Leaf fall detection method

Abstract

To provide a leaf abscission detection system and a leaf abscission detection method for accurately detecting leaf abscission of trees, and detecting leaf abscission due to seasonality and diseases with ease and precision.SOLUTION: A leaf abscission detection system 1 comprises: a cloud computing environment 21 including an obtaining unit (24) which obtains a specific index value capable of evaluating a condition of vegetation of trees (Hevea brasiliensis PG), the index value being calculated from a specific product having an appropriately-corrected observance angle for noise reduction as a result of prescribed processing being performed on various observance information relevant to the vegetation of the trees, the information being acquired by a satellite (AS) with frequency of once or more every month for a specific rubber plantation (GF) which is an arbitrary point on the ground surface in regions (RE) with much rainfall on the earth (BE); and a detection unit (local PC 31) which detects a leaf abscission status of the trees (PG) on the basis of the specific index value obtained by the obtaining unit (24).SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention evaluates the vegetation status of trees in a farm. Leaf fall detection method In particular, detecting the fallen leaves of rubber trees (Hevea brasiliensis) Leaf fall detection method Regarding. [Background technology]

[0002] Conventionally, a disease determination method capable of distinguishing between diseased trees (infected trees) and healthy trees (non-infected trees) has been proposed as one method for evaluating the vegetation state of Para rubber trees (see Patent Document 1).

[0003] The disease assessment method described in Patent Document 1 compares monochrome images of rubber tree plantations acquired using a satellite with images sensitive to the near-infrared region to determine whether the rubber trees are infected with white root rot. This method does not require the assessor to have any special experience or skill, making it easy to determine whether the rubber trees are infected with the disease. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2014-10058 Summary of the Invention [Problem to be solved by the invention]

[0005] However, determining whether a tree is suffering from leaf blight or other seasonal or disease-related defoliation requires the expert to have special experience and skill, as the above-mentioned conventional method of determining whether a rubber tree is suffering from leaf blight or other disease-related defoliation cannot determine whether the tree is suffering from leaf blight or other disease-related defoliation.

[0006] In particular, rubber tree plantations are widely distributed in tropical regions where there is a high probability of cloud formation and abundant rainfall.

[0007] Artificial satellites usually use optical sensors to capture (photograph) images of the Earth's surface. Therefore, depending on the optical sensor installed, satellites may not be able to capture images of the Earth's surface if there are clouds.

[0008] In other words, in order to accurately determine whether a rubber tree is infected with leaf blight, a system that can capture images of changes in the condition of the rubber tree over time using an artificial satellite, or a method for determining whether the tree is infected, was needed.

[0009] To achieve this, it is necessary to use an optical sensor that takes images frequently. In other words, by selecting only cloud-free images with good ground surface conditions from the many images taken and averaging them, it becomes possible to accurately capture changes in the condition of the rubber trees over time.

[0010] However, satellite optical sensor data is very noisy, making it difficult to accurately capture time-series changes in the condition of rubber trees using images that are close to the raw data.

[0011] Furthermore, simply averaging sensor data from multiple days can reduce noise, but it also eliminates signals that indicate time-series changes in the condition of the rubber trees. As a result, averaged data cannot capture the true changes in the condition of the rubber trees.

[0012] Thus, there has been a demand for the establishment of a technique that can easily detect seasonal and disease-induced defoliation of trees without requiring experience or skill.

[0013] The object of the present invention is to provide a method for accurately detecting defoliation of trees, and to easily and accurately detect defoliation due to seasonal and disease damage. Leaf fall detection method The purpose is to provide [Means for solving the problem]

