Disease detection method and disease detection device

Abstract

To provide a disease detection method having high robustness.SOLUTION: A disease detection device conducts: a step (step S11) of obtaining a target image which shows vegetation; a step (step S13) of computing vegetation index values for each pixel in any target region in the target image; a step (step S14) of classifying each pixel in the target region R into a disease class and a healthy class based on the vegetation index values and calculating class-to-class variance between the disease class and the healthy class; and a step (step S15) of determining whether the target region is a disease-affected area by comparing the class-to-class variance with a predetermined determination threshold.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a vegetation disease detection method and a vegetation disease detection device. [Background technology]

[0002] Conventionally, a method using image analysis is known as a method for detecting vegetation diseases. For example, in the method disclosed in Patent Document 1, a vegetation index value of a crop planted in a field is calculated based on an image of the field. Then, a representative vegetation index value of the field is compared with a vegetation index value at a predetermined position in the field, and if the vegetation index value at the predetermined position deviates from the representative vegetation index value, it is estimated that a disease has occurred at the predetermined position. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2021-106554 A Summary of the Invention [Problem to be solved by the invention]

[0004] In general, vegetation index values ​​are prone to variation among individual vegetation depending on the degree of growth. For this reason, when determining the occurrence of disease by setting a certain threshold value for the vegetation index value, it is difficult to ensure the accuracy of the determination. Even when determining the occurrence of disease using the deviation from a representative vegetation index value, as in the method disclosed in the above Patent Document 1, it is difficult to improve the accuracy of the determination because it is affected by the variation in the vegetation index value. Therefore, a highly robust disease detection method is required. [Means for solving the problem]

[0005] A disease detection method according to one aspect of the present disclosure is a method that performs the steps of acquiring a target image in which vegetation is captured, calculating a vegetation index value for each unit area in a target area in the target image, classifying each unit area in the target area into a disease class and a health class based on the corresponding vegetation index value and calculating the inter-class variance between the disease class and the health class, and determining whether or not disease has occurred in the target area by comparing the inter-class variance with a predetermined judgment threshold.

[0006] A disease detection device according to one embodiment of the present disclosure includes an image acquisition unit that acquires a target image in which vegetation is captured, a vegetation index value calculation unit that calculates a vegetation index value for each unit area in any target area within the target image, an inter-class variance calculation unit that classifies each unit area in the target area into a disease class and a health class based on the vegetation index value and calculates the inter-class variance between the disease class and the health class, and a judgment unit that judges whether or not disease has occurred in the target image by comparing the inter-class variance with a predetermined judgment threshold. [Brief description of the drawings]

[0007] [Figure 1] FIG. 1 is a block diagram showing a schematic configuration of a disease detection system according to one embodiment of the present invention. [Diagram 2] 3 is a flowchart showing a disease detection method according to the present embodiment. [Diagram 3] FIG. 13 is a diagram illustrating an example of a target image (healthy target image) in which healthy vegetation is captured. [Figure 4] FIG. 2 is a diagram illustrating an example of a target image (disease target image) in which diseased vegetation is captured. [Diagram 5] 4 is a histogram illustrating the vegetation index value for each pixel in a health target image; [Figure 6] 1 is a histogram illustrating the vegetation index value of each pixel in a disease target image. [Figure 7] 4 is a flowchart showing a method for determining a determination threshold value for a vegetation index value according to the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] Hereinafter, one embodiment of the present invention will be described. Fig. 1 is a schematic diagram showing a general configuration of a disease detection system 1 of this embodiment. In Fig. 1, the disease detection system 1 includes an imaging unit 10 that captures images of vegetation, and a disease detection device 20 that detects disease based on an image of the vegetation captured by the imaging unit 10 (i.e., a target image I). Note that the disease detection device 20 of this embodiment is configured as a separate entity capable of communicating with the imaging unit 10 by any means, but may also be configured integrally with the imaging unit 10. The vegetation may be any vegetation, such as agricultural crops, etc. The type of disease is not particularly limited as long as it is a disease that can occur in vegetation.

[0009] [Imaging unit] The imaging unit 10 includes an imaging section 11, a GPS receiving section 12, and an imaging control section 13. The imaging unit 11 of this embodiment may be an RGB camera or a spectroscopic camera. That is, the target image I captured by the imaging unit 11 may be an RGB image or a spectroscopic image. When the imaging unit 11 is a spectroscopic camera, the imaging unit 11 is configured to include an imaging optical system, a spectroscopic element, and an imaging element, and captures a spectroscopic image corresponding to one or more wavelengths as the target image I. For example, a wavelength-variable Fabry-Perot etalon can be used as the spectroscopic element.

