JP7137426B2 - Harvest prediction system for greenhouse-grown fruits - Google Patents

Description

The present invention relates to a harvest prediction system for greenhouse-grown fruit, for example, regarding the harvest time, harvest amount, size, etc. of greenhouse-grown fruits such as strawberries, predicting the harvest time, harvest amount, and size for each individual fruit. It relates to a harvest prediction system for greenhouse-grown fruits. It also relates to a harvest prediction system for greenhouse-grown fruit that predicts harvest time, yield, and size with higher accuracy using artificial intelligence. In this specification, the cultivation of “strawberry” is explained as an example of the target of greenhouse-grown fruits, but it is not limited to this, for example, tomatoes, publica, oranges, grapes, etc. It is also applied to other greenhouse-grown fruits. be. In addition, the varieties of strawberries exemplified in this specification are all varieties of strawberries such as "Benihoppe" and "Tochiotome".

FIG. 6 shows an embodiment of a greenhouse 100 to which the harvest prediction system 1 is applied. The harvest prediction system 1 is applied to strawberries 2, which are fruits 23 cultivated in a cultivation bed 108 provided in a greenhouse 100 in which the indoor environment is controlled. This greenhouse 100 has a frame made up of pillars 101 and beams 102, and a transparent covering material 107 such as fluorine film or glass is attached to the framework. In addition, the roof is provided with a skylight 103 with an insect repellent net and a stirring fan 104, and the ceiling is provided with an automatic light-shielding curtain 105 for controlling solar radiation and an automatic heat-retaining curtain 106 for averaging the indoor temperature. , the indoor environment of the greenhouse 100 is controlled. The greenhouse 100 shown in FIG. 6 is an example of a facility for so-called "house cultivation", and the greenhouse 100 of the present invention is not limited to this example.

A cultivation bed 108 for cultivating strawberries 2 as cultivated fruits 109 is juxtaposed in the greenhouse 100 , and a seedling of the strawberry 2 is provided on the cultivation bed 108 . Then, an automatic traveling robot 113 travels along the traveling path 110 between the cultivation beds 108, and an imaging camera 114 mounted on the automatic traveling robot 113 captures an image of the strawberry 2. This automatic traveling robot 113 is provided with an image monitor 112 and a transmitting/receiving antenna 111, displays information about the cultivated strawberries 2 on the image monitor 112, and displays information about the cultivated strawberries 2 on a centralized monitor or the like. Send and receive. In addition, the automatic traveling robot 113 is provided with a notification means such as a warning sound generator 115, and when a defective product such as a deformed fruit is generated in the cultivated strawberry 2, or when the strawberry 2 undergoes mutation, A warning sound is generated to inform the person in charge of the situation.

Strawberries 2 are easily affected by natural environments such as typhoons and long rains, so in recent years, they are often cultivated in greenhouses 100 in which the indoor environment is controlled. In this way, cultivation in the greenhouse 100 has made it possible to control the harvest time. However, pests and diseases typified by "mildew" easily spread within the greenhouse 100 once they occur, so quality control of the strawberries 2 requires manual labor.

Patent Literature 1 discloses a harvest prediction device and method that can reduce time and cost and predict the yield of crops in a wide area. Here, the image analysis unit acquires spectral data from the image data showing crops, compares the acquired spectral data with the spectral data stored in the spectrum DB, identifies the type of crop and the stage of cultivation, and Calculate area. Predicting the crop yield from the crop type, growing stage, and planted area obtained by the image analysis unit, and from the crop growing period and yield per unit area stored in the crop information DB. is described.

Patent Literature 2 discloses a harvesting device capable of reliably and quickly automatically harvesting even a yellow-green crop. In this embodiment, a self-propellable mobile body having a picking section and a travel control section for harvested materials and a position detecting device of the mobile body from a transmitter and a receiver are provided, and the mobile body detects the blooming of the harvested material. Equipped with an image processing device for recognizing conditions and fruiting conditions, a memory for storing flowering positions and times recognized by the image processing device, and a database for calculating the time for fruiting from the flowering conditions, and inputting data into the database It is described that a calculation unit is provided for predicting the time and position of fruiting based on the content and the detection result of the image processing device.