[0014] According to one aspect of the present invention Leaf fall detection methodteeth, The leaf fall detection system is implemented in a cloud computing environment. Based on tree vegetation acquired by satellite at least once a month View For measurement information Place The observation angle is properly corrected to reduce noise by applying a certain process. Ru, R NIR is the spectral reflectance in the near infrared region, and R Red is the spectral reflectance of red light, (R NIR -R Red ) / (R NIR +R Red ) to obtain the normalized vegetation index as a specific index value. and, The leaf fall detection system performs the following on a local PC without using on-site leaf fall data of the vegetation. Based on the obtained specific index value, and a step of detecting a decrease in the Normalized Difference Vegetation Index as defoliation due to seasonality or disease, wherein the step of obtaining the Normalized Difference Vegetation Index as the specific index value is characterized in that the observation information is a satellite image taken at a resolution of 5 km or less by MODIS mounted on the artificial satellite Terra or Aqua, and the specific product is an MCD43A4 product that has been subjected to the predetermined processing of the observation information, including correction for the MODIS shooting angle and sunlight angle and correction for cloud image removal, and noise removal by moving average, and the Normalized Difference Vegetation Index is calculated. [Effects of the Invention]

[0016] Like this Leaf fall detection method According to the study, by using specific index values ​​that can properly evaluate the state of tree vegetation extracted from satellite observation data, it is possible to easily and accurately detect changes in tree condition (defoliation) due to seasonality and disease.

[0017] In particular, seasonal and disease-induced defoliation of trees can be easily detected through analysis in a cloud computing environment (cloud analysis) without downloading observation data from satellites. [Brief explanation of the drawings]

[0018] [Figure 1] FIG. 1 is a schematic diagram illustrating an example of the configuration of a fallen leaf detection system according to an embodiment. [Figure 2] 2(a), 2(b), and 2(c) are graphs showing the normalized difference vegetation index MCD43A4_NDVI used in the leaf fall detection system according to the embodiment in comparison with the vegetation indices of other MODIS products. [Figure 3] Figure 3 is a graph illustrating field defoliation data (actual annual coverage variation). [Figure 4]Figures 4(a), 4(b), 4(c), and 4(d) are schematic diagrams for explaining the annual fluctuations in coverage due to defoliation. [Figure 5] FIG. 5 shows an example of evaluation of the leaf fall detection system according to the embodiment, where FIG. 5(a) is a graph of annual variation in coverage, and FIG. 5(b) is a graph of the Normalized Difference Vegetation Index MCD43A4_NDVI. [Figure 6] FIG. 6 is a schematic diagram showing an example of an analysis screen display of the fallen leaf detection system according to the embodiment. [Figure 7] FIG. 7 is a diagram for explaining an analysis environment applicable to the fallen leaf detection system according to the embodiment. [Figure 8] Fig. 8 is a schematic diagram illustrating another embodiment of a defoliation prediction using a time series model. Fig. 8(a) is a time series diagram showing actual vegetation index data and defoliation prediction using a time series model, Fig. 8(b) is a time series diagram showing the annual fluctuation trend of the vegetation index, and Fig. 8(c) is a diagram showing seasonal changes in the vegetation index. [Figure 9] FIG. 9 is a schematic diagram showing an example of a screen display for visualizing fallen leaves using a spatial hazard map, as another embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0019] Hereinafter, a leaf falling detection system according to an embodiment will be described with reference to the drawings.

[0020] (Schematic configuration) Fig. 1 shows an example of the configuration of a leaf fall detection system 1 according to an embodiment. As shown in Fig. 1, the leaf fall detection system 1 is configured to detect seasonal and disease-induced leaf fall from trees based on sensor data (satellite data) from an optical sensor IS mounted on an Earth observation satellite AS.

[0021] Here, we will explain a case where the target of evaluation of the vegetation state (leaf growth) of trees is, for example, defoliation of a rubber tree PG in a specific rubber tree plantation GF, which is an arbitrary point on the ground surface in a rainy region RE on Earth BE. The defoliation of the rubber tree PG can be easily detected by the defoliation detection system 1.

[0022] That is, the leaf fall detection system 1 is configured such that, for example, a cloud computing environment 21 is connected to a local PC (personal computer) 31 of the rubber tree farm GF, a smartphone 32, etc. via an internet line INC.

[0023] The cloud computing environment 21 is an online platform for analyzing satellite images. For example, the "Google Earth Engine (GEE)" developed by Google is used as the cloud computing environment 21. This cloud computing environment 21 makes it easy to calculate, process, and analyze images of observation data (sensor data).

