Determination method, determination apparatus, machine learning method, machine learning apparatus, and program

Abstract

To provide a technique that enables easy and highly accurate determination of whether a crop has undergone young panicle differentiation.SOLUTION: A determination method includes: an acquisition step of acquiring sensing result data indicating a result of sensing an internode of a target crop collected in a field that is exposed by removing a leaf sheath of the target crop; a determination step of determining whether the target crop has undergone young panicle differentiation by inputting the sensing result data acquired in the acquisition step into a trained model that has been machine-learned using, as input data, data indicating a result of sensing an internode of a crop exposed by removing a leaf sheath of the crop, and using, as output data, the presence or absence of young panicle differentiation of the crop, thereby learning a correlation relationship between the input data and the output data; and an output step of outputting a determination result in the determination step.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a technique for determining whether or not panicle differentiation has occurred in a crop. [Background technology]

[0002] Methods for determining whether panicle differentiation has occurred in rice and other crops have been proposed. Because the panicle differentiation period in rice is particularly important for yield and quality, determining whether panicle differentiation has occurred is important for effective fertilization and efficient water management. For example, Non-Patent Document 1 describes a method in which, after removing a plant from the field, the base is carefully dissected using a scalpel or tweezers under a microscope, and the exposed panicle is observed under a microscope to determine whether it is at the first bract primordium differentiation stage (panicle differentiation stage) or a later developmental stage. Non-Patent Document 2 also describes that the internodes of rice plants turn green before and after panicle differentiation, and this is used as an indicator of the transition from the vegetative to the reproductive growth stage. Patent Document 1 also describes a method for determining the rice growth stage, including panicle differentiation, from images of rice canopies. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6638121 [Non-patent literature]

[0004] [Non-Patent Document 1] Fushimi et al., 2021, A quantitative staging system for describing rice panicle development and its application for a crop phenological model. Agronomy Journal,113:5040-5053. [Non-patent document 2] Counce et al., 2000, A Uniform, Objective, and Adaptive System for Expressing Rice Development. Crop Science,40:436-443. Summary of the Invention [Problem to be solved by the invention]

[0005] However, the method described in Non-Patent Document 1 requires advanced techniques to dissect rice plants under a microscope, which is time-consuming and labor-intensive. First, the plant must be carefully dissected using a scalpel or tweezers to avoid damaging the shoot apex where the flower buds are located, and then the young panicles must be observed under a microscope. This makes it impossible (and therefore unrealistic) to determine whether or not young panicle differentiation has occurred in actual production sites. Furthermore, the method described in Non-Patent Document 2 involves visual inspection, which results in non-quantitative and variable results. Furthermore, the method described in Non-Patent Document 2 focuses not only on the young panicle differentiation stage but also on a broad range of young panicle developmental stages, making it impossible to pinpoint the most important young panicle differentiation stage. Furthermore, the method described in Patent Document 1 makes it difficult to distinguish morphological changes in young panicles on a 0.1 mm scale, which are invisible from the outside, from changes in the shape of the rice canopy. Thus, the above-mentioned conventional techniques have the drawback of being unable to easily and accurately determine whether or not a crop has young panicle differentiation.

[0006] An object of one aspect of the present invention is to realize a technique for easily and accurately determining whether or not panicle differentiation has occurred in a crop. [Means for solving the problem]

[0007] In order to solve the above problem, a method for determining young panicle differentiation according to one embodiment of the present invention includes an acquisition step of acquiring sensing result data indicating the results of sensing the internodes of a target crop that are exposed by removing the leaf sheaths of the target crop collected in the field; a discrimination step of determining the presence or absence of young panicle differentiation of the target crop by inputting the sensing result data acquired in the acquisition step into a trained model that has machine-learned the correlation between the input data and the output data, with the data indicating the results of sensing the internodes of the target crop that are exposed by removing the leaf sheaths of the crop as input data and the presence or absence of young panicle differentiation of the target crop as output data; and an output step of outputting the discrimination result of the discrimination step.

[0008] In addition, in order to solve the above-mentioned problem, a machine learning method according to one embodiment of the present invention includes an acquisition step of acquiring first data indicating the results of sensing the internodes of a crop that are exposed by removing the leaf sheaths of the crop collected in a field, and second data indicating whether or not young panicle differentiation has occurred in the crop, and a training step of using the first data as input data and the second data as output data, associating the two, and machine learning a learned model.

[0009] In addition, in order to solve the above-mentioned problems, a discrimination device according to one embodiment of the present invention comprises an acquisition unit that acquires sensing result data indicating the results of sensing the internodes of a target crop that are exposed by removing the leaf sheaths of the target crop collected in a field; a discrimination unit that uses the data indicating the results of sensing the internodes of the target crop that are exposed by removing the leaf sheaths as input data and the presence or absence of young panicle differentiation of the target crop as output data, and inputs the sensing result data acquired by the acquisition unit into a trained model that has machine-learned the correlation between the input data and the output data; and an output unit that outputs the discrimination result obtained by the discrimination unit.

[0010] In addition, in order to solve the above-mentioned problems, a machine learning device according to one embodiment of the present invention includes an acquisition unit that acquires first data indicating the results of sensing the internodes of a crop that are exposed by removing the leaf sheaths of the crop collected in a field, and second data indicating whether or not young panicle differentiation has occurred in the crop, and a training unit that uses the first data as input data and the second data as output data, associates the two, and machine-learns a learned model.

