Method for predicting the optimal harvest time for corn

Abstract

To provide a method for predicting the optimum corn harvesting stage that can accurately predict the optimum harvesting stage by calculating a predicted silking date from growing corn image data obtained by imaging growing corn.SOLUTION: A computer executes: a data acquisition step 10 of acquiring growing corn image data by imaging growing corn; a determination step 20 of determining each growing tassel imaged in the growing corn image data into at least three stages, non-flowered, early flowering, and late flowering, using a flowering stage estimation model; a silking date prediction step 30 of calculating a predicted silking date using the number of non-flowered tassels of growing tassels determined as non-flowered, the number of early flowering tassels of growing tassels determined as early flowering, and the number of late flowering tassels of growing tassels determined as late flowering; and a harvesting stage prediction step 40 of predicting the optimum harvesting stage using the integrated temperature with the predicted silking date as the starting date.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a method for predicting the optimal harvest time of corn that predicts the optimal harvest time of corn. Background Art

[0002] Sweet corn, which is a high-profit crop, is also known as a crop whose quality decreases significantly after harvesting, and planned harvesting operations are desired. Currently, producers touch ears by hand immediately before harvesting and determine whether harvesting is possible based on the swelling of the ears. However, this method requires frequent trips to the field during the harvesting period, which hinders efficient cultivation. In addition, since it is unavoidable to determine the optimal harvest time based on only a small number of samples from a vast field, it is difficult to accurately determine the optimal harvest time and perform planned harvesting operations. Incidentally, Patent Document 1 proposes a method for accurately predicting the yield of corn at an early stage. Further, Patent Document 2 proposes an apparatus for estimating the harvest time when agricultural products have good taste quality, and Patent Document 3 proposes an information processing apparatus, a harvest time prediction program, and a harvest time prediction method that can predict the harvest time of cultivated crops with higher accuracy. Prior Art Documents Patent Documents

[0003] Patent Document 1 Japanese Unexamined Patent Publication No. 2022-64327 Patent Document 2 Japanese Unexamined Patent Publication No. 2016-19512 Patent Document 3 Japanese Unexamined Patent Publication No. 2013-42668 Summary of the Invention Problem to be Solved by the Invention

[0004] Patent Document 1 does not aim to predict the optimal harvest time of corn, and Patent Documents 2 and 3 are not suitable for predicting the optimal harvest time of corn.

[0005] Therefore, the present invention aims to provide a method for predicting the optimal harvest time for corn, which can accurately predict the optimal harvest time by calculating the predicted silk extraction date from image data of corn growing during its growth. [Means for solving the problem]

[0006] The method for predicting the optimal harvest time for corn according to claim 1 of the present invention is a method for predicting the optimal harvest time for corn using a flowering stage estimation model that has been machine-trained using training image data obtained by adding flowering stage information of the male inflorescence to training image data obtained by imaging the male inflorescence of corn, wherein the flowering stage estimation model is a color change model in the center of the male inflorescence, and the method comprises a data acquisition step 10 in which a computer acquires growing corn image data obtained by imaging the corn during growth, and the data acquired in the data acquisition step 10 A determination step 20 determines, using the flowering stage estimation model, that each growing corn tassel captured in the growing corn image data is in at least three stages: unflowered, early flowering, and late flowering. A silk extraction date prediction step 30 calculates a predicted silk extraction date using the number of unflowered tassels of the growing corn determined to be unflowered, the number of early flowering tassels of the growing corn determined to be in the early flowering stage, and the number of late flowering tassels of the growing corn determined to be in the late flowering stage. The predicted silk extraction date calculated in the silk extraction date prediction step 30 is used as the starting date. Calculated using actual temperature or predicted temperature. The method is characterized by performing a harvest timing prediction step 40, which predicts the optimal harvest time using the accumulated temperature. The present invention as described in claim 2 is a method for predicting the optimal harvest time for corn as described in claim 1, characterized in that the center of the tassel is yellow in the early flowering stage and the center of the tassel is brown in the late flowering stage. The present invention as described in claim 3 is a method for predicting the optimal harvest time of corn as described in claim 1, characterized in that, as the flowering stage estimation model, a flowering status model based on the ratio of flowered parts to unflowered parts for one of the male inflorescences is used together with the color change model, in the determination step 20 a determination is made to four or more stages, and in the silk extraction date prediction step 30 the predicted silk extraction date is calculated using the number of male inflorescences determined to be four or more stages. Claim 4 The present invention, as described in claim 1, is a method for predicting the optimal harvest time for corn, characterized in that in the optimal harvest time prediction step 40, the accumulated temperature is determined according to the variety of corn being grown. Claim 5 The present invention described herein is as follows: Claim 4 The method for predicting the optimal harvest time for corn according to any one of the above, characterized in that the growing corn image data is an aerial image of the growing corn. [Effects of the Invention]