[0024] Google Earth Engine is an analytical environment that uses remote sensing and a cloud-based GIS (geographic information system). With Google Earth Engine, instructions can be sent from the local PC 31 at the rubber tree plantation GF to Google's server via the internet line INC, and the results can be sent back after all the analysis has been completed. This means that, for example, at the rubber tree plantation GF, advanced image analysis, such as detecting fallen leaves, can be easily performed and shared without the need for expensive satellite image analysis software.

[0025] As shown in FIG. 1, the cloud computing environment 21 includes, for example, a map vector data storage unit 22, an acquisition unit 24 that acquires a normalized difference vegetation index MCD43A4_NDVI as a specific index value, which will be described later, and the like.

[0026] When detecting fallen leaves of a rubber tree PG, the vector data storage unit 22 temporarily stores vector data of the corresponding area in response to an instruction from the local PC 31 of the rubber tree plantation GF, for example.

[0027] The vector data is, for example, map data in which contour correction is applied to a map of a specific area including the rubber tree plantation GF, the site to be detected, and mass data for displaying blocks is overlaid. This vector data is prepared in advance and, for example, uploaded from the local PC 31 of the rubber tree plantation GF when fallen leaves are detected.

[0028] The acquisition unit 24 acquires, for example, a portion corresponding to the vector data stored in the vector data storage unit 22 from a specific product (MCD43A4 product) stored in the cloud computing environment 21. For the acquired portion of the product, the acquisition unit 24 acquires the normalized difference vegetation index MCD43A4_NDVI, which can evaluate the state of tree vegetation.

[0029] In the cloud computing environment 21, after the vector data is read into the vector data storage unit 22, the acquisition unit 24, for example, calculates the annual daily fluctuation value (time series data) of the Normalized Difference Vegetation Index MCD43A4_NDVI for each plot.

[0030] Generally, tree leaves absorb red light (620-690nm) and reflect light in the near-infrared range (720-1200nm). Therefore, the normalized difference vegetation index MCD43A4_NDVI is (R NIR -R Red ) / (R NIR +R Red ) where R NIR is the spectral reflectance in the near infrared region, and R Red is the spectral reflectance of red light.

[0031] That is, in this embodiment, the Normalized Difference Vegetation Index (MCD43A4_NDVI) calculated from the MCD43A4 product ("Nadir BRDF-Adjusted Reflectance 16-Day L3 Global 500m"), which is a MODIS product, is used as an index for detecting defoliation (leaf growth) of Para rubber tree PG due to seasonality and disease.

[0032] The local PC 31 and smartphone 32 at the rubber tree farm GF are used to obtain various products derived from the observation data acquired by the satellite AS via the Internet line INC, and to input instructions necessary for detecting fallen leaves.

[0033] In this embodiment, the detection unit is configured by the local PC 31. That is, the local PC 31 has, for example, a database server, and the database server uses any software to graph the annual daily fluctuation values ​​of the Normalized Difference Vegetation Index MCD43A4_NDVI calculated by the acquisition unit 24. Furthermore, by analyzing the increase or decrease in the daily fluctuation values, defoliation of the Hevea trees PG in the rubber tree plantation GF is detected. The any software may be, for example, Microsoft Excel.

[0034] The analysis for detecting fallen leaves does not have to be performed on the database server in the local PC 31. For example, it may be performed in the cloud computing environment 21 or in another cloud computing environment. Examples of the other cloud computing environment include "Microsoft Azure," "Amazon Web Services (AWS)," or "Google Cloud Platform," which will be described later, other than GEE.

[0035] Local PC31 is also used to evaluate detected Hevea PG defoliation using field defoliation data (details will be discussed later).

[0036] Here, the observation data acquired by the artificial satellite AS is received by a ground observation center 11, which serves as a management unit. The ground observation center 11 makes various products derived from the observation data available to the public on the Internet line INC.