[0010] The GPS receiving section 12 receives signals from GPS satellites to identify the current position of the imaging unit 10 and generate position information.

[0011] The imaging control section 13 is configured with a microcomputer that controls the overall operation of the imaging unit 10, a memory that stores various data, etc. The microcomputer controls the imaging section 11 to generate image data based on an image signal output from the imaging section 11. The microcomputer also stores the generated image data in the memory in association with position information at the time of imaging.

[0012] [Disease detection device] The disease detection device 20 of this embodiment calculates a vegetation index value for each pixel (corresponding to a unit area of ​​the present invention) from a target image I, and determines the presence or absence of disease on the basis of the calculated vegetation index value. Here, the vegetation index value may be a pixel value of an arbitrary wavelength, a pixel value of an arbitrary color, or a normalized difference vegetation index (NDVI). For example, the NDVI can be calculated by the following formula. NDVI=(IR-R) / (IR+R) R: Reflectance of red light in the visible range IR: Near infrared reflectance There is no particular limitation on the method of calculating the reflectance of each wavelength from the target image I. For example, the reflectance may be calculated by dividing the pixel value of each wavelength by a reference value acquired in advance.

[0013] The disease detection device 20 has the basic components of a computer. For example, the disease detection device 20 has a storage unit 21 configured with a storage circuit such as a memory, and a processor 22 configured with an arithmetic circuit such as a CPU (Central Processing Unit). In addition, although not shown, the disease detection device 20 has an operation unit that accepts user operations, and a display unit such as a display.

[0014] The storage unit 21 is an information storage device configured with a memory, a hard disk, etc. The storage unit 21 stores various programs including a disease detection program for detecting disease from the target image I, and various data used when executing the various programs.

[0015] For example, the storage unit 21 of this embodiment stores a threshold table indicating the correspondence between the type of disease, the type of vegetation index value (wavelength, color, or NDVI), and the judgment threshold value Td for the vegetation index value. The storage unit 21 also stores countermeasure information corresponding to the type of disease (type of pesticide, application time and method, and advice information such as leaf removal). The data in the threshold table can be determined based on the symptoms of disease in vegetation, and may be obtained from any database or may be determined by prior experiments or simulations.

[0016] The processor 22 reads and executes the programs stored in the memory unit 21, thereby functioning as an image acquisition unit 221, an area setting unit 222, a vegetation index value calculation unit 223, an inter-class variance calculation unit 224, a determination unit 225, an output unit 226, a label unit 227, and a threshold setting unit 228. Each of these functions will be described later.

[0017] (Disease detection method) A disease detection method using the disease detection device 20 of this embodiment will be described. Before disease detection by disease detection device 20 begins, imaging unit 10 captures images of vegetation in a field such as a farm field in response to user operations or setting information. For example, when imaging unit 10 is mounted on a small flying object such as a drone or a small self-propelled aircraft, imaging section 11 may capture images each time imaging unit 10 moves a predetermined distance in the field, and imaging control section 13 may generate target image I corresponding to position information at the time of capturing the image.

[0018] The disease detection device 20 starts the flowchart of FIG. 2 in response to a user operation or the like. In the following, for simplicity of explanation, a case will be described in which one RGB image is acquired as the target image I. Also, it is assumed that the type of disease to be detected is set in advance. Note that the type of disease may be set based on information input by the user, or may be set by another method.

[0019] First, the image acquisition section 221 acquires a target image I from the imaging unit 10, and stores it in the storage section 21 (step S11). Note that the image acquisition section 221 may acquire position information corresponding to the target image I together with the target image I.

[0020] Next, the region setting unit 222 sets a target region R for the target image I acquired in step S11 (step S12). For example, the region setting unit 222 may detect the edge of any part of the vegetation (e.g., leaves or fruits) in the target image I by using a known image processing technique, and set the range surrounded by the edge as the target region R. Furthermore, when the edge cannot be detected, for example, when a part of the vegetation is reflected in the entire target image I, the region setting unit 222 may set the entire region of the target image I as the target region R.