[0011] The discrimination device according to each aspect of the present invention may be realized by a computer. In this case, the discrimination device program that causes the computer to operate as each part (software element) of the discrimination device to realize the discrimination device on a computer, and the computer-readable recording medium on which the program is recorded, also fall within the scope of the present invention.

[0012] In addition, the machine learning device according to each aspect of the present invention may be realized by a computer. In this case, the machine learning device program that realizes the machine learning device on a computer by causing the computer to operate as each part (software element) of the machine learning device, and the computer-readable recording medium on which the program is recorded, also fall within the scope of the present invention. [Effects of the Invention]

[0013] According to one aspect of the present invention, panicle differentiation of a crop can be easily and accurately determined. [Brief explanation of the drawings]

[0014] [Figure 1] FIG. 1 is a diagram showing the growth process of rice. [Figure 2] FIG. 1 is a diagram showing an example of a method for observing young panicles of rice. [Figure 3] FIG. 1 is a diagram showing a schematic diagram of the relationship between the growth process of rice and the shape of the young panicle. [Figure 4] FIG. 1 is a diagram showing an example of photographed images of internodes of rice plants in the vegetative growth stage and the reproductive growth stage. [Figure 5]1 is a block diagram showing the configuration of a system for determining panicle differentiation according to an embodiment. FIG. [Figure 6] FIG. 4 is a diagram illustrating an example of data indicating a sensing result obtained by the sensing device according to the embodiment. [Figure 7] FIG. 1 is a flow chart showing the flow of a method for determining the presence or absence of panicle differentiation according to an embodiment. [Figure 8] FIG. 1 is a diagram for simply explaining a method for determining whether or not panicle differentiation has occurred according to an embodiment. [Figure 9] FIG. 1 is a flow chart showing the flow of a method for generating a trained model according to an embodiment. [Figure 10] FIG. 1 is a diagram briefly explaining a method for generating a trained model according to an embodiment. [Figure 11] FIG. 3 is a diagram showing a specific example of first data included in training data according to the embodiment. [Figure 12] FIG. 2 is a diagram showing an example of a developmental stage after panicle differentiation according to an embodiment. [Figure 13] 1 is a block diagram showing the configuration of a system for determining panicle differentiation according to an embodiment. FIG. [Figure 14] FIG. 1 is a flow chart showing the flow of a method for determining the presence or absence of panicle differentiation and the developmental stage after panicle differentiation according to an embodiment. [Figure 15] FIG. 1 is a diagram briefly explaining a method for generating a trained model according to an embodiment. [Figure 16] FIG. 10 is a diagram showing an example of a determination result obtained by the determination system according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0015] [Embodiment 1] <Outline of the discrimination system> An embodiment of the present invention will be described in detail below. The discrimination system according to this embodiment is a system for discriminating whether or not young panicle differentiation has occurred in a crop. An example of a crop to be discriminated is paddy rice. However, the crop to be discriminated is not limited to this, and may be other crops such as wheat, corn, barley, rye, sugarcane, etc. Crop fields include, but are not limited to, paddy fields and fields.

[0016] (Rice growth process) Here, the growth process of rice, an example of a crop, will be explained with reference to the drawings. Figure 1 shows the growth process of rice. As shown in Figure 1, the growth process of rice is broadly divided into the vegetative growth period, the reproductive growth period, and the ripening period. The vegetative growth period is the period from budding to the formation of the panicle primordium. The reproductive growth period is the period from the formation of the panicle primordium to heading and flowering. The ripening period is the period from heading and flowering to maturity. The panicle primordium differentiates at the meristem, a process called panicle differentiation. Panicle differentiation is a physiological and ecological response involved in the transition from vegetative growth to reproductive growth, and the panicle differentiation period is the critical period that most affects yield and quality.

[0017] FIG. 2 is a diagram showing an example of a method for observing young panicles of paddy rice. The method shown in FIG. 2 includes steps S1 to S4. In step S1, a paddy rice producer or the like harvests paddy rice from a paddy field, which is a farm field. In step S2, the producer or the like removes the leaf sheaths from the harvested crop to expose the internodes (lower stems) of the paddy rice. In step S3, the producer or the like carefully dissects the internodes of the paddy rice using a scalpel or tweezers to expose the young panicles. In step S4, the producer or the like observes the exposed young panicles under a microscope. Because the young panicles are very small, measuring approximately 0.1 mm, they must be dissected carefully, and observing the young panicles under a microscope requires skilled techniques.

[0018] Figure 3 is a diagram showing the relationship between the growth process of rice and the shape of the young panicle. In Figure 3, rice plant 201 is an example of the appearance of rice plant in the vegetative growth stage, and young panicle 202 is a young panicle exposed by breaking down internode 201a of rice plant 201. Young panicle 202 is in a state before young panicle differentiation and includes shoot apex 202a and leaf primordium 202b.

