Methods for determining the activity status of plants

Abstract

To provide a method for determining the active state of plants which enables accurate determination without the need for expertise or skill.SOLUTION: A method for determining the active state of a plant includes a measurement step (S1) for measuring the surface temperature of a plant and its surrounding temperature; a calculation step (S2) for using the measured surface temperature and surrounding temperature to calculate an estimated value by a statistical or machine learning process; and a relation creating step (S3) for, based on the calculated estimated value, creating a relational expression or diagram indicating the correspondence between the surface temperature and the surrounding temperature, and the active state, used to determine the active state of the plant.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present disclosure relates to a method for determining the activity state of plants.

Background Art

[0002] Conventionally, in a farm, workers in the farm have been determining whether the cultivated trees are diseased. If diseased, management such as applying appropriate treatment is required. The disease diagnosis for distinguishing diseased trees (infected trees) from healthy trees (non-infected trees) has been based on the experience and skill of the workers. Infection is, for example, infection with white root disease (WRD).

[0003] On the other hand, for example, Patent Document 1 discloses a method for determining the activity state of Paragonimus trees including healthy, diseased, etc. based on the measured surface temperature and the ambient temperature.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] The technique of Patent Document 1 enables a system to determine the activity state of Paragonimus trees based on data rather than by humans. That is, the technique of Patent Document 1 makes it possible to determine the activity state of Paragonimus trees without requiring experience and skill. Here, comparison with a threshold value may be performed in determining the activity state, and it is possible to further improve the determination accuracy by appropriately setting the threshold value.

[0006] In view of such circumstances, an object of the present disclosure is to provide a method for determining the activity state of plants that can accurately determine the activity state of plants without requiring experience and skill. [Means for solving the problem]

[0007] A method for determining the activity state of a plant according to one embodiment of the present disclosure includes: a measurement step of measuring the surface temperature of the plant and the ambient temperature of the plant; a calculation step of calculating a predicted value using the measured surface temperature and ambient temperature by a statistical method or a machine learning method; and a relationship generation step of generating a relationship expression or relationship diagram showing the correspondence between the surface temperature, ambient temperature and the activity state, which is used to determine the activity state of the plant based on the calculated predicted value. This configuration makes it possible to accurately determine the activity level of plants without requiring experience or expertise.

[0008] In one embodiment of this disclosure, the statistical method is logistic regression. This configuration can improve the accuracy of determining the activity status of plants.

[0009] In one embodiment of the present disclosure, in the calculation step, p is the probability of disease, n is an integer of 2 or more, and a i (i=1, 2, ..., n) are the regression coefficients, b is a constant, x i A logistic regression is performed using the following formula, with (i=1, 2, ..., n) as variables including the surface temperature and the ambient temperature.

number

[0010] In one embodiment of the present disclosure, the calculation step includes calculating a criterion for the predicted value using an ROC curve. This configuration provides a highly accurate criterion for determining the activity level of plants.

[0011] As one embodiment of the present disclosure, the relationship generation step generates the relationship diagram showing the correspondence between the surface temperature and the ambient temperature and the activity state of the plant including at least diseased and healthy based on the calculated predicted value and the reference. With this configuration, it becomes possible to quickly perform highly accurate discrimination without requiring experience or skill.

[0012] As one embodiment of the present disclosure, the plant is Ficus elastica. With this configuration, it is possible to accurately discriminate whether Ficus elastica is diseased or not, and improve the productivity of the Ficus elastica farm.

Advantages of the Invention

[0013] According to the present disclosure, it is possible to provide a method for discriminating the activity state of a plant that can accurately discriminate the activity state of a plant without requiring experience or skill.

Brief Description of the Drawings

[0014] [Figure 1] FIG. 1 is a schematic diagram showing an activity state discrimination system. [Figure 2] FIG. 2 is a diagram showing an example of temperature data acquired by a temperature measuring device. [Figure 3] FIG. 3 is a flowchart showing an activity state discrimination method according to an embodiment of the present disclosure. [Figure 4] FIG. 4 is a diagram showing an example of a relationship diagram showing the correspondence between temperature and the activity state of a plant. [Figure 5] FIG. 5 is a diagram showing an example of a ROC curve when the statistical method is logistic regression. [[ID=3q]] [Figure 6] FIG. 6 is a diagram showing an example of a ROC curve when the machine learning method is a neural network.

Modes for Carrying Out the Invention

[0015] The following describes a method for determining the activity state of a plant according to an embodiment of the present disclosure with reference to the drawings. In each figure, the same or corresponding parts are denoted by the same reference numerals. In the description of this embodiment, the description of the same or corresponding parts will be omitted or simplified as appropriate.