[0037] In this embodiment, a Moderate Resolution Imaging Spectroradiometer (MODIS) is used as the optical sensor IS. This MODIS is currently installed on artificial satellites (also called MODIS satellites) such as the Terra satellite and the Aqua satellite managed by the National Aeronautics and Space Administration (NASA) and the like.

[0038] That is, the ground observation center 11 includes a product generation unit 12 that generates various MODIS products based on MODIS data captured by an optical sensor IS as the satellite AS moves. The product generation unit 12 is, for example, a NASA server. In particular, the product generation unit 12 generates a specific product (MCD43A4 product, 500m resolution, daily Earth surface reflectance) as a MODIS product.

[0039] In this embodiment, observation information according to tree vegetation is acquired at least once a month by an artificial satellite AS. For example, images of the rubber tree plantation GF are taken regularly every day by MODIS (resolution of 250 m to 1 km, at least 5 km or less). Then, MODIS data with a resolution of 500 m is acquired as observation information.

[0040] For example, even if the plantation area of ​​a rubber tree plantation GF is vast, by using an artificial satellite AS, it is possible to easily photograph the entire plantation area using remote sensing.

[0041] The specific product (MCD43A4 product) is an average of MODIS data for eight days before and after the data has been subjected to noise reduction, such as moving averaging, removal of cloud images, and correction of observation angles (such as the shooting angle and sunlight angle). The MCD43A4 product eliminates the effects of rain clouds and other factors, reducing noise to provide clear observation information. The MCD43A4 product enables the leaf fall detection system 1 to obtain accurate observation information that more faithfully reproduces the original vegetation conditions of the rubber tree PG when detecting defoliation of the rubber tree PG. Therefore, by calculating, for example, a specific index value (MCD43A4_NDVI) from the MCD43A4 product, changes in the vegetation conditions of the rubber tree PG can be easily captured. The calculated specific index value (MCD43A4_NDVI) includes a signal indicating time-series changes in conditions due to factors such as defoliation.

[0042] (Evaluation of the Normalized Difference Vegetation Index MCD43A4_NDVI) Figures 2(a) to 2(c) show a comparison of the time-series variation patterns of vegetation indices from MODIS products. Here, we explain using observation data from January to December 2019 as an example.

[0043] For example, Figure 2(a) shows the vegetation index of a low-level processed MODIS product (TERRA_500m_daily). Figure 2(b) shows the vegetation index (spatial resolution 250m, time composite 16 days) of a MODIS product (MOD13Q1), which is processed data with twice the resolution of Figure 2(a) and spatially averaged and time composited. Figure 2(c) shows the normalized vegetation index (MCD43A4_NDVI) of the MCD43A4 product (spatial resolution 500m mesh), which is processed data that has been corrected for observation angle, spatially averaged, and averaged over 16 days.

[0044] Figure 3 shows on-site defoliation data (on-site defoliation data) for the year 2019. Here, on-site defoliation data (on-site defoliation data) refers to the annual fluctuations in the coverage rate of Para rubber trees (PG) within the actual rubber tree plantation GF (annual coverage rate fluctuations).

[0045] The coverage rate of Hevea brasiliensis PG drops significantly with defoliation during the dry season, as is clear from Figure 3. In other words, as shown in the March data indicated by the arrow in Figure 3, the defoliation of Hevea brasiliensis PG appears as a decrease in coverage rate.

[0046] As shown in Figure 2(c), the normalized difference vegetation index (MCD43A4_NDVI) of the MCD43A4 product has a fluctuation pattern similar to that of the field leaf fall data shown in Figure 3.

[0047] In other words, since the graph of the Normalized Difference Vegetation Index MCD43A4_NDVI is similar to the graph of actual annual variation in coverage, it is possible to accurately detect seasonal defoliation of rubber tree PG, as well as defoliation due to diseases such as leaf blight.