[0019] Paddy rice 203 is an example of the appearance of paddy rice during the reproductive growth stage, and young panicle 204 is a young panicle exposed by breaking down internode 203a of paddy rice 203. Young panicle 204 is in a state after young panicle differentiation and includes a first bract primordium 204a and a flag leaf primordium 204b. Paddy rice 205 is an example of the appearance of paddy rice during the ripening stage, and young panicle 206 is a young panicle exposed by breaking down internode 205a of paddy rice 205. Young panicle 206 includes a primary rachis-branch primordium 206a and a bract primordium 206b.

[0020] FIG. 4 shows examples of photographed images of internodes of rice plants in the vegetative and reproductive stages. In FIG. 4, image 301 is an image of an internode (the lower part of the stem exposed by removing the leaf sheath) of rice plants in the vegetative stage. Image 302 is an image of an internode (the lower part of the stem exposed by removing the leaf sheath) of rice plants in the reproductive stage. As shown in FIG. 4, a green ring appears in the internode of the rice plant in image 302. This ring is also called a green ring. The green ring can be used as an indicator of the transition from the vegetative stage to the reproductive stage, but it is difficult to accurately determine whether panicle differentiation has occurred by visual inspection.

[0021] <Configuration of the discrimination system> FIG. 5 is a block diagram showing the configuration of a discrimination system 1 according to this embodiment. The discrimination system 1 is a system for discriminating whether or not young panicle differentiation has occurred in a crop, and includes a discrimination device 10 and a sensing device 20. Hereinafter, the crop to be discriminated will also be referred to as the "target crop." The discrimination device 10 is a device for discriminating whether or not young panicle differentiation has occurred in a target crop, and is, for example, a general-purpose computer. The sensing device 20 is a device for sensing the target crop. An example of the sensing device 20 is a spectroscopic reflectometer that measures spectral reflectance, but is not limited to this. An example of the sensing device 20 is an imaging device that captures an image of a target and outputs image data. In this case, the image data output by the imaging device may be still image data or may be moving image data.

[0022] In the example of Fig. 5, the discrimination device 10 and the sensing device 20 are configured as separate devices, but the sensing device 20 may be built into the discrimination device 10. For example, the discrimination device 10 may be built into an imaging device. Furthermore, in the example of Fig. 5, one sensing device 20 is illustrated, but the discrimination system 1 may include multiple sensing devices 20. For example, the discrimination system 1 may include both a spectroreflectometer and an imaging device.

[0023] (discrimination device) The discrimination device 10 includes a control unit 11, a storage unit 12, a communication unit 13, and an input / output unit 14. The communication unit 13 communicates with devices external to the discrimination device 10 via a communication line. The specific configuration of the communication line does not limit the present exemplary embodiment, but examples of the communication line include a wireless LAN (Local Area Network), a wired LAN, a WAN (Wide Area Network), a public line network, a mobile data communication network, or a combination thereof. The communication unit 13 transmits data supplied from the control unit 11 to other devices and supplies data received from other devices to the control unit 11.

[0024] The input / output unit 14 receives input to the discrimination device 10, and the discrimination device 10 outputs data. Input / output devices such as a keyboard, a mouse, a display, a printer, and a touch panel are connected to the input / output unit 14. The input / output unit 14 receives various types of information input to the discrimination device 10 from the connected input devices. Furthermore, the input / output unit 14 outputs various types of information to connected output devices under the control of the control unit 11. An example of the input / output unit 14 is an interface such as a USB (Universal Serial Bus).

[0025] The input / output unit 14 is connected to the sensing device 20. The sensing result of the sensing device 20 is input to the input / output unit 14. The sensing device 20 may be connected to the communication unit 13.

[0026] The memory unit 12 stores various data used by the discrimination device 10. In particular, the memory unit 12 stores instructions of a computer program executed by the control unit 11. The memory unit 12 also stores a trained model 121. Here, storing the trained model 121 in the memory unit 12 means that parameters defining the trained model 121 are stored in the memory unit 12.

[0027] (Pre-trained model) The trained model 121 is a trained model that uses data indicating the results of sensing the internodes of a crop that are exposed by removing the leaf sheaths of the crop as input data, and the presence or absence of panicle differentiation of the crop as output data, and that machine-learns the correlation between the input data and the output data. As an example, the trained model 121 may be an SVM (support vector machine), or may be a model having a neural network structure such as a CNN (convolutional neural network). However, the trained model 121 is not limited to the above examples, and may be another model generated by machine learning.

[0028] (Input of trained model) The input of the trained model 121 is sensing result data indicating the results of sensing the internodes of the target crop that are exposed by removing the leaf sheaths of the target crop collected in the field. The sensing result data includes, for example, data indicating the measurement results of the spectral reflectance of the internodes of the target crop measured by a spectroscopic reflectometer. More specifically, the sensing result data may be a numerical data set indicating the spectral reflectance, or may be image data representing a spectral reflectance curve. The sensing result data may also include image data obtained by capturing an image of the internodes of the target crop. In this case, the image data may be data representing a still image or may be data representing a moving image. The image data may also be data representing an image with a background of a predetermined color (e.g., black).

[0029] The sensing result data may also be data obtained by processing image data of an internode of a target crop. For example, the acquiring unit 111 may process the image data by performing a predetermined color correction process on the image data of an internode of a target crop, and the image data obtained by processing may be used as the sensing result data.