In Patent Document 3, based on data obtained by sensing with an unmanned aerial vehicle such as a drone, management of farmland, etc., growth prediction of agricultural crops, harvest prediction, etc. are automatically performed, and proper management of agricultural land and crops, etc. Disclosed is an agricultural management forecasting system that achieves: Here, in the unmanned aerial vehicle and the server device that can freely communicate with the management terminal, a communication unit that receives image data sent from the management terminal, a vegetation indexing unit that converts the vegetation index based on the image data, an index Equipped with a standard deviation unit for standard deviation of information and a semantic unit for semanticizing the standard deviation. It is described that management prediction of agriculture is performed on the server device side.

JP-A-2003-6612 JP-A-6-133624 JP 2018-46787 A

In the greenhouse cultivation of strawberries, there is a problem that the loss rate increases due to overripeness because strawberries do not last for a long time. There was a problem that it was necessary to carry out the work, and the labor cost increased.

In addition, in the greenhouse cultivation of strawberries, there is a problem that it is impossible to accurately predict the growing period of the fruit, the harvest amount, etc., or that it is necessary to rely on the intuition of an expert. In addition, there is a problem that the size of the fruit, which greatly affects the selling price, cannot be accurately predicted.

The purpose of the present application is to solve such problems and accurately predict the harvest date, harvest amount, size, etc. of the fruit, so that the product value is high by securing a cheap and reliable harvest labor force and an excellent and reliable shipping destination. It is to provide a harvest prediction system for greenhouse-grown fruits.

In addition, in the greenhouse cultivation of fruit, the individual feature values of each fruit are extracted by artificial intelligence neural networks, etc., and individual management of each fruit is performed, and the harvest date, yield, size, etc. of the fruit can be predicted without human intervention. It is possible to provide a harvest prediction system for greenhouse-grown fruits with improved commercial value.

Furthermore, in the greenhouse cultivation of fruits, it is possible to predict the harvest date, harvest amount, size, etc. of fruits with higher accuracy by referring to past big data by deep learning of artificial intelligence, etc. of harvest prediction system can be provided.

In order to achieve the above object, the yield prediction system for greenhouse-grown fruits according to the present invention captures images of fruits grown in a greenhouse in which the indoor environment is controlled with an imaging camera, and the time-lapse images captured by the imaging camera are used to predict the yield of fruits. A harvest date prediction unit that determines the flowering date and sets the harvest date based on the planned integrated temperature in the greenhouse after the flowering date, and a fruit harvest number prediction unit that integrates information on the number of fruit blooms and predicts the number of fruit harvests. and a fruit size prediction unit for estimating fruit size information from the flower bed volume.

With the above configuration, for each fruit cultivated in a greenhouse where the indoor environment is controlled, it is possible to accurately predict the harvest date, harvest number, and size of the fruit from time-lapse images captured by the imaging camera. For example, it is possible to determine the flowering date of the fruit from periodic time-lapse images taken by an imaging camera, and set the harvest date of the fruit from the planned accumulated temperature in the greenhouse after the flowering date. Further, by accumulating the predicted fruit harvest dates for all the fruits in the greenhouse, it is possible to predict the number of fruits harvested in the greenhouse as a whole. Furthermore, the weight of fruits such as strawberries grown in greenhouses is almost proportional to the flower bed volume. Therefore, by measuring the flower bed volume of the fruit, it is possible to predict in advance the yield of large fruits such as strawberries, which are sold at a high unit price.

In addition, the yield prediction system for greenhouse-grown fruit includes an identification number assigning unit that assigns an identification number to each fruit, and a feature amount extraction unit that analyzes the time-lapse image of the fruit, extracts the individual feature amount, and associates it with the identification number. is preferably provided. In this way, a time-lapse image of a fruit such as a strawberry is analyzed by a neural network of artificial intelligence or the like to extract the individual feature amount of the image and assign an identification number to the image. In this way, by individually recognizing and individually managing fruits such as strawberries to be cultivated, more accurate quality control is possible.