[0007] According to the present invention's method for predicting the optimal harvest time for corn, the predicted silk extraction date can be calculated by using a color change model in the center of the male inflorescence, thereby enabling the prediction of the optimal harvest time. [Brief explanation of the drawing]

[0008] [Figure 1] Flowchart illustrating the data acquisition step and determination step in a method for predicting the optimal harvest time for corn according to one embodiment of the present invention. [Figure 2] Flowchart illustrating the silk extraction date prediction step in the method for predicting the optimal harvest time for the corn. [Figure 3] Flowchart illustrating the harvest timing prediction step and output step in the method for predicting the optimal harvest timing of the corn. [Figure 4] Explanatory photograph illustrating the predictions made by the color change model in this embodiment. [Figure 5] This figure shows a demonstration example of the method for predicting the optimal harvest time for corn. [Figure 6] A photograph illustrating the flowering stage model used as a flowering stage estimation model in the method for predicting the optimal harvest time for the corn. [Modes for carrying out the invention]

[0009] The first embodiment of the present invention provides a method for predicting the optimal harvest time for corn, in which the flowering stage estimation model is a color change model in the center of the tassel, and the computer acquires growing corn image data of corn in the process of growing; a determination step in which the computer uses the flowering stage estimation model to determine that each growing tassel captured in the growing corn image data acquired in the data acquisition step is in at least three stages: unflowered, early flowering, and late flowering; a silk extraction date prediction step calculates a predicted silk extraction date using the number of unflowered tassels for growing tassels determined to be unflowered, the number of early flowering tassels for growing tassels determined to be early flowering, and the number of late flowering tassels for growing tassels determined to be late flowering; and the predicted silk extraction date calculated in the silk extraction date prediction step is used as the starting date. Calculated using actual temperature or predicted temperature. This method involves performing a harvest timing prediction step, which uses accumulated temperature to predict the optimal harvest time. According to this embodiment, by using a color change model in the center of the male inflorescence, the predicted silk extraction date can be calculated, thus allowing for the prediction of the optimal harvest time.

[0010] A second embodiment of the present invention utilizes the fact that the center of the tassel is yellow in the early flowering stage and brown in the late flowering stage, in addition to the method for predicting the optimal harvest time of corn according to the first embodiment. According to this embodiment, the flowering stage can be accurately determined by focusing on the color change in the center of the tassel.

[0011] According to a third embodiment of the present invention, in the method for predicting the appropriate harvesting time of corn according to the first embodiment, as a flowering stage estimation model, together with a color tone change model, a flowering status model based on the ratio of the flowering part to the non-flowering part of one tassel is used, determination is performed in four or more stages in the determination step, and the predicted silking date is calculated using the respective numbers of tassels determined in four or more stages in the silking date prediction step. According to this embodiment, by using the flowering status model together with the color tone change model as the flowering stage estimation model, determination can be performed in four or more stages, and the predicted silking date can be calculated with higher accuracy.

[0012] According to the present invention 4th embodiment, in the method for predicting the appropriate harvesting time of corn according to the first embodiment, in the appropriate harvesting time prediction step, the integrated temperature is set according to the variety of growing corn. According to the present embodiment, by setting the integrated temperature according to the corn variety, the appropriate harvesting time can be predicted with higher accuracy.