[0030] (output of trained model) The output of the trained model 121 is data indicating whether or not young panicle differentiation has occurred in the target crop. The output of the trained model 121 may be, for example, binary data indicating whether or not young panicle differentiation has occurred, or may be a real value indicating the probability of whether or not young panicle differentiation has occurred.

[0031] (Control unit) The control unit 11 includes an acquisition unit 111, a discrimination unit 112, an output unit 113, a training data acquisition unit 114, and a training unit 115. The acquisition unit 111, the discrimination unit 112, the output unit 113, the training data acquisition unit 114, and the training unit 115 are realized by the control unit 11 reading and executing instructions of a computer program stored in the memory unit 12.

[0032] The acquiring unit 111 acquires sensing result data indicating the results of sensing by the sensing device 20. The acquiring unit 111 may use the data output by the sensing device 20 as the sensing result data as is, or may acquire sensing result data generated using the data output by the sensing device 20. More specifically, the acquiring unit 111 may use, for example, the data output by a spectroscopic reflectometer as the sensing result data as is, or may use the data output by the spectroscopic reflectometer to generate image data representing a spectral reflectance curve and use the generated image data as the sensing result data.

[0033] Furthermore, the acquisition unit 111 may use the image data output by the imaging device as the sensing result data, or may use data obtained by performing predetermined image processing on the image data output by the imaging device as the sensing result data.

[0034] The acquisition unit 111 may also receive sensing result data from another device connected via the communication unit 13. The acquisition unit 111 may also acquire the data by reading the data from a storage destination (which may be a storage device within the discrimination device 10 or a storage device outside the discrimination device 10) designated by the user of the discrimination device 10.

[0035] The discrimination unit 112 discriminates whether or not young panicle differentiation has occurred in the target crop by inputting the sensing result data acquired by the acquisition unit 111 into the trained model 121. By inputting the sensing result data into the trained model 121, data indicating whether or not young panicle differentiation has occurred in the target crop is output from the trained model 121. The discrimination unit 112 may supply the data output from the trained model 121 directly to the output unit 113, or may discriminate whether or not young panicle differentiation has occurred based on the data output from the trained model 121 and supply the discrimination result to the output unit 113.

[0036] The output unit 113 outputs the discrimination result by the discrimination unit 112. As an example, the output unit 113 outputs data indicating the discrimination result to an output device connected to the input / output unit 14. Here, examples of the output device include, but are not limited to, a display, a printer, a projector, or a speaker. The output unit 113 may output the data indicating the discrimination result by writing the data to a storage destination designated by the user of the discrimination device 10 (which may be a storage device within the discrimination device 10 or a storage device external to the discrimination device 10). The output unit 113 may output the data by transmitting the data to another device connected via the communication unit 13. More specifically, the output unit 113 may transmit the data to, for example, a user terminal (such as a smartphone) carried by a farm manager or the like.

[0037] The training data acquisition unit 114 acquires training data including first data indicating the results of sensing internodes of a crop that are exposed by removing the leaf sheaths of the crop collected in the field, and second data indicating whether or not young panicle differentiation has occurred in the crop. The training data acquisition unit 114 may receive the data from another device connected via the communication unit 13, or may acquire the data input via the input / output unit 14. The training data acquisition unit 114 may also acquire the data by reading the data from a storage location specified by the user of the discrimination device 10 (which may be a storage device within the discrimination device 10 or a storage device external to the discrimination device 10).

[0038] The training unit 115 uses training data including the first data and the second data to train the trained model 121 by machine learning. In other words, the training unit 115 uses the first data as input data and the second data as output data, associates the two, and trains the trained model 121 by machine learning. More specifically, as an example, the training unit 115 updates the model parameters of the trained model 121 so as to reduce the loss calculated from the predicted value obtained from the input data and the output data.

[0039] (Sensing device) The sensing device 20 senses the crop to be discriminated and outputs the sensing results to the discrimination device 10. If the sensing device 20 is a spectroreflectometer, the sensing device 20 outputs data indicating the measurement results of the spectral reflectance. As the spectroreflectometer, a spectroreflectometer generally available on the market can be used. Alternatively, a dedicated device for measuring the spectral reflectance of internodes of crops may be used as the spectroreflectometer.

[0040] FIG. 6 is a diagram showing an example of data indicating the sensing results obtained by the sensing device 20. In the example of FIG. 6, the data indicating the sensing results is data representing a spectral reflectance curve. In FIG. 6, the horizontal axis indicates wavelength (nm) and the vertical axis indicates reflectance. Curve 611 shows the measurement results of the spectral reflectance of an internode before panicle differentiation. Curve 612 shows the measurement results of the spectral reflectance of an internode after panicle differentiation. As shown in FIG. 6, the spectral reflectance curves before and after panicle differentiation have different characteristics.

[0041] <Flow of the determination method> Fig. 7 is a flow chart showing an example of the flow of a method for determining whether or not young panicle differentiation has occurred according to this embodiment. Fig. 8 is a diagram for simply explaining the determination method shown in Fig. 7. In the examples of Figs. 7 and 8, a case will be described in which the crop to be determined is paddy rice. First, in step S11, a producer or the like pulls out and harvests paddy rice from a paddy field, which is a farm field. Paddy rice 81 in Fig. 8 is an example of paddy rice harvested in step S11.