Further, in the greenhouse-grown fruit harvest prediction system, it is preferable that the individual feature amount of each fruit photographed by the feature amount extraction unit is stored as the individual feature amount in the individual fruit management database together with the identification number. In this way, fruits such as strawberries are individually selected by individual feature amounts such as the shape of the strawberry itself, or the peripheral shape related to the strawberry itself such as "sepal", which differs for each fruit. can be identified. Based on this individual fruit management database, more accurate quality control becomes possible.

In addition, in the greenhouse-grown fruit harvest prediction system, the feature quantity extraction unit compares the individual feature quantity of the fruit photographed by the imaging camera with the individual feature quantity of the fruit stored in the individual fruit management database, and identifies the fruit. preferable. As a result, fruits such as strawberries photographed by an imaging camera grow periodically and change in shape. can be managed.

In addition, in the system for predicting the yield of fruits grown in a facility, it is preferable that the individual fruit management database stores history information of temporal images of flowerbeds of each fruit. As a result, time-lapse images of flowerbeds of fruits such as strawberries can be saved as big data, and utilized in predicting the harvest date, harvest number, and size of fruit cultivation every year.

Further, in the greenhouse-grown fruit harvest prediction system, the harvest date prediction unit preferably corrects the set harvest date from the time-lapse images of the fruit stored in the individual fruit management database. As a result, if the harvest date set from the flowering date of fruits such as strawberries and the planned integrated temperature in the house after the flowering date deviates from the initial harvest date due to other factors such as sunshine conditions, individual It is possible to correct the harvest date set by artificial intelligence deep learning etc. by reflecting the past time-lapse image of fruits such as strawberries stored in the fruit management database.

In addition, in the greenhouse-grown fruit yield prediction system, the fruit size prediction unit predicts the size of the fruit based on a comparison between past big data on the size of the fruit stored in the individual fruit management database and images of the fruit over time. is preferred. As a result, artificial intelligence can accurately predict the final size of fruits such as strawberries based on past progress information when the flower beds of fruits such as strawberries become enlarged and become fruits. Note that the “size related to the fruit such as strawberry” may be the weight of the fruit such as strawberry.

In addition, in the system for predicting the yield of fruits grown in a facility, it is preferable that the imaging camera is mounted on an automatic traveling robot and that the robot travels side by side along a cultivation bed adjusted to a height that allows close-up photography. As a result, the automatic traveling robot can take close-up photographs of flowerbeds, fruits, etc. of each strawberry or other fruit provided on the cultivation bed, and analyze them in detail using artificial intelligence.

Furthermore, the system for predicting the yield of fruits grown in greenhouses is further equipped with a fruit quality confirmation unit, and if an abnormality occurs in the fruit from the photographed image of the fruit by the automatic traveling robot, an alarm can be issued as a defective product. preferable. As a result, abnormalities in fruits such as strawberries that have relied on visual inspection until now. It can be detected in advance by artificial intelligence. In addition, it is possible to detect whether or not a disease such as powdery mildew has occurred not only on the fruit but also on the leaves.

As described above, according to the harvest prediction system for greenhouse-grown fruit according to the present invention, by accurately predicting the harvest date, harvest amount, size, etc. of the fruit, a cheap and reliable harvest labor force and an excellent and reliable By securing a shipping destination, a harvest prediction system for greenhouse-grown fruits with high commercial value can be provided.

In addition, in the greenhouse cultivation of fruit, the individual feature values of each fruit are extracted by artificial intelligence neural networks, etc., and individual management of each fruit is performed, and the harvest date, yield, size, etc. of the fruit can be predicted without human intervention. It is possible to provide a harvest prediction system for greenhouse-grown fruits with improved commercial value.