[0021] 3 and 4 are diagrams showing schematic examples of a target image I in which a target region R has been set in step S12. Here, FIG. 3 shows a target image I in which healthy vegetation has been captured, and FIG. 4 shows a target image I in which diseased vegetation has been captured. Within the target region R of the target image I shown in FIG. 3, almost the entire surface is green. Meanwhile, within the target region R of the target image I shown in FIG. 4, disease spots (hatched regions in the figure) caused by disease are present locally. The region in which the disease spots exist has physical property values ​​(such as color) that are different from other regions in the target region R.

[0022] Next, the vegetation index value calculation unit 223 selects a type of vegetation index value corresponding to the set disease type, and calculates the vegetation index value for each pixel in the target region R (step S13). The following illustrates an example in which a red image value is calculated as the vegetation index value corresponding to the disease type to be detected (for example, "downy mildew").

[0023] Next, the inter-class variance calculation unit 224 classifies each pixel in the target region R into a disease class and a health class based on the vegetation index value of the pixel, and calculates the variance of the vegetation index value between the disease class and the health class in the target region R, i.e., the inter-class variance σ 2 bis calculated (step S14).

[0024] Step S14 will now be described in detail. The inter-class variance calculation unit 224 classifies each pixel in the target region R into a disease class and a health class based on the vegetation index value of the pixel, and calculates the inter-class variance σ 2 b The average pixel value of the entire target region R is calculated as m t , variance is σ 2 t Let the number of pixels in the health class be ω 1 , the average is m 1 , variance is σ 2 1 Let the number of pixels in the disease class be ω 2 , the average is m 2、 Variance is σ 2 2 Let us assume that.

number

[0025] The threshold value t for the above classification may be determined so as to maximize the degree of separation X between the healthy class and the disease class. The degree of separation X is calculated using the following formula (2). In addition, the within-class variance σ 2 w is calculated by the following formula (3).

number

number

[0026] Here, the total variance of pixel values ​​in the target region R is σ 2 t Since is expressed by the following formula (4), the degree of separation X can be transformed into the following formula (5).

number

number

[0027] In the above formula (5), the total variance σ 2 t is constant. Therefore, to maximize the degree of separation X, the inter-class variance σ 2 b The threshold value t can be determined so that From the above, the inter-class variance σ 2 b is calculated.

[0028] Fig. 5 illustrates a histogram of vegetation index values ​​calculated based on a target image I in which healthy vegetation as shown in Fig. 3 is captured. In this histogram, one peak appears corresponding to the normal color of vegetation, and the inter-class variance σ 2 b tends to be smaller.

[0029] 6 illustrates a histogram of vegetation index values ​​calculated based on a target image I in which vegetation in a diseased state as shown in FIG. 4 is captured. In such a histogram, a peak corresponding to the health class and a peak corresponding to the disease class appear, and therefore the inter-class variance σ 2 b Therefore, the inter-class variance of the vegetation index value shown in Fig. 6, σ 2 b is the inter-class variance of the vegetation index values ​​shown in Figure 5. 2 b Greater than.

[0030] Next, the determination unit 225 uses the inter-class variance σ 2 b With a predetermined determination threshold Td, it is determined whether or not a disease has occurred in the target region R (step S15). Note that the determination threshold Td is a value stored in the threshold table corresponding to the type of disease to be detected.

[0031] In step S15, the inter-class variance σ 2 b If it is equal to or smaller than the determination threshold value Td (step S15; YES), the determination unit 225 determines that the target region R is a healthy area. After that, the output unit 226 performs a process for when the target region R is a healthy area, for example, a process of outputting to the display that the target region R is healthy (step S16).

[0032] On the other hand, in step S15, the inter-class variance σ 2 b is greater than the judgment threshold value Td (step S15; NO), the judgment unit 225 judges that the target area R is a disease-infested area. Thereafter, the output unit 226 performs processing for the case where the target area R is a disease-infested area (step S17). For example, the output unit 226 may output countermeasure information corresponding to the type of disease to a display or the like, or may output position information corresponding to the target image I of the target area R judged to be a disease-infested area to a display or the like.

[0033] This completes the process of the flowchart in FIG. In the above description, the process of the disease detection device 20 for one target image I has been described for the sake of simplicity, but the process may be performed simultaneously for multiple target images I. For example, the disease detection device 20 may acquire multiple target images I in step S11, perform steps S12 to S16 in order for each target image I, and output the determination results (presence or absence of disease and the type of disease) corresponding to each target image I in list form. At this time, the output unit 226 may output position information corresponding to the target image I determined to have a disease, together with the determination results.