[0048] In contrast, the influence of noise in the MODIS product in Figure 2(a) makes it difficult to see the trend in the time series vegetation index for 2019. Similarly, the MODIS product in Figure 2(b) does not confirm the decline in the vegetation index due to defoliation seen in the March data for the coverage rate of Para rubber trees (PG) in Figure 3. For this reason, it is difficult to detect the trend in the time series vegetation index for 2019 in the MODIS products shown in Figures 2(a) and 2(b).

[0049] In this way, according to the Normalized Difference Vegetation Index MCD43A4_NDVI shown in Figure 2(c), seasonal defoliation of Hevea brasiliensis PG can be determined from a decrease in the Normalized Difference Vegetation Index MCD43A4_NDVI in March, etc. Furthermore, defoliation due to disease can be determined from a decrease in the Normalized Difference Vegetation Index MCD43A4_NDVI in other months.

[0050] Whether the decline in the Normalized Difference Vegetation Index MCD43A4_NDVI in March is due to seasonality or disease can be determined, for example, by the recovery status of the Normalized Difference Vegetation Index MCD43A4_NDVI after March. In other words, if the decline in the Normalized Difference Vegetation Index MCD43A4_NDVI does not recover even in April, it can be assumed that the defoliation in March is not simply due to seasonality.

[0051] Here, we will briefly explain how to calculate the on-site leaf fall data.

[0052] Figures 4(a) to 4(c) are schematic diagrams for explaining the annual variation in coverage due to defoliation of Para rubber trees (PG). Figure 4(a) illustrates an example of a method for taking panoramic photographs (AP). Figure 4(b) illustrates an example of a panoramic photograph (AP) when the canopy coverage is 26.9%. Figure 4(c) illustrates an example of a panoramic photograph (AP) when the canopy coverage is 72.4%. Figure 4(d) illustrates a graph of the annual variation in coverage.

[0053] An all-sky photograph AP is a panoramic image taken periodically from directly above the forest floor of the Para rubber tree PG at each observation point (block), as shown in Figure 4(a), for example. In other words, an all-sky photograph AP is the result of fixed-point observation of the leaf expansion / defoliation status of the Para rubber tree PG for each arbitrarily divided block (actual, on-site defoliation data). Note that the all-sky photograph AP may be taken with a digital camera DC using a circular fisheye lens (all-sky lens).

[0054] The canopy coverage rate is the percentage of the sky covered by the leaves of Para rubber trees (PG), calculated using free all-sky photo analysis software (e.g., lia32 by Kazukiyo Yamamoto https: / / www.agr.nagoya-u.ac.jp / ~shinkan / LIA32 / ) from all-sky photos AP taken at each observation point once a month. Therefore, it can be used as a quantitative value for the change in the amount of leaves at the site.

[0055] The annual change in canopy coverage is the change in canopy coverage over 12 months (time series data) for each block of mass data at each observation point. The annual change in canopy coverage can be graphed using any software (for example, Microsoft Excel).

[0056] Therefore, by comparing the graph of the time series data of the Normalized Difference Vegetation Index MCD43A4_NDVI with the graph of the on-site defoliation data in other cloud computing environments or local PC31, it is possible to evaluate the detection of defoliation of Para rubber tree PG.

[0057] 5 shows an example of an evaluation of the leaf fall detection system according to the embodiment. Fig. 5(a) is a graph of the annual coverage fluctuation (field leaf fall data) in 2019, and Fig. 5(b) is a graph of the time series data of the Normalized Difference Vegetation Index (MCD43A4_NDVI) in 2019.

[0058] As is clear from Figures 5(a) and 5(b), the graphs of annual coverage fluctuations and the graphs of the time-series data of the Normalized Difference Vegetation Index MCD43A4_NDVI show nearly identical (synchronized) fluctuation patterns. This means that seasonal and disease-induced defoliation of Hevea brasiliensis PG can be accurately detected based on the graphs of the time-series data of the Normalized Difference Vegetation Index MCD43A4_NDVI.

[0059] That is, the inventors of the present application discovered that the graph of the Normalized Difference Vegetation Index MCD43A4_NDVI of the MCD43A4 product (see Figure 2(c)) when the resolution is set to, for example, 500 m closely matches the field defoliation data in Figure 4(c). By using the Normalized Difference Vegetation Index MCD43A4_NDVI, it is possible to accurately detect seasonal and disease-induced defoliation of Hevea brasiliensis PG without using field defoliation data where it is difficult to obtain panoramic photographs AP.