Furthermore, in the greenhouse cultivation of fruits, it is possible to predict the harvest date, harvest amount, size, etc. of fruits with a higher degree of accuracy by referring to past big data using artificial intelligence deep learning, etc., and high-quality greenhouse cultivation. A fruit harvest prediction system can be provided.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a schematic configuration of an embodiment of a system for predicting yield of greenhouse-grown fruit according to the present invention; It is explanatory drawing which shows the individual fruit management database which manages fruits, such as a strawberry cultivated in a greenhouse, individually. It is explanatory drawing which shows the structure which predicts a harvest date by this harvest prediction system. It is explanatory drawing which shows the structure which predicts a yield by this harvest prediction system. It is explanatory drawing which shows the structure which predicts the size of a fruit by this harvest prediction system. 1 is a plan view showing one embodiment of a greenhouse to which this harvest prediction system is applied; FIG. 1 is an explanatory diagram showing one schematic structure of an automatic traveling robot equipped with an imaging camera in a greenhouse; FIG. FIG. 2 is an explanatory diagram showing names of strawberry flowers and fruits, which is one example of fruits.

(Configuration of fruit harvest prediction system)
Hereinafter, the yield prediction system 1 for greenhouse-grown fruit according to the present invention will be described in detail with reference to the drawings. FIG. 1 shows a block diagram of a schematic configuration of one embodiment of the harvest prediction system 1. As shown in FIG. In this specification, the case where the fruit 23 cultivated in a facility is "strawberry" will be described, but the present invention is not limited to this, and can also be applied to other strawberries 2 cultivated in a facility such as tomatoes, publica, mandarin oranges, and grapes. be. In addition, the varieties of strawberries 2 exemplified in this specification are all varieties of strawberries 2 such as “Benihoppe” and “Tochiotome”.

(automatic driving robot)
FIG. 7 shows one schematic structure of the automatic traveling robot 113 on which the imaging camera 114 is mounted. The automatically traveling robot 113 that travels inside the greenhouse 100 automatically travels around the cultivation bed 108 . The automatic traveling robot 113 is equipped with an imaging camera 114 so that the height of the cultivation bed 108 can be taken close-up. This enables the automatic traveling robot 113 to take close-ups of all the strawberries 2 inside the greenhouse 100 . The imaging camera 114 periodically captures temporal images 20 (see FIG. 1) of all the strawberries 2 cultivated in the greenhouse 100 .

(Configuration of yield prediction system for greenhouse-grown fruit)
FIG. 1 shows the configuration of a harvest prediction system 1. As shown in FIG. Each strawberry 2 cultivated in the cultivation bed 108 is periodically photographed as time-lapse images 20 by the imaging camera 114 . This time-lapse image 20 is basically assumed to be taken once a day when flowering is completed, but it is not limited to this time point and may be taken as necessary. Then, the harvest prediction system 1 predicts the harvest date, harvest number, and fruit size of the strawberries 2 based on the time-lapse image 20 of each strawberry 2 that has been photographed. The harvest prediction system 1 comprises a fruit harvest date prediction unit 3, a fruit harvest number prediction unit 4, a flower bed volume calculation unit 5, and a fruit size prediction unit 6.

(Fruit harvest date prediction, harvest number prediction, fruit size prediction)
FIG. 8 shows names of flowers and fruits of cultivated strawberries 2 . 8(a) is a strawberry flower, and FIG. 8(b) is a strawberry fruit 23. FIG. The fruit harvest date prediction unit 3 determines the flowering date of the strawberry 2 from the time-lapse image 20 captured by the imaging camera 114 . After the flowering date, the number of days from the flowering date of the strawberry 2 to the harvest date is set from the planned integrated temperature 27 of the greenhouse 100 . The unit of this planned cumulative temperature 27 is expressed in (°C.day). In the case of the strawberry 2, the number of days from flowering to harvest can be obtained from the varieties such as “Benihoppe” and “Tochiotome” and the accumulated temperature in the greenhouse 100, for example. In other words, the harvest date can be almost accurately predicted in a greenhouse where the temperature is controlled. The fruit harvest number prediction unit 4 predicts the number of strawberries 2 to be harvested by accumulating the information on the flowering dates of the strawberries 2 predicted by the fruit harvest date prediction unit 3 , that is, the flowering number information 12 . The number of healthy fruits that can be harvested from blooming flowers varies depending on the variety, climate, disease, pests, actions of honeybees, etc., but can be roughly predicted from past data. As one calculation formula, the number of healthy fruits can be expressed as the number of flowering times the rate of fruiting x (1-defective product rate). Here, the defect rate means the probability of pest damage or deformation. The flower bed volume calculation unit 5 calculates the volume of the flower bed 19a of the strawberry 2 photographed by the imaging camera 114 . This flower bed volume 14 (V) is obtained by V=radius (R) of flower bed 19a×π×height/3. That is, a seed is attached to the flower bed 19a and one seed is included. As the number of seeds on the flowerbed 19a increases, the strawberry 2 grows larger. From this, it can be said that there is a correlation between the flower bed volume 14 and the size of the strawberry 2 . The fruit size prediction unit 6 predicts the size information 13 of the strawberry 2 from the volume of the flower bed 19a calculated by the flower bed volume calculation unit 5. FIG. This is because "the volume 14 of the flower bed of the strawberry 2 and the weight of the harvested fruit are in a substantially proportional relationship" as described above.