[0042] In step S12 of Figure 7, the producer or the like peels and removes the leaves (leaf sheaths) around the stems of the harvested rice plants. This exposes the internodes of the rice plants. No special equipment or the like is required for this step. Internode 82 in Figure 8 is an example of an internode obtained in step S12.

[0043] In step S13 of FIG. 7, the producer or the like senses the internodes of the rice plant using the sensing device 20. If the sensing device 20 is a spectroreflectometer, the producer or the like performs an operation to measure the spectral reflectance of the internodes of the rice plant using the spectroreflectometer. At this time, the producer or the like may perform the measurement with the background set to a predetermined color (black, etc.). The spectroreflectometer outputs data indicating the measurement results to the discrimination device 10. Also, if the sensing device 20 is an imaging device, the producer or the like performs an operation on the imaging device or the sensing device 20 to capture an image of the internodes of the rice plant. At this time, the producer or the like may perform the image capture with the background set to a predetermined color (black, etc.). The imaging device outputs image data of the captured image to the discrimination device 10.

[0044] The acquisition unit 111 of the discrimination device 10 acquires data indicating the sensing result. The acquisition unit 111 may use the data acquired from the sensing device 20 as sensing result data to be input to the trained model 121, or may generate sensing result data using the data acquired from the sensing device 20. The spectral reflectance curve 83 in FIG. 8 is an example of sensing result data acquired in step S13.

[0045] When the sensing device 20 is an imaging device, the acquisition unit 111 may, for example, perform predetermined image processing on image data acquired from the imaging device, and use the data obtained as the sensing result data. More specifically, the acquisition unit 111 may, for example, perform processing to convert the background color of the image data to a predetermined color (black, etc.).

[0046] In step S14, the discrimination unit 112 discriminates whether or not young panicle differentiation has occurred in the target crop by inputting the acquired sensing result data into the trained model 121. Data indicating the discrimination result of whether or not young panicle differentiation has occurred is output from the trained model 121. In step S15, the output unit 113 outputs the discrimination result from step S14.

[0047] <Flow of how to generate a trained model> FIG. 9 is a flow diagram showing the flow of a method for generating the trained model 121. FIG. 10 is a diagram briefly explaining the method for generating the trained model 121 shown in FIG. 9. In step S21 of FIG. 9, the training data acquisition unit 114 acquires training data including first data indicating the results of sensing the internodes of a crop harvested in a field that are exposed by removing the leaf sheaths of the crop, and second data indicating whether or not the crop has differentiated into young panicles. The spectral reflectance curve 91 of FIG. 10 is an example of the first data. The first data may be, for example, data indicating the results of sensing the internodes with the background colored a predetermined color.

[0048] The second data is, for example, data indicating the results of an expert or the like observing young rice panicles using the method shown in Fig. 2. The second data may be, for example, data input by an expert or the like who has observed the young rice panicles using an input device (keyboard, mouse, etc.) connected to the input / output unit 14 of the discrimination device 10, or may be data received from another device (for example, a user terminal held by the expert) connected via the communication unit 13.

[0049] FIG. 11 is a diagram showing a specific example of the first data included in the training data. In the example of FIG. 11, images 401 to 403 are images of internodes taken before panicle differentiation. Images 411 to 413 are images of internodes taken after panicle differentiation. The image data included in the training data may be, for example, image data taken against a uniform background of a predetermined color (such as black). In this case, by using image data of internodes of a crop to be identified against a background similar to that of the training data, the estimation accuracy of the trained model 121 can be improved with a small amount of training data.

[0050] In step S22 of FIG. 9, the training unit 115 uses the first data as input data and the second data as output data, associates the two, and trains the trained model 121 by machine learning.

[0051] <Effects of the embodiment> Conventional methods for determining the panicle differentiation stage require sampling rice plants, then dissecting them under a microscope using a scalpel or tweezers to expose and observe the young panicles. In contrast, with the discrimination system 1 according to the present embodiment, after sampling rice plants, the leaf sheaths are simply peeled off until the green rings at the internodes (lower stems) are visible, eliminating the need to disassemble the rice plants until the young panicles are exposed. Therefore, according to this embodiment, it is possible to easily and accurately determine whether or not a crop has undergone panicle differentiation. Furthermore, according to this embodiment, changes in the green rings can be quantitatively evaluated from changes in spectral reflectance, allowing for pinpoint determination of the panicle differentiation stage. This allows for accurate and rapid evaluation of the panicle differentiation stage, for example, in rice production sites and in rice research. Furthermore, this can be used to facilitate the development of cultivation plans, such as fertilization and water management.

[0052] <Modification> The functions of the above-described discrimination device 10 may be shared and implemented by a plurality of devices. For example, the above-described discrimination device 10 may be realized as a system in which two or more devices are connected via a communication network. In this case, the discrimination system may be, for example, a system including a first device including an acquisition unit 111, a discrimination unit 112, and an output unit 113, and a second device including a training data acquisition unit 114 and a training unit 115. In this case, the functions of the discrimination device 10 are realized by the cooperation of the first device and the second device.