[0034] (Determination of the judgment threshold Td) A method for determining the determination threshold value Td in this embodiment will be described with reference to the flowchart of FIG.

[0035] First, the image acquisition unit 221 acquires previously prepared teacher data (step S21). Here, the teacher data includes a plurality of images of healthy vegetation and a plurality of images of diseased vegetation.

[0036] Next, the region setting unit 222 sets a target region R for each image acquired in step S21 (step S22). Note that the method of setting the target region R is substantially the same as that of step S12 above.

[0037] Next, the vegetation index value calculation unit 223 calculates the vegetation index value for each pixel in the target region R of each image (step S23). At this time, the vegetation index value calculation unit 223 calculates one or more types of vegetation index values ​​for each pixel.

[0038] Next, the inter-class variance calculation unit 224 classifies each pixel in the target region R of each image into a disease class and a healthy class, and calculates the variance of the vegetation index value between the disease class and the healthy class, that is, the inter-class variance σ 2 b (Step S24). Note that the inter-class variance σ 2 b The calculation method is substantially the same as that in step S14. In addition, when a plurality of types of vegetation index values ​​are calculated, the inter-class variance σ 2 b It is sufficient to calculate

[0039] Next, the label unit 227 calculates the inter-class variance σ 2 b The images are labeled (step S25). Here, the label indicates the presence or absence of disease and the type of disease. The specific labeling method is not particularly limited, and known techniques can be used. As a result, the training data includes a label corresponding to each image.

[0040] Next, the threshold setting unit 228 calculates the inter-class variance σ 2 bA classifier such as a support vector machine is trained on the above, and for each disease type, the classifier calculates the inter-class variance σ 2 b A classification threshold for classifying the vegetation into a healthy group and a diseased group is calculated as a judgment threshold Td (step S26). Note that, based on the correspondence between the disease type and the vegetation threshold type in the threshold table, the judgment threshold Td for the corresponding vegetation threshold type may be calculated for each disease type.

[0041] Thereafter, the threshold setting unit 228 stores the determination threshold value Td obtained for each type of disease in step S26 in the threshold value table of the storage unit 21 (step S27). This completes the process of the flowchart in FIG.

[0042] [Effects of this embodiment] As described above, the disease detection method of this embodiment includes the steps of acquiring a target image I in which vegetation is captured (step S11), calculating a vegetation index value for each pixel in an arbitrary target region R in the target image I (step S13), classifying each pixel in the target region R into a disease class and a health class based on the vegetation index value, and estimating an inter-class variance σ 2 b (Step S14) and calculating the inter-class variance σ 2 b and a step of judging whether or not the target region R is a disease-occurring area by comparing the difference with a predetermined judgment threshold value Td (step S15). In this method, the inter-class variance σ 2 b The value of varies depending on the state of the vegetation (whether healthy or diseased), but is not easily affected by the variation in vegetation index value for each individual plant due to the growth rate. Therefore, even for vegetation with large variation in growth rate for each individual plant, it is possible to accurately determine the presence or absence of disease. In other words, according to this embodiment, a highly robust disease detection method is provided.

[0043] Furthermore, in the method of Patent Document 1 described above, when calculating a representative vegetation index value, data from healthy areas and disease-infested areas are averaged, which causes a problem in that the accuracy of determining whether or not a disease has occurred is reduced, particularly in the early stages of a disease. In contrast, in the disease detection method of this embodiment, data averaging is not performed, and the accuracy of determining whether or not a disease has occurred can be improved even in the early stages of a disease. In other words, the disease detection method of this embodiment allows for early detection of disease. This allows measures against disease to be taken in the early stages of disease occurrence.

[0044] Furthermore, by using the disease detection device 20 that performs the disease detection method of the present embodiment, disease detection can be automated, which can reduce the workload, particularly in large-scale farm fields. This can contribute to the promotion of smart agriculture.

[0045] In this embodiment, the target image I is a spectral image or an RGB image corresponding to one or more wavelengths, and the vegetation index value may be a pixel value of any wavelength, a pixel value of any color, or a normalized vegetation index. Furthermore, in the disease detection method of the present embodiment, before the step of calculating the vegetation index value (step S13), the type of vegetation index value is selected according to the type of disease set as the detection target. According to such a method, it is possible to accurately determine whether or not a disease has occurred depending on the type of disease.