(Individual recognition and individual management of fruits)
FIG. 1 shows an identification number assigning unit 7 that analyzes the time-lapse image 20 captured by the imaging camera 114 and assigns an identification number 18 to the strawberry 2, and a feature amount extraction unit that extracts the individual feature amount 15 of the strawberry 2. 8 is shown. Also shown is an individual fruit management database 10 for individually recognizing and managing strawberries 2 cultivated in the greenhouse 100 . The individual fruit management database 10 stores an identification number 18 and an individual feature amount 15 relating to the strawberries 2 grown in the greenhouse 100 .

As described above, the harvest prediction system 1 includes the identification number assigning unit 7 and the feature amount extracting unit 8, and the identification number assigning unit 7 assigns an identification number 18 to each strawberry 2 for identification. The individual feature quantity 15 of the strawberry 2 includes, for example, the number, position, and shape of pistils 19b and stamens 19c of the flower bed 19a shown in FIG. The number, position, and shape of the petals 19g, etc., and the characteristics by which each strawberry 2 can be distinguished from other strawberries 2. Then, the individual feature amount 15 is either the individual feature amount 15 at the stage of the "flower bed 19a" of the strawberry 2 or the individual feature amount 15 at the stage where the flower bed 19a has enlarged and developed to become flesh. is also included. The feature amount extraction unit 8 extracts these individual feature amounts 15 and associates them with the identification number 18 .

The feature amount extraction unit 8 compares the individual feature amount 15 of the strawberry 2 with the individual feature amount 15 of the strawberry 2 stored in the individual fruit management database 10 each time the image is captured by the imaging camera 114, and determines that the strawberry 2 is identify That is, the strawberry 2 is periodically photographed by the imaging camera 114 , and each time the individual fruit management database 10 is searched, the strawberry 2 can be identified from the individual feature amount 15 . As a result, the latest image of strawberry 2 is acquired every time strawberry 2 is imaged by imaging camera 114 . In this way, the individual fruit management database 10 stores the history information 17 for the temporal image 20 of each strawberry 2 .

FIG. 2 shows the contents of the individual fruit management database 10 that individually manages the strawberries 2 grown in the greenhouse 100. As shown in FIG. As shown in the upper part of FIG. 2, in the greenhouse 100, the strains of strawberries 2 are indicated as "strain 1", "strain 2", . . . "strain n". In each strain, 19 g of the petals of the strawberries 2 become flesh, but these strawberries 2 are assigned an identification number 18 and managed individually. For example, stock 1 is given stock 1-1 to stock 1-4 as an identification number 18, stock 2 is given stock 2-1 to stock 2-3 as an identification number 18, and stock n is given an identification number. 18, strain n-1 to strain n-3. Then, in the individual fruit management database 10, the individual feature quantity 15 associated with the identification number 18 for each strawberry 2 is stored.