[0016] FIG. 1 is a schematic diagram showing an activity state determination system 1 that executes a method for determining the activity state of a plant according to this embodiment (hereinafter, may be simply referred to as the "activity state determination method"). The activity state determination system 1 targets various plants, and the type of the target plant is not limited. In this embodiment, Paraserianthes falcataria will be described as an example. The activity state determination method according to this embodiment is executed to determine whether Paraserianthes falcataria is diseased from, for example, the surface temperature of the trunk of Paraserianthes falcataria. The productivity of the Paraserianthes falcataria farm can be improved by the highly accurate determination described below.

[0017] The activity state determination system 1 that executes the activity state determination method according to this embodiment includes a plate Br, a temperature measurement device 10, and an arithmetic device 20. In the activity state determination system 1, the temperature measurement device 10 measures the surface temperature of the trunk of the tree Tr and the ambient temperature of the plate Br attached around the trunk of the tree Tr. In this embodiment, the tree Tr to be measured is Paraserianthes falcataria. The arithmetic device 20 executes an arithmetic operation for determining the activity state of the tree Tr based on the measured surface temperature and ambient temperature.

[0018] The temperature measurement device 10 and the arithmetic device 20 can transmit and receive data to and from each other. The method of data transmission and reception is not limited. For example, the temperature measurement device 10 and the arithmetic device 20 may transmit and receive data via a network such as the Internet. The temperature measurement device 10 and the arithmetic device 20 may perform wired communication or wireless communication. Also, for example, the temperature measurement device 10 and the arithmetic device 20 may be provided with a holder for removably storing a storage medium. The data stored in the storage medium stored in the temperature measurement device 10 and removed and stored in the arithmetic device 20 may be transferred. The storage medium is, for example, a memory card.

[0019] Board Br is a thin wooden board. In this embodiment, the shape of board Br is square, but it is not limited to any particular shape. The shape of board Br may be, for example, rectangular. Also, in this embodiment, the length of one side of board Br is about 50 cm, but it is not limited to any particular length. The length of one side of board Br may be determined according to the width of the trunk of tree Tr, so as to be sufficiently larger than the width of the trunk of tree Tr. The length of one side of board Br may be set to, for example, 3 to 9 times the width of the trunk of tree Tr.

[0020] Here, a sufficiently high emissivity can be obtained by using wood for the plate Br. Emissivity is the ratio of the amount of energy radiated from an object by thermal radiation to the amount of energy radiated by blackbody radiation. As an example, the emissivity of both the plate Br and the wood Tr can be made approximately 97%.

[0021] A wooden frame may be provided around the periphery of the board Br. By providing a frame, the shape of the board Br can be maintained even when it is touched or exposed to wind. The width and thickness of the frame may be approximately 2 cm, but are not limited to this.

[0022] The temperature measuring device 10 may be, for example, an infrared (IR) camera that captures an image based on infrared radiation arriving from an object contained within the field of view Vf. The image captured by the infrared camera shows the temperature measured based on the spectrum of infrared radiation arriving at each pixel, i.e., the temperature distribution of the object. The temperature measuring device 10 is oriented so that the field of view Vf includes the board Br and the trunk of the tree Tr. Preferably, the temperature measuring device 10 is positioned so that the entire surface of the board Br, which is placed behind the trunk of the tree Tr, is represented in most of the field of view Vf. Most of the field of view Vf is, for example, more than 1 / 4 of the horizontal width or more than 1 / 3 of the vertical height of the captured image. The temperature measuring device 10 outputs temperature data showing the temperature distribution of the tree Tr trunk and the board Br to the calculation device 20.

[0023] Here, the temperature measuring device 10 measures the temperature of the board Br in addition to the temperature of the tree trunk Tr, so that the measured temperatures can be used as a reference for comparison. By using an infrared camera as the temperature measuring device 10, temperature data for multiple areas can be easily acquired at once without contacting the object being measured. The temperature measuring device 10 then performs the measurement process described later.

[0024] The computing device 20 performs calculations to predict and determine the activity state of the tree Tr. The activity state can be, for example, "healthy," "sick," "suspected to be sick," or "dead." In summary, the computing device 20 uses the measured surface temperature of the plant (Rumex pallidum in this embodiment) and the ambient temperature to calculate predicted values ​​using statistical or machine learning methods, and generates a relational expression or diagram showing the correspondence between the plant's surface temperature, ambient temperature, and the plant's activity state based on the predicted values. Statistical methods include, for example, logistic regression. Machine learning methods include, for example, neural networks, support vector machines, and random forests. The predicted value is the probability of a certain activity state (for example, sickness) (for example, the probability of being sick). The relational expression or diagram is used to determine the activity state of a target plant. The computing device 20 performs the calculation process and relation generation process described later. If the activity state determination system 1 also continues to perform the determination process described later, the computing device 20 may perform the determination process together with the temperature measuring device 10.