[0053] Furthermore, in the above-described embodiment, the case where the trained model 121 is stored in the memory unit 12 of the discrimination device 10 has been described, but the trained model 121 may be stored in a device other than the discrimination device 10. In this case, the discrimination device 10 transmits sensing result data to the device in which the trained model 121 is stored, and receives data indicating whether panicle differentiation has occurred or not, transmitted from the device in response to the transmitted sensing result data.

[0054] [Embodiment 2] Another embodiment of the present invention will be described below. Note that the same reference numerals are used to designate components having the same functions as those described in the above embodiment, and the description thereof will not be repeated.

[0055] The discrimination system 2 according to this embodiment discriminates not only whether or not a crop has differentiated into young panicles, but also the detailed developmental stage (FS: floral stage) after panicle differentiation. Here, an example of the developmental stage after panicle differentiation of paddy rice, which is an example of a crop, will be described with reference to the drawings. FIG. 12 is a diagram showing an example of the developmental stage after panicle differentiation of paddy rice. In the example of FIG. 12, the period from the panicle differentiation stage to the panicle formation stage is divided into a number of developmental stages ("0", "0.5", "1", "1.5", "2", "2.5", and "3").

[0056] However, the method for dividing the developmental stages is not limited to the example in Figure 12, and the developmental stages may be divided by other methods. For example, the undifferentiated stage may be defined as "0," and the period from the panicle differentiation stage to the panicle formation stage may be divided into seven developmental stages: "0.5," "1," "1.5," "2," "2.5," "3," and "3.5."

[0057] 13 is a block diagram showing the configuration of the discrimination system 2. The discrimination system 2 differs from the above-described discrimination system 1 in that the discrimination system 2 includes a discrimination unit 212, a training data acquisition unit 214, a training unit 215, and a trained model 221 instead of the discrimination unit 112, the training data acquisition unit 114, the training unit 115, and the trained model 121 of the discrimination system 1.

[0058] The trained model 221 is a trained model that uses data indicating the results of sensing the internodes of a crop that are exposed by removing the leaf sheaths of the crop as input data, and the presence or absence of panicle differentiation of the crop and the developmental stage after panicle differentiation as output data, and has machine-learned the correlation between the input data and the output data. As an example, the trained model 221 may be an SVM, or may be a model having a neural network structure such as a CNN. However, the trained model 221 is not limited to the above examples, and may be another model generated by machine learning.

[0059] The input of the trained model 221 is sensing result data indicating the results of sensing the internodes of the target crop exposed by removing the leaf sheaths of the target crop collected in the field, and is the same as the input of the trained model 121 according to the above-mentioned embodiment 1. On the other hand, the output of the trained model 221 is data indicating whether or not young panicle differentiation has occurred in the target crop and the developmental stage after young panicle differentiation. As an example, the output of the trained model 121 is a real value indicating whether or not young panicle differentiation has occurred and the developmental stage after young panicle differentiation. Furthermore, as an example, the output of the trained model 221 may include data indicating the probability of each of a plurality of developmental stages being in that developmental stage.

[0060] The discrimination unit 212 discriminates whether or not young panicle differentiation has occurred in the target crop and the developmental stage after young panicle differentiation by inputting the sensing result data acquired by the acquisition unit 111 into the trained model 221. By inputting the sensing result data into the trained model 221, data indicating whether or not young panicle differentiation has occurred in the target crop and the developmental stage after young panicle differentiation is output from the trained model 221. The discrimination unit 112 may supply the data output from the trained model 121 directly to the output unit 113, or may discriminate whether or not young panicle differentiation has occurred and the developmental stage based on the data output from the trained model 121 and supply the discrimination result to the output unit 113.

[0061] The training data acquisition unit 214 acquires training data including first data indicating the results of sensing the internodes of a crop that are exposed by removing the leaf sheaths of the crop collected in the field, and third data indicating whether or not the crop has differentiated into young panicles and the developmental stage after differentiation of the young panicles.

[0062] The training unit 215 uses training data including the first data and the third data to train the trained model 221. In other words, the training unit 115 can also train the trained model 221 by machine learning, associating the first data as input data and the third data as output data. More specifically, as an example, the training unit 215 updates the model parameters of the trained model 221 so as to reduce the loss calculated from the predicted value obtained from the input data and the output data.

[0063] <Flow of the determination method> Fig. 14 is a flow chart showing an example of the flow of a method for determining whether or not young panicle differentiation has occurred and the developmental stage after young panicle differentiation according to this embodiment. The determination method shown in Fig. 14 differs from the determination method shown in Fig. 7 described above in that it includes step S34 instead of step S14. In step S34, the discrimination unit 212 determines whether or not young panicle differentiation has occurred and the developmental stage of the target crop by inputting the acquired sensing result data into the trained model 221. Data indicating the discrimination results of whether or not young panicle differentiation has occurred and the developmental stage are output from the trained model 221.

[0064] <Flow of how to generate a trained model> Fig. 15 is a flow diagram showing the flow of a method for generating the trained model 221. In step S41 of Fig. 15, the training data acquisition unit 214 acquires training data including first data indicating the results of sensing the internodes of a crop harvested in a field that are exposed by removing the leaf sheaths of the crop, and third data indicating the presence or absence of panicle differentiation and the developmental stage of the crop.