[0046] The disease detection method of this embodiment further includes a step of acquiring training data including images of healthy vegetation and images of diseased vegetation (step S21), and a step of training a classifier on the inter-class variance of vegetation index values ​​calculated from each image of the training data, and determining a classification threshold Td as a judgment threshold Td for the classifier to classify each image of the training data into a healthy group and a diseased group (step S26). According to such a method, an appropriate determination threshold value Td can be set, and as a result, the occurrence or non-occurrence of disease can be determined with high accuracy.

[0047] In this embodiment, the training data trained by the classifier further includes labels indicating the presence or absence of disease and the type of disease corresponding to each image, and the judgment threshold Td is determined for each type of disease based on the labels. According to this method, an appropriate determination threshold value Td can be set depending on the type of disease.

[0048] The disease detection method of this embodiment may further include a step of providing a user with countermeasure information corresponding to the type of disease when the target area R is determined to be a disease-occurring area based on a judgment threshold Td corresponding to any type of disease. This method can prompt the user to take appropriate measures after a disease outbreak.

[0049] In addition, in this embodiment, the target image I is associated with imaging position information, and the disease detection method of this embodiment may further include a step of providing the location of the disease occurrence to the user based on the imaging position information when the target area R is determined to be a disease occurrence area. This method is suitable for processing a large number of target images I acquired in a large field at once, and can contribute to large-scale agriculture. In addition, by analyzing the location of disease occurrence, users can determine the range of pesticide spraying for the field.

[0050] [Variations] The present invention is not limited to the above-described embodiment, and modifications and improvements within the scope of the present invention that can achieve the object of the present invention are included in the present invention.

[0051] In the disease detection method of the above embodiment, one target region R is set in one target image I, but multiple target regions R may be set. In this case, the presence or absence of disease may be determined for each target region R by calculating the inter-class variance for each target region R.

[0052] In the disease detection method of the above embodiment, a type of vegetation index value corresponding to the type of disease set as the detection target is selected, and the selected type of vegetation index value is calculated, but the present invention is not limited to this. For example, in the disease detection method of the above embodiment, the above step S13 may be omitted, and in the above step S14, the vegetation index value may be calculated for all types present in the threshold table. In this case, in the above step S15, the inter-class variance σ 2 b In step S16, the vegetation index value of the corresponding type may be compared with the corresponding judgment threshold value Td for each type of disease based on the threshold table. This makes it possible to determine the type of disease that has occurred in the vegetation.

[0053] In the above embodiment, the judgment threshold Td is obtained by machine learning using teacher data, but the present invention is not limited to this. For example, the threshold setting unit 228 may obtain the judgment threshold Td from an arbitrary database.

[0054] In the above embodiment, an RGB image is used as the target image I, but a spectral image may be used as the target image I. In this case, the pixel value of any wavelength included in the spectral image may be used as the vegetation index value.

[0055] In the above embodiment, in order to detect any type of disease, one type of vegetation index value corresponding to the disease is used, but multiple types of vegetation index values ​​may be used in combination. For example, pixel values ​​of multiple wavelengths or pixel values ​​of multiple colors may be used as vegetation index values ​​corresponding to the disease.

[0056] In the above embodiment, a pixel is used as the unit region of the target region R, but a plurality of pixels may be used as the unit region. In this case, the average value of the pixel values ​​of the unit region may be used to calculate the vegetation index value of the unit region.

[0057] [Summary of this disclosure] The disease detection method according to the present disclosure includes the steps of: acquiring a target image in which vegetation is captured; and calculating a vegetation index value for each unit area in an arbitrary target area within the target image; The method includes the steps of classifying each unit area in the target area into a disease class and a health class based on the vegetation index value, calculating the inter-class variance between the disease class and the health class, and determining whether the target area is a disease-infested area by comparing the inter-class variance with a predetermined judgment threshold. According to such a method, the presence or absence of disease can be determined with high accuracy even in vegetation where the growth rate varies from individual to individual, thereby realizing a highly robust disease detection method.

[0058] In the disease detection method of the present disclosure, it is preferable that the target image is a spectral image or an RGB image corresponding to one or more wavelengths, and the vegetation index value is a pixel value of an arbitrary wavelength, a pixel value of an arbitrary color, or a normalized vegetation index.