[0025] Here, the arithmetic unit 20 may be implemented as a computer. In this case, a program for executing the activity state determination method described later may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be loaded into the computer system and the processing executed. Here, the computer system may consist of an OS (Operating System) and peripheral devices, etc. Furthermore, a computer-readable recording medium refers to a portable medium such as a flexible disk, magneto-optical disk, ROM (Read-only Memory), CD (Compact Disc)-ROM, or a storage device such as a hard disk built into the computer system.

[0026] Figure 2 shows an example of temperature data acquired by the temperature measuring device 10. The surface temperature of the object is represented by varying shades of gray. Areas with higher temperatures are represented brighter, and areas with lower temperatures are represented darker. The bright rectangle in the center of Figure 2 shows the temperature distribution at the periphery of the plate Br. The darker area enclosed by the periphery shows the temperature distribution on the surface of the plate Br. The brighter area that intersects the rectangle vertically at approximately the center of Figure 2 shows the temperature distribution on the surface of the tree trunk Tr. In the example in Figure 2, the surface of the tree trunk Tr is represented brighter than the surface of the plate Br, indicating a higher temperature.

[0027] When the roots of a tree trabeculata are infected with the WRD pathogen, the absorption of water by the roots and its transport to the leaves are inhibited. At this time, the stomata of the tree trabeculata leaves close, and water dissipation from the leaves is suppressed. In other words, because transpiration from the tree trabeculata is suppressed, the heat of vaporization is reduced, and the temperature of the tree trabeculata trunk becomes higher than that of a healthy tree. The fact that the trunk temperature of a diseased tree trabeculata rises can be used to determine whether the tree trabeculata is diseased. Furthermore, this tendency is not limited to rubber parasol trees but is generally observed in plants. Here, the temperature of the leaves of the tree trabeculata fluctuates with the ambient temperature around the tree trabeculata, which also affects water dissipation. Therefore, as described above, the temperature measuring device 10 measures the temperature of the board Br in addition to the temperature of the tree trabeculata, so that the measured temperatures can be used as a comparison point.

[0028] (Method for determining activity status) Figure 3 is a flowchart showing the activity state determination method according to this embodiment.

[0029] (Measurement process) The temperature measuring device 10 measures the temperature distribution of the trunk and board Br of the tree Tr as subjects included within the field of view Vf (step S1, measurement process). Here, the temperature measuring device 10 may perform multiple measurements for each individual tree Tr. Regarding the timing of the measurement, it is preferable to measure during the daytime when healthy trees in a healthy state are actively transpirating. Specifically, this is the time of day from 10:00 to 14:00 when the sun's altitude is higher than at other times. Regarding the season of measurement, it is preferable to measure during periods other than the deciduous period when transpiration does not occur or is inactive.

[0030] In this embodiment, the temperature measuring device 10 measures the temperature of the trunk (surface temperature) and the temperature of the board Br (ambient temperature) of multiple tree trees. Furthermore, for each of the multiple tree trees, the actual activity status is identified through observation, etc., and stored as performance data in a storage device (e.g., memory or hard disk) or storage medium accessible by the computing device 20. In this embodiment, the activity status is identified as whether the tree is diseased or healthy. Table 1 shows an example of performance data.

[0031] [Table 1]

[0032] In Table 1, the identification number is a unique number assigned to each of the multiple tree Trs. The actual activity status is indicated by 1 if the tree Tr is diseased and 0 if it is healthy.

[0033] (calculation process) The calculation unit 20 calculates predicted values ​​using actual data by statistical methods or machine learning methods (step S2, calculation step). In this embodiment, the calculation unit 20 calculates predicted values ​​by statistical methods. In this embodiment, a statistical method is used, and the statistical method is logistic regression. In this embodiment, the predicted value is the probability of disease. Specific examples will be described later, but the inventors compared and studied several statistical methods and machine learning methods and found that the accuracy of discriminating the activity state of plants was relatively high when logistic regression was used as the statistical method. In other words, by using logistic regression as the statistical method, the accuracy of discriminating the activity state of plants can be improved.

[0034] Logistic regression is a method for obtaining the probability of an objective variable occurring from several explanatory variables, and the following formula can be used.