[0013] According to the present invention Fifth embodiment, from the first embodiment 4th , in the method for predicting the appropriate harvesting time of corn according to any one of the embodiments, the image data of growing corn is obtained by imaging the growing corn from the air. According to this embodiment, by imaging the growing corn from the air, it is easy to capture an image of the central part of the tassel. [EXAMPLE]

[0014] Hereinafter, a method for predicting the appropriate harvesting time of corn according to one example of the present invention will be described. The method for predicting the appropriate harvesting time of corn according to the present example predicts the appropriate harvesting time of corn by using a flowering stage estimation model that has been machine-learned using teacher data obtained by adding tassel flowering stage information to learning image data obtained by imaging corn tassels, wherein a computer executes a data acquisition step, a determination step, a silking date prediction step, an appropriate harvesting time prediction step, and an output step. In this example, sweet corn is used as the corn.

[0015] Figure 1 is a flowchart illustrating the data acquisition step and the determination step in the method for predicting the optimal harvest time for corn according to this embodiment. In data acquisition step 10, growing corn image data is acquired by taking images of corn plants in the process of growing. The image data of growing corn was taken from above while the corn was growing. By taking images of growing corn from above in this way, it is easier to capture the center of the male inflorescence. Drones are suitable for aerial imaging, as they allow for uniform aerial photography of the entire field, enabling the acquisition of images of growing corn across the entire field. A drone flight speed of 1-4 m / s and an altitude of 3-7 m are ideal.

[0016] In the determination step 20, each growing corn tassel captured in the growing corn image data acquired in the data acquisition step 10 is determined to be in at least three stages: not yet flowered, early flowering, and late flowering, using a flowering stage estimation model. By using AI to detect developing male inflorescences in multiple stages according to their growth condition, the determination can be made accurately, quickly, and easily. The AI-based flowering stage estimation model in this embodiment is a color change model for the center of the male inflorescence. The color change model was created using classification criteria that focus on the color and shape of the male inflorescence, and is called the "Coloring-model." Among the growing male inflorescences in aerial photographs, those that do not show flowering are defined as "non-flowering (NF)," those that show flowering and have a yellow center are defined as "early stage of flowering (EF)," and those that show flowering and have a brown center are defined as "late stage of flowering (LF)." Of all the developing male inflorescences detected, as growth progresses, the number of unflowering male inflorescences decreases and the number of early-flowering male inflorescences increases. As growth progresses further, the number of early-flowering male inflorescences decreases and the number of late-flowering male inflorescences increases.

[0017] Figure 2 is a flowchart illustrating the silk extraction date prediction step in the corn harvest timing prediction method according to this embodiment. In the silk extraction date prediction step 30, the predicted silk extraction date is calculated using the number of unflowered tassels of developing tassels determined to be unflowered, the number of early-flowering tassels of developing tassels determined to be in the early-flowering stage, and the number of late-flowering tassels of developing tassels determined to be in the late-flowering stage. Step 30, which predicts the silk extraction date, consists of Step 31, which calculates the number of male inflorescences at each stage, and Step 32, which predicts the date. Step 31, which calculates the number of male inflorescences at each stage, calculates the number of unflowered male inflorescences for growing male inflorescences that have been determined to be in the unflowered stage, the number of early-flowering male inflorescences for growing male inflorescences that have been determined to be in the early-flowering stage, and the number of late-flowering male inflorescences for growing male inflorescences that have been determined to be in the late-flowering stage.

[0018] Figure 2 shows that the calculation results in step 31 for calculating the number of male inflorescences at each stage are, for example, 157 unflowered male inflorescences, 940 early flowering male inflorescences, and 304 late flowering male inflorescences. In prediction step 32, the silk extraction date is predicted using the number of unflowered thorns, the number of thorns in the early flowering stage, and the number of thorns in the late flowering stage, which were calculated in step 31 for each stage, along with the imaging date. For predicting the silk extraction date, for example, a simple linear regression model or a multiple regression model can be used. NF stands for unflowered male inflorescence, and LF stands for late-flowering male inflorescence. The calculation result in prediction step 32 will output, for example, a predicted silk extraction date of July 23rd.