[0059] In the disease detection method according to the present disclosure, before the step of calculating the vegetation index value, a type of the vegetation index value may be selected according to a type of disease set as a detection target.

[0060] The disease detection method according to the present disclosure may further include the steps of acquiring training data including images of vegetation in a healthy state and images of vegetation in a diseased state, and training a classifier on the inter-class variance of the vegetation index values ​​calculated from each image of the training data, and determining, as the judgment threshold, a classification threshold used by the classifier to classify each image of the training data into a healthy group and a diseased group.

[0061] In the disease detection method according to the present disclosure, the training data learned by the classifier may further include labels indicating the presence or absence of a disease and the type of disease corresponding to each image, and the judgment threshold may be determined for each type of disease based on the labels.

[0062] The disease detection method of the present disclosure may further include a step of providing a user with countermeasure information corresponding to the type of disease when the target area is determined to be a disease-infested area based on the judgment threshold corresponding to any type of disease.

[0063] In the disease detection method of the present disclosure, the target image is associated with imaging location information, and if it is determined that the target area is a disease occurrence area, a step of providing a user with the location of the disease occurrence based on the imaging location information may be further implemented.

[0064] The disease detection device of the present disclosure comprises an image acquisition unit that acquires a target image in which vegetation is captured, a vegetation index value calculation unit that calculates a vegetation index value for each unit area in any target area within the target image, an inter-class variance calculation unit that classifies each unit area in the target area into a disease class and a health class based on the vegetation index value and calculates the inter-class variance between the disease class and the health class, and a judgment unit that judges whether or not disease has occurred in the target image by comparing the inter-class variance with a predetermined judgment threshold. According to the disease detection device of the present disclosure, it is possible to achieve the same effects as the above-mentioned disease detection method. [Explanation of symbols]

[0065] 1...disease detection system, 10...imaging unit, 11...imaging section, 12...GPS receiving section, 13...imaging control section, 20...disease detection device, 21...memory section, 22...processor, 221...image acquisition section, 222...area setting section, 223...vegetation index value calculation section, 224...inter-class variance calculation section, 225...judgment section, 226...output section, 227...label section, 228...threshold setting section, I...target image, R...target area.

Claims

1. acquiring a target image in which vegetation is captured; calculating a vegetation index value for each unit area in any target area within the target image; classifying each of the unit areas in the target area into a disease class and a health class based on the vegetation index value, and calculating an inter-class variance between the disease class and the health class; A disease detection method that includes a step of determining whether the target area is a disease-infested area by comparing the inter-class variance with a predetermined decision threshold.

2. the target image is a spectral image or an RGB image corresponding to one or more wavelengths, The disease detection method according to claim 1 , wherein the vegetation index value is a pixel value of an arbitrary wavelength, a pixel value of an arbitrary color, or a normalized vegetation index.

3. 3. The disease detection method according to claim 2, further comprising the step of selecting a type of the vegetation index value according to a type of disease set as a detection target, before the step of calculating the vegetation index value.

4. acquiring training data including images of healthy vegetation and images of diseased vegetation; 2. The disease detection method according to claim 1, further comprising the steps of: training a classifier on the inter-class variance of the vegetation index value calculated from each image of the training data; and determining, as the judgment threshold, a classification threshold by which the classifier classifies each image of the training data into a healthy group and a diseased group.

5. The training data trained by the classifier further includes a label indicating the presence or absence of a disease and a type of the disease corresponding to each of the images, The disease detection method according to claim 4 , wherein the determination threshold is determined for each type of disease based on the label.

6. The disease detection method described in claim 1, further comprising the step of providing a user with countermeasure information corresponding to the type of disease when the target area is determined to be a disease-infested area based on the judgment threshold corresponding to any type of disease.

7. The target image is associated with imaging position information, The disease detection method according to claim 1 , further comprising the step of providing a user with a location of the disease occurrence based on the image capture position information when the target area is determined to be a disease occurrence area.

8. an image acquisition unit for acquiring a target image in which vegetation is captured; a vegetation index value calculation unit that calculates a vegetation index value for each unit area in an arbitrary target area within the target image; an inter-class variance calculation unit that classifies each of the unit areas in the target area into a disease class and a health class based on the vegetation index value and calculates an inter-class variance between the disease class and the health class; A disease detection device comprising: a judgment unit that judges whether or not a disease has occurred in the target image by comparing the inter-class variance with a predetermined judgment threshold.

Images