice stem borer pest stress based on UAV hyperspectral imagery according to claim 1, characterized in that, In step three, the rice identification index is determined. threshold and threshold When the sample ,and If so, the land cover category corresponding to this sample is rice; The determination method is as follows: calculate the number of rice and non-rice type sample points in the UAV hyperspectral image. The pixel values ​​corresponding to the pixels were used to create a box plot, based on the rice sample points. Lower quartiles and non-rice sample points The mean of the upper quartile determines the threshold. ; The determination method is as follows: calculate the number of rice and non-rice type sample points in the UAV hyperspectral image. And create a box plot based on rice sample points. Lower quartiles and non-rice sample points The mean of the upper quartile determines the threshold. .

3. The method for identifying rice stem borer pest stress based on UAV hyperspectral imagery according to claim 2, characterized in that, In step three, the threshold for the rice stem borer pest identification index is determined. The samples with surface cover category of rice If so, the rice at that sample location is healthy rice; The determination method is as follows: The sample points of healthy rice and rice under rice stem borer stress are calculated in the UAV hyperspectral imagery. The pixel value of the corresponding pixel is determined by the healthy rice sample points. The threshold is obtained by taking the mean of the lower quartile and the upper quartile of the rice sample points under rice stem borer stress. .

Images