[0060] FIG. 6 shows an example of an analysis screen displayed by a web browser on the screen MN of the local PC 31 when detecting fallen leaves of a rubber tree PG in a rubber tree plantation GF.

[0061] In other words, when detecting fallen leaves of Para rubber tree PG, the analysis screen displays data selection for selecting the site to be detected, the instructed analysis content, data and graphs for image analysis, and analysis images for detecting fallen leaves.

[0062] In this way, by using a web browser, cloud analysis is possible in a cloud computing environment without having to download large amounts of satellite data. In other words, by utilizing the computer resources on the network cloud, it is possible to effectively detect defoliation of Hevea brasiliensis PG within the vast rubber tree plantation GF.

[0063] (Operation procedure) Next, a process flow for detecting fallen leaves of a rubber tree PG (a method for detecting fallen leaves) will be described using a specific example.

[0064] For example, a specific product (MCD43A4 product) is generated in the product generation unit 12 of the ground observation center 11 daily from January 2019 based on MODIS data including the rubber tree plantation GF photographed by the optical sensor IS of the satellite AS.

[0065] In response to this, first, the vector data of the rubber tree farm GF, which has been prepared in advance, is uploaded from the local PC 31 and read into the vector data storage unit 22 of the cloud computing environment 21 via the Internet line INC.

[0066] Thereafter, the acquisition unit 24 of the cloud computing environment 21 imports a specific product (MCD43A4 product) corresponding to the vector data read into the vector data storage unit 22. Then, the normalized difference vegetation index MCD43A4_NDVI, which can be used to evaluate the vegetation state of the rubber tree PG, is extracted.

[0067] Furthermore, for example, the acquisition unit 24 in the cloud computing environment 21 uses "Google Earth Engine" to calculate the annual daily fluctuation value of the Normalized Difference Vegetation Index MCD43A4_NDVI for each block of the rubber tree plantation GF. Then, the annual daily fluctuation value of the Normalized Difference Vegetation Index MCD43A4_NDVI is output using, for example, a CSV file (Comma Separated Value File) and transferred to the local PC 31 of the rubber tree plantation GF via the Internet line INC.

[0068] Meanwhile, the local PC 31 creates a graph of the time-series data of the Normalized Difference Vegetation Index MCD43A4_NDVI based on the received annual daily fluctuation values, and then analyzes the increase / decrease trends in the graph of the annual daily fluctuation values ​​of the time-series data of the Normalized Difference Vegetation Index MCD43A4_NDVI to detect defoliation of the Para rubber tree PG due to seasonality and disease.

[0069] At this time, an analysis screen such as that shown in Fig. 6 is displayed on the screen MN of the local PC 31. The analysis results of the local PC 31 are also displayed on the screen MN. In this way, the series of processes is completed.

[0070] As described above, according to the embodiment, a leaf fall detection system 1 can be provided that can accurately detect leaf fall of rubber trees PG and can easily detect leaf fall due to seasonality and disease without requiring experience or skill.

[0071] (Actions and Effects) According to the above-described embodiment, the following actions and effects can be obtained.

[0072] That is, in this embodiment, MODIS data from a MODIS satellite capable of taking images daily is used, and the vegetation index NDVI is extracted from the MCD43A4 product, which is a MODIS product, to detect defoliation of the rubber trees PG in the rubber tree plantation GF.

[0073] This makes it possible to use the fluctuation pattern of the annual daily fluctuation value of the normalized difference vegetation index MCD43A4_NDVI to detect defoliation of Para rubber tree PG.

[0074] Therefore, even if the rubber tree plantation GF is located in a particularly rainy area, the impact of data loss due to rain clouds can be reduced, and the defoliation status of the Para rubber tree PG can be detected more accurately.