[0019] Figure 3 is a flowchart illustrating the harvest timing prediction step and output step in the corn harvest timing prediction method according to this embodiment. In the harvest timing prediction step 40, the optimal harvest timing is predicted using the accumulated temperature, with the predicted silk extraction date calculated in the silk extraction date prediction step 30 as the starting date. Step 40 of predicting the optimal harvest time consists of Step 41, which calculates the accumulated temperature using the predicted silk extraction date as the starting date; Step 42, which identifies the variety; and Steps 43 and 44, which calculate the predicted optimal harvest time using the accumulated temperature according to the variety. Figure 3 shows an example where variety A in Step 43 is Megumi Star, and variety B in Step 44 is Honey Bantam® 20. Other varieties of sweet corn include Gold Rush®, Pure White®, and Peter Corn. Step 41, which calculates the accumulated temperature using the predicted silk extraction date as the starting date, uses predicted meteorological data. For predicted meteorological data, for example, data provided by the Japan Meteorological Agency, the Agro-Meteorological Grid Square Data (NARO) provided by the National Agriculture and Food Research Organization (NARO), and data provided by other organizations can be used. When using predicted meteorological data, the location name and latitude / longitude information that identify the field should be used. In harvest timing prediction step 40, the accumulated temperature is calculated using the predicted temperature after the predicted silk extraction date, allowing for early prediction of the optimal harvest time. Furthermore, if the predicted silk extraction date has passed by the time of prediction, using the actual temperature up to that point reduces the temperature prediction error in the accumulated temperature, resulting in a more accurate prediction of the optimal harvest time. Furthermore, in the harvest timing prediction step 40, the accumulated temperature can be used according to the variety of corn being grown, allowing for an even more accurate prediction of the optimal harvest time. In output step 50, the harvest time predicted in harvest time prediction step 40 is output. Preferably, the output includes the start and end dates of the predicted harvest time.

[0020] Figure 4 is a photograph illustrating the predictions made by the color change model in this embodiment. Figure 4(a) shows the male and female inflorescences, and Figure 4(b) shows the changes in the male inflorescence over a specified period. As shown in Figure 4(a), the male inflorescence is located at the upper end of the plant body, making aerial photography suitable for its imagery. Figure 4(b) shows images of developing male inflorescences from July 16 to July 28. As growth progresses, the number of flowering male inflorescences increases, and while the center of the male inflorescence is yellow in the early flowering stage, it changes to brown in the later flowering stage. In this way, by focusing on the color change in the center of the male inflorescence, the flowering stage can be determined with high accuracy. Furthermore, by classifying the male inflorescences into three stages using a color change model in the center of the inflorescence, it is possible to calculate the predicted silk extraction date using a linear simple regression model, thereby predicting the optimal harvest time.

[0021] Figure 5 shows a demonstration example of the method for predicting the optimal harvest time for corn according to this embodiment. Figure 5 shows the overlap between the predicted harvest time for corn using this embodiment and the actual harvest time. An overlap indicates that the prediction is correct. Aerial photography was conducted using a drone (Mavic Air 2 or Mavic 2 Pro, DJI) at two sweet corn fields (variety: Megumi Star or Honey Bantam®) at site A and site B. From early July to early August, during the stage from inflorescence to silk emergence, sweet corn inflorescences were photographed at an altitude of approximately 7m with the drone's attached RGB camera pointed directly downwards. The acquired images were 3840 x 2160 pixels, and the central part of the image was then cropped to 1920 x 1080 pixels. Each cropped image contains approximately 10 to 20 inflorescences. Images where the inflorescences were unclear due to overexposure or wind effects were deleted. As shown in Figure 5, the overlap between the two verification sites, A and B, exceeded 84%.