[0065] The third data is, for example, data indicating the results of an expert's observation of the presence or absence of young panicle differentiation and the developmental stage after panicle differentiation in paddy rice plants using the method shown in FIG. 2. As an example, the third data is data in which undifferentiated rice is designated as "0," and the period from the young panicle differentiation stage to the young panicle formation stage is divided into seven developmental stages: "0.5," "1," "1.5," "2," "2.5," "3," and "3.5." The third data takes any of the values ​​"0," "0.5," "1," "1.5," "2," "2.5," "3," and "3.5." For example, the third data may be data input by an expert or the like who observed the young panicles of paddy rice using an input device (keyboard, mouse, etc.) connected to the input / output unit 14 of the discrimination device 10, or may be data received from another device (e.g., a user terminal owned by the expert) connected via the communication unit 13.

[0066] In step S42, the training unit 215 uses the first data as input data and the third data as output data, associates the two, and trains the trained model 121 through machine learning.

[0067] Fig. 16 is a diagram showing an example of the discrimination results obtained by the discrimination system 2. In Fig. 16, graphs 501 to 503 are graphs showing the discrimination results for the rice varieties Koshihikari, Hitomebore, and Nikomaru, respectively. Graph 504 is a graph showing the discrimination results for these three varieties together. In graphs 501 to 504, the horizontal axis shows the estimated value of the developmental stage, which is the discrimination result obtained by the discrimination system 2, and the vertical axis shows the observed value of the developmental stage. As is clear from graphs 501 to 504, there is a correlation between the estimated value and the observed value of the developmental stage for multiple rice varieties.

[0068] [Software implementation example] The functions of the discrimination device 10 (hereinafter referred to as the "device") can be realized by a program that causes a computer to function as the device, and a program that causes a computer to function as each control block of the device (particularly each part included in the control unit 11).

[0069] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the program, thereby realizing the functions described in each of the above embodiments.

[0070] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0071] Furthermore, some or all of the functions of the control blocks can be realized by logic circuits. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of the control blocks can also be realized by, for example, a quantum computer.

[0072] Furthermore, each process described in each of the above embodiments may be executed by AI (Artificial Intelligence). In this case, the AI ​​may run on the control device or on another device (for example, an edge computer or a cloud server).

[0073] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0074] <Summary> The discrimination method according to aspect 1 of the present invention includes an acquisition step of acquiring sensing result data indicating the results of sensing the internodes of a target crop that are exposed by removing the leaf sheaths of the target crop collected in a field; a discrimination step of discriminating whether or not young panicle differentiation is present in the target crop by inputting the sensing result data acquired in the acquisition step into a trained model that has machine-learned the correlation between the input data and the output data, with the data indicating the results of sensing the internodes of the target crop that are exposed by removing the leaf sheaths of the crop as input data and the presence or absence of young panicle differentiation in the target crop as output data; and an output step of outputting the discrimination result in the discrimination step.

[0075] According to the above aspect, it is possible to easily and accurately determine whether or not young panicle differentiation has occurred in a crop.

[0076] A discrimination method according to a second aspect of the present invention is the discrimination method according to the first aspect, wherein the sensing result data includes data indicating the measurement results of the spectral reflectance of the internodes of the target crop measured by a spectroscopic reflectometer.

[0077] According to the above aspect, the presence or absence of panicle differentiation of a crop can be accurately determined simply by performing the simple task of removing the leaf sheath of the crop and measuring the internodes of the crop with a spectroreflectometer.

[0078] A discrimination method according to a third aspect of the present invention is the discrimination method according to the first or second aspect, wherein the sensing result data includes image data obtained by capturing an image of an internode of the target crop.

[0079] According to the above aspect, the presence or absence of panicle differentiation of a crop can be accurately determined simply by performing the simple task of removing the leaf sheath of the crop and photographing the internodes of the crop.

[0080] A discrimination method according to a fourth aspect of the present invention is the discrimination method according to any one of the above-mentioned first to third aspects, wherein the target crop is paddy rice. According to the above-mentioned aspect, it is possible to easily and accurately determine whether or not young panicle differentiation has occurred in paddy rice.

[0081] A discrimination method according to a fifth aspect of the present invention is the discrimination method according to the third aspect, wherein the image data represents an image of an internode of the target crop against a predetermined color background. According to the above aspect, it is possible to more accurately discriminate whether or not young panicle differentiation has occurred in paddy rice.

[0082] The machine learning method according to aspect 6 of the present invention includes an acquisition step of acquiring first data indicating the results of sensing the internodes of a crop that are exposed by removing the leaf sheaths of the crop collected in a field, and second data indicating whether or not young panicle differentiation has occurred in the crop, and a training step of using the first data as input data and the second data as output data, associating the two, and machine learning a learned model.

[0083] According to the above aspect, it is possible to generate a trained model that can easily and accurately determine whether or not young panicles have differentiated in a crop.

[0084] A discrimination device according to aspect 7 of the present invention comprises an acquisition unit that acquires sensing result data indicating the results of sensing the internodes of a target crop that are exposed by removing the leaf sheaths of the target crop collected in a field; a discrimination unit that uses the data indicating the results of sensing the internodes of the target crop that are exposed by removing the leaf sheaths as input data and the presence or absence of young panicle differentiation of the target crop as output data, and inputs the sensing result data acquired by the acquisition unit into a trained model that has machine-learned the correlation between the input data and the output data; and an output unit that outputs the discrimination result obtained by the discrimination unit.