[0035]

number

[0036] Here, p is the probability of becoming ill. In other words, p corresponds to the probability of the occurrence of the dependent variable, where becoming ill is "1" and being healthy is "0". Let n be an integer greater than or equal to 2, a i (i=1, 2, ..., n) are the regression coefficients. Also, b is a constant. i (i=1, 2, ..., n) is a variable that includes surface temperature and ambient temperature. In this embodiment, n is 2, and the variables are surface temperature and ambient temperature. However, as shown in the above formula, there may be three or more variables, for example, wind speed during measurement may be added. Alternatively, there may be only one variable, for example, the relative temperature difference between surface temperature and ambient temperature may be used.

[0037] The computing device 20 can determine the regression coefficients and constants in the above equation using actual data as shown in Table 1, and calculate predicted values. Here, the predicted values ​​are given as the probability of occurrence (probability of disease in this embodiment). Therefore, in order to make the final prediction of determining the activity state of the plant (determination of diseased or healthy in this embodiment), it is necessary to set a standard (threshold) to compare with the calculated predicted values. For example, if the probability of disease is 0.5, a standard of 0.51 will determine that it is a healthy tree, and a standard of 0.49 will determine that it is a diseased tree. Thus, the accuracy of determining the activity state of the plant is affected by how the standard is set. In this embodiment, the computing device 20 calculates the standard as described below, enabling the generation of a relational expression or relational diagram (see Figure 4) that correlates surface temperature, ambient temperature and activity state with high accuracy.

[0038] The calculation unit 20 calculates a criterion for the predicted value using an ROC (Receiver Operating Characteristic) curve in the calculation process. By using the ROC curve, a highly accurate criterion for determining the activity state of plants can be obtained. Figure 5 shows an example of an ROC curve when the statistical method is logistic regression. The vertical axis is the true positive rate (TPR) or sensitivity. The horizontal axis is the false positive rate (FPR) or "1-specificity". The calculation unit 20 selects a criterion to maximize the AUC (Area Under Curve).

[0039] (Relationship generation process) The computing device 20 generates a relational expression or diagram showing the correspondence between surface temperature, ambient temperature, and the activity state, based on the calculated predicted values, which is used to determine the activity state of the plant (step S3, relation generation step). The generated relational expression or diagram may be stored in a memory device accessible by the computing device 20. The computing device 20 may also display the relational diagram on a display device (for example, the display of a mobile terminal) located at a location where the determination step described later is performed, via a network or other means.

[0040] Figure 4 shows an example of a relationship diagram illustrating the correspondence between temperature and plant activity. In the relationship diagram in Figure 4, surface temperature is shown vertically, and ambient temperature is shown horizontally. At the intersection, the activity status (sick, suspected of being sick, or healthy) is shown, determined by the predicted value (p, which is the disease probability in the above formula) calculated based on the combination of surface temperature and ambient temperature, and the calculated reference value.

[0041] In the relationship generation process of this embodiment, the arithmetic unit 20 generates a relationship diagram showing the correspondence between surface temperature and ambient temperature and the activity state of the plant, including at least diseased and healthy states, based on the calculated predicted values ​​and criteria. By using such a relationship diagram, it becomes possible to quickly make highly accurate determinations without requiring experience or skill. For example, even an inexperienced worker can quickly and accurately determine the activity state by measuring the surface temperature and ambient temperature in the location where the rubber tree is growing and comparing these temperatures with the relationship diagram. Here, the arithmetic unit 20 may store the relationship diagram in a memory device in table format. The relationship formula may be a mathematical expression of the relationship diagram shown in Figure 4.

[0042] Here, in the relationship diagram of Figure 4, "suspected illness" is included as an activity state to be determined, but it may be omitted. In other words, a relationship diagram that shows only illness and health, allowing for clearer determination, may be generated. The computing unit 20 can calculate predicted values ​​using multiple statistical methods, multiple machine learning methods, or combinations including one or more statistical methods and machine learning methods, and a criterion can be set for each of the multiple methods using the ROC curve described above. In such cases, "suspected illness" may be used in cases where some methods determine illness, while other methods determine health.

[0043] (Discrimination process) After the calculation unit 20 generates a relational expression or relational diagram, the activity status of the target plant is determined using the generated relational expression or relational diagram. The target plant is a different individual plant from the plant from which the above performance data was obtained. If the activity status determination system 1 continues to perform the determination (YES in step S4), the temperature measuring device 10 and the calculation unit 20 perform the determination of the activity status of the target plant (step S5, determination process). Alternatively, the surface temperature of the target plant and the ambient temperature may be measured by the temperature measuring device 10 or another device, and the activity status may be determined by a human or another computer (e.g., a mobile terminal) using the generated relational expression or relational diagram. In other words, the determination process may be performed by a human or other person other than the activity status determination system 1. In such cases, the activity status determination system 1 does not continue to perform the determination (NO in step S4) and terminates the series of processes.