[0022] Figure 6 is a photograph illustrating the flowering status model used as a flowering stage estimation model in the method for predicting the optimal harvest time for corn according to this embodiment. The flowering status model is a classification criterion called "Flowering-mode" that focuses on the flowering status of the male inflorescence. Among the growing male inflorescences in aerial images, those in which no flowering is observed are defined as "non-flowering (NF)", those in which flowering is observed but non-flowering parts can also be seen are defined as "partially flowering (PF)", and those in which flowering is observed in all parts of the growing male inflorescence are defined as "fully flowering (FF)". Figure 6(a) shows the flowers before blooming, Figure 6(b) shows the flowers partially in bloom, and Figure 6(c) shows the flowers in full bloom.

[0023] As shown in Figure 6, a flowering status model can also be used that is based on the ratio of flowering to unflowering parts for a single male inflorescence. In particular, by using such a flowering status model together with a color change model, it is possible to determine four or more stages in the judgment step 20, and then in the silk extraction date prediction step 30, the predicted silk extraction date can be calculated using the number of male inflorescences for each of the four or more stages determined. Thus, by using a flowering status model along with a color change model as a flowering stage estimation model, it is possible to determine four or more stages and calculate the predicted silk extraction date with even greater accuracy. [Industrial applicability]

[0024] The present invention is particularly suitable for sweet corn, but can be applied to popcorn, dent corn, flint corn, waxy corn, soft corn, pod corn, and other types of corn. [Explanation of symbols]

[0025] 10 Data Acquisition Steps 20. Judgment Steps 30. Prediction Steps for Silk Thread Extraction Date 31 Steps for calculating the number of male inflorescences at each stage 32 prediction steps 40. Steps for predicting the optimal harvest time 41 Steps to calculate the accumulated temperature 42 Variety Identification Steps 43, 44 Steps for calculating predicted optimal harvest time 50 output steps

Claims

1. A method for predicting the optimal harvest time for corn, which uses a flowering stage estimation model trained with training image data obtained by adding flowering stage information of corn tassels to training image data obtained by imaging corn tassels, to predict the optimal harvest time for corn, The flowering stage estimation model is a color change model in the center of the male inflorescence, Computers A data acquisition step to acquire growing corn image data obtained by imaging the corn during its growth, A determination step is performed to determine, using the flowering stage estimation model, that each growing male tassel captured in the growing corn image data acquired in the data acquisition step is in at least three stages: not yet flowered, early flowering, and late flowering. A silk extraction date prediction step, which calculates the predicted silk extraction date using the number of unflowered male inflorescences of the growing male inflorescences that have been determined to be unflowered, the number of early-flowering male inflorescences of the growing male inflorescences that have been determined to be in the early-flowering stage, and the number of late-flowering male inflorescences of the growing male inflorescences that have been determined to be in the late-flowering stage. A harvest timing prediction step predicts the optimal harvest time using the accumulated temperature calculated using the actual temperature or predicted temperature, with the predicted silk extraction date calculated in the silk extraction date prediction step as the starting date. Execute A method for predicting the optimal harvest time for corn, characterized by the features described above.

2. In the early flowering stage, the center of the male inflorescence is yellow, and in the later flowering stage, the center of the male inflorescence is brown. The method for predicting the optimal harvest time for corn according to feature 1.

3. As the flowering stage estimation model, along with the color change model, a flowering status model based on the ratio of flowering to unflowering parts for one of the male inflorescences is used. In the aforementioned determination step, the determination is made in four or more stages. In the silk extraction date prediction step, the predicted silk extraction date is calculated using the number of male inflorescences determined in the four or more stages. The method for predicting the optimal harvest time for corn according to feature 1.

4. In the aforementioned harvest timing prediction step, The cumulative temperature shall be determined according to the variety of corn being grown. The method for predicting the optimal harvest time for corn according to feature 1.

5. The aforementioned image data of growing corn was taken from above of the growing corn. A method for predicting the optimal harvest time for corn according to any one of claims 1 to 4.

Images