[0085] According to the above aspect, it is possible to easily and accurately determine whether or not young panicle differentiation has occurred in a crop.

[0086] A program according to aspect 8 of the present invention is a program for causing a computer to function as the discrimination device of aspect 7, and is a program for causing a computer to function as the acquisition unit, the discrimination unit, and the output unit.

[0087] According to the above aspect, it is possible to easily and accurately determine whether or not young panicle differentiation has occurred in a crop.

[0088] A machine learning device according to aspect 9 of the present invention includes an acquisition unit that acquires first data indicating the results of sensing the internodes of a crop that are exposed by removing the leaf sheaths of the crop collected in a field, and second data indicating whether or not young panicle differentiation has occurred in the crop, and a training unit that uses the first data as input data and the second data as output data, associates the two, and machine-learns a learned model.

[0089] According to the above aspect, it is possible to generate a trained model that can easily and accurately determine whether or not young panicles have differentiated in a crop.

[0090] A program according to aspect 10 of the present invention is a program for causing a computer to function as the machine learning device described in aspect 9 above, and is a program for causing a computer to function as the acquisition unit and the training unit.

[0091] According to the above aspect, it is possible to generate a trained model that can easily and accurately determine whether or not young panicles have differentiated in a crop.

[0092] A discrimination method according to aspect 11 of the present invention is a discrimination method according to any one of aspects 1 to 5 above, wherein the output data includes information indicating whether or not young panicle differentiation has occurred in the crop and the developmental stage after young panicle differentiation, and the discrimination process determines whether or not young panicle differentiation has occurred in the target crop and the developmental stage after young panicle differentiation by inputting the sensing result data acquired in the acquisition process into the trained model.

[0093] According to the above aspect, it is possible to easily and accurately determine whether or not young panicle differentiation has occurred in a crop and the developmental stage thereof. [Explanation of symbols]

[0094] 1, 2 Discrimination system 10 Discrimination device 11 Control section 12 Storage section 13 Communications Department 14 Input / output section 20 Sensing Device 111 Acquisition Department 112, 212 Discrimination section 113 Output section 114, 214 Training data acquisition section 115, 215 Training Department 121, 221 trained models

Claims

1. an acquiring step of acquiring sensing result data indicating the results of sensing internodes of a target crop that are exposed by removing leaf sheaths of the target crop collected in a field; a discrimination process for discriminating whether or not young panicle differentiation of the target crop has occurred by inputting the sensing result data acquired in the acquisition process into a trained model that has machine-learned the correlation between the input data and the output data, with data indicating the results of sensing the internodes of the crop that are exposed by removing the leaf sheaths of the crop as input data and the presence or absence of young panicle differentiation of the crop as output data; an output step of outputting a determination result in the determination step; A method of determining

2. The sensing result data includes data indicating measurement results of the spectral reflectance of internodes of the target crop measured by a spectroscopic reflectometer. The method of claim 1 .

3. The sensing result data includes image data obtained by imaging an internode of the target crop. The method according to claim 1 or 2.

4. The target crop is paddy rice. The method according to claim 1 or 2.

5. The image data is data representing an image of an internode of the target crop, the image having a background of a predetermined color. The method according to claim 3 .

6. an acquiring step of acquiring first data indicating the results of sensing internodes of a crop that are exposed by removing leaf sheaths of the crop collected in a field, and second data indicating whether or not young panicle differentiation has occurred in the crop; a training process in which the first data is used as input data, the second data is used as output data, and the two are associated with each other to machine-train a learned model; Machine learning methods, including

7. an acquisition unit that acquires sensing result data indicating the results of sensing internodes of a target crop that are exposed by removing leaf sheaths of the target crop collected in a field; a discrimination unit that determines whether or not young panicle differentiation has occurred in the target crop by inputting the sensing result data acquired by the acquisition unit into a trained model that has machine-learned the correlation between the input data and the output data, with input data indicating the results of sensing the internodes of the crop that are exposed by removing the leaf sheaths of the crop and output data indicating whether or not young panicle differentiation has occurred in the target crop; an output unit that outputs a determination result by the determination unit; A discrimination device comprising:

8. A program for causing a computer to function as the discrimination device according to claim 7, the program causing a computer to function as the acquisition unit, the discrimination unit, and the output unit.

9. an acquisition unit that acquires first data indicating the results of sensing internodes of a crop that are exposed by removing leaf sheaths of the crop collected in a field, and second data indicating whether or not young panicle differentiation has occurred in the crop; a training unit that uses the first data as input data and the second data as output data, associates the two, and performs machine learning to generate a learned model; A machine learning device comprising:

10. 10. A program for causing a computer to function as the machine learning device according to claim 9, the program causing a computer to function as the acquisition unit and the training unit.

11. The output data includes information indicating whether or not panicle differentiation of the crop has occurred and the developmental stage after panicle differentiation, The discrimination step determines whether or not panicle differentiation of the target crop has occurred and the developmental stage after panicle differentiation by inputting the sensing result data acquired in the acquisition step into the trained model. The method according to claim 1 or 2.

Images