[0075] While the example given above uses MODIS data from the MODIS satellite as observation information from the artificial satellite AS, optical sensors such as those mounted on Landsat satellites, which have a resolution of 30m and a 16-day imaging frequency, also have sufficient specs for image analysis. However, when using MODIS data with an extremely high imaging frequency, it is possible to further reduce issues such as the loss of important signals showing time-series changes in the condition of the Para rubber tree PG when noise is reduced.

[0076] Moreover, seasonal and disease-induced defoliation can be easily detected without requiring special experience or skill.

[0077] In particular, it is possible to easily grasp the disease status of leaf blight, one of the major diseases of rubber tree PG, over a wide area within a rubber tree plantation GF, for which no effective diagnostic method has been established.

[0078] In the above-described embodiment, the normalized difference vegetation index MCD43A4_NDVI of the MCD43A4 product when the resolution of the MODIS data is set to 500 m is used to detect defoliation of the Para rubber tree PG, but this is not limited to this.

[0079] That is, when detecting leaf fall of various trees with a resolution of 500 m or more, it is possible to arbitrarily combine different satellites and index values ​​(such as the Extended Vegetation Index EVI and the Normalized Water Index NDWI) depending on the application, as shown in Figure 7.

[0080] For example, by taking into account the water status (the amount of water in vegetation indicated by the Normalized Water Index (NDWI) or other indicators) and temperature status, it will be possible to detect defoliation of trees with higher accuracy.

[0081] In particular, we decided to use "Google Earth Engine" due to the amount of data and ease of analysis, but it is also possible to use analysis environments other than "Google Earth Engine," such as Amazon's "Amazon Web Services (AWS)" and "Earth on AWS," Microsoft's "AI for Earth," "Microsoft Azure," or "Google Cloud Platform."

[0082] In this way, depending on the combination of satellite type, product type, index value, and analysis environment, the system can be applied to other sustainable applications other than detecting fallen leaves on trees for farm management purposes, such as resource management, exploration, and disasters.

[0083] In addition, as other embodiments, for example, it is possible to predict leaf fall using a time series model (see FIG. 8) or visualize leaf fall using a spatial hazard map (see FIG. 9).

[0084] The above describes the leaf fall detection system 1 based on an embodiment, but the present invention is not limited to this, and the configuration of each part can be replaced with any configuration that has similar functions.

[0085] The entire contents of Patent Application No. 2020-173119 (filing date: October 14, 2020) are incorporated herein by reference. [Explanation of symbols]

[0086] 1 Leaf fall detection system 11 Ground observation center 21 Cloud computing environment 22 Vector data storage unit 24 Acquisition unit 31 Local PC (detection unit) AS Satellite (MODIS satellite) GF Rubber tree plantation IS Optical sensor (MODIS) INC Internet connection NDVI Normalized Difference Vegetation Index (specific index value) PG Para rubber tree (tree)

Claims

1. The leaf fall detection system acquires, as a specific index value, a normalized vegetation index calculated from a specific product in which the observation angle is appropriately corrected for noise reduction by performing predetermined processing on observation information according to tree vegetation acquired by a satellite at least once per month in a cloud computing environment, where R NIR is the spectral reflectance in the near-infrared region and R Red is the spectral reflectance of red light; a step of detecting a decrease in the normalized difference vegetation index as defoliation due to seasonality or disease based on the specific index value obtained by the defoliation detection system on a local PC without using defoliation data of the vegetation at the site; Equipped with The step of obtaining the normalized difference vegetation index as the specific index value includes: The observation information is a satellite image taken at a resolution of 5 km or less by MODIS mounted on the artificial satellite Terra or Aqua, A leaf fall detection method characterized in that the specified processing of the observation information involves correction of the MODIS shooting angle and sunlight angle, and correction to remove cloud images, and the MCD43A4 product from which noise has been removed using a moving average is used as the specific product to calculate the normalized vegetation index.

2. 2. The method for detecting fallen leaves according to claim 1, wherein, in addition to the normalized difference vegetation index, one of water status and temperature status is further obtained as the specific index value.

Images