[0044] In the identification process, the surface temperature of the plant to be identified and the ambient temperature are measured by a temperature measuring device 10 or another device. Furthermore, the disease probability may be calculated based on the measured temperatures. Table 2 shows an example of temperature data and disease probability for the plant to be identified. The identification number, as in Table 1, is a unique number used to identify the tree Tr.

[0045] [Table 2]

[0046] In the discrimination process, the computing device 20, another computer, or a human being uses the generated relational expression or relational diagram to determine the activity state of the plant to be discriminated.

[0047] As described above, in the plant activity status determination method according to this embodiment, the activity status determination system 1 provides criteria for determining the plant's activity status based on actual data. Therefore, it becomes possible to determine the plant's activity status without requiring experience or skill. Furthermore, the above criteria are determined based on predicted values ​​calculated using statistical or machine learning methods with actual data, thereby improving the accuracy of the determination. Thus, the plant activity status determination method according to this embodiment makes it possible to determine the activity status of rubber trees without requiring experience or skill.

[0048] (Examples) The effects of this disclosure will be described in detail below based on the examples, but this disclosure is not limited to the contents of the examples.

[0049] As described in the above embodiment, the calculation device 20 calculated predicted values ​​using logistic regression and generated a relationship diagram showing the correspondence between surface temperature, ambient temperature, and activity status. As actual data, temperature data from a total of 252 rubber trees were used: 54 diseased trees and 198 healthy trees. The ROC curve shown in Figure 5 was obtained and the relationship diagram was generated. The surface temperature and ambient temperature of another 75 rubber trees were measured, and the activity status was determined (predicted) based on the generated relationship diagram. When using logistic regression, the positive predictive value was 42%, the negative predictive value was 92%, and the average accuracy was 67%, demonstrating that disease of rubber trees could be determined with higher accuracy than conventional techniques.

[0050] In another embodiment, the computing device 20 used a neural network to calculate predicted values ​​and generated a relationship diagram showing the correspondence between surface temperature, ambient temperature, and activity status. The same actual data as in the logistic regression case was used. The ROC curve shown in Figure 6 was obtained and the relationship diagram was generated. As in the logistic regression case, the surface temperature and ambient temperature of 75 rubber trees were measured, and the activity status was determined based on the generated relationship diagram. When using the neural network, the positive predictive value was 40%, the negative predictive value was 90%, and the average accuracy was 66%, achieving almost the same discrimination accuracy as in the logistic regression case.

[0051] Furthermore, the computing unit 20 calculated predicted values ​​using machine learning methods such as support vector machines, random forests, and gradient boosting. These machine learning methods also yielded average accuracy rates equivalent to or better than those obtained with neural networks. Therefore, it was found that the methods for calculating predicted values ​​in the calculation process are not limited to statistical methods, and machine learning methods are also effective. [Explanation of symbols]

[0052] 1. Activity Status Discrimination System 10 Temperature measuring device 20 Arithmetic unit Br board Tr tree

Claims

1. A measurement step for measuring the surface temperature of a plant and the ambient temperature of the plant, A calculation step of calculating a predicted value using a statistical method or a machine learning method with the measured surface temperature and ambient temperature, The process includes a relationship generation step of generating a relationship formula or relationship diagram that shows the correspondence between the surface temperature, the ambient temperature, and the activity state, which is used to determine the activity state of the plant based on the calculated predicted values, The calculation step includes calculating a standard for the predicted value using the ROC curve, The relationship generation step generates a relationship diagram showing the correspondence between the surface temperature and ambient temperature and the activity state of the plant, including at least diseased and healthy states, based on the calculated predicted values ​​and criteria. A method for determining the activity state of a plant, wherein the plant is a rubber tree, and the surface temperature and ambient temperature are in the range of 30°C to 45°C.

2. The method for determining the activity state of a plant according to claim 1, wherein the statistical method is logistic regression.

3. In the calculation step described above, p is the probability of disease, n is an integer of 2 or more, and a ｉ (i = 1, 2, ..., n) are regression coefficients, b is a constant, x ｉ The method for determining the activity state of a plant according to claim 2, wherein logistic regression is performed using the following formula, with (i = 1, 2, ..., n) being variables including the surface temperature and the ambient temperature. [Math 1]

Images