As an example, the individual fruit management database 10 stores the individual feature amount 15 for the strawberry 2 with the identification number 18 (strain 2-3), a time-lapse image 20 of the flowering date of the strawberry 2, and the fruiting of the strawberry 2. A day-lapse image 20 and a time-lapse image 20 at the time of harvesting the strawberries 2 are stored together with the shooting date and time. Then, in the time-lapse image 20 of the flowering date of the strawberry 2, information that it has bloomed is written together with the record of the year, month, day and time, and in the time-lapse image 20 of the fruiting date of the strawberry 2 Along with recording the time, information indicating that the strawberry 2 has fruited is written, and a time-lapse image 20 of the strawberry 2 at the time of harvesting is stored together with the date and time of photography. Then, in the chronological image 20 of the harvest date of the strawberry 2, information indicating that the strawberry 2 was harvested is written together with the record of the year, month, day and time. From this information, the number of days information (N1) from the flowering date to the fruiting date of all the strawberries 2 and the number of days information (N2) from the fruiting date to the harvest date are calculated and recorded in the individual fruit management database 10. be.

(Calculation of Corrected Harvest Date)
FIG. 3 is a block diagram showing a configuration for predicting the harvest date of each strawberry 2 by the fruit harvest date prediction unit 3. As shown in FIG. The fruit harvest date prediction unit 3 corrects the harvest date set from the temporal image 20 of the strawberry 2 stored in the individual fruit management database 10 . That is, as described above, the fruit harvest date prediction unit 3 determines the flowering date information 25 based on the cultivation plan 26 of the strawberries 2 from the temporal images 20 captured by the imaging camera 114 . Then, after the flowering date, the number of days from the flowering date of the strawberry 2 to the expected harvest date 28a is set based on the planned accumulated temperature 27 in the greenhouse 100. FIG. In order to set the planned integrated temperature 27, the product type data 31a is required. This method is based on the fact that "if the environment is controlled and cultivated in the greenhouse 100, it is possible to predict the number of days to maturity from the date of flowering to the date of harvest to some extent." However, the set harvest date may change due to climate change related to the growth of the strawberries 2, the occurrence of diseases and pests, the working conditions of honeybees and the like. For example, the strawberries 2 are grown in the greenhouse 100 where the indoor environment is controlled, so the indoor temperature should be constant, but they may be affected by climate change due to typhoons, cold waves, and the like. On the other hand, the corrected harvest date 28b can be calculated by predicting the harvest date 30a by deep learning of artificial intelligence or the like.

For calculating the corrected harvest date 28b, as shown in FIG. It is possible to adopt a method of predicting the number of days (N2) from the harvest date to the harvest date as big data using artificial intelligence. That is, the fruit harvest date prediction unit 3 can correct the harvest date set from the temporal image 20 of the strawberry 2 stored in the individual fruit management database 10 . That is, the number of days of maturity from the flowering date to the expected harvesting date 28a of the strawberry 2 is set by the expected integrated temperature 27 within 100 from the flowering date to the harvesting date. However, as described above, the number of days for ripening may vary due to factors other than the planned integrated temperature 27 , and this is corrected by the temporal image 20 of the strawberry 2 stored in the individual fruit management database 10 . For example, regarding the maturity of the strawberry 2, if the maturity at a certain point is earlier or later than the past data, the corrected harvest date 28b is calculated considering the error (X) of the number of days. .

(Example of using artificial intelligence)
The fruit harvest date prediction unit 3 uses artificial intelligence deep learning to obtain the maturity of the strawberries 2 from the time-lapse images 20 of all the strawberries 2 stored in the individual fruit management database 10. The state can be arranged in a plane. For example, deep learning is used to analyze how much the ripening of strawberries at which coordinate position or height position is earlier or later when viewed from past data. As a result, it is possible to ascertain whether the cause of the error (X) in the number of days described above is climate change, the occurrence of disease or insect pests, the working condition of bees or the like, or other reasons.

Artificial intelligence can detect changes in the number of days to ripen due to factors other than the scheduled integrated temperature 27 without overlooking subtle changes in the time-lapse images 20 of the strawberries 2 that are taken periodically. Accumulating this variation can then accurately correct the final harvest date. In addition, when cultivating outside the greenhouse 100, it is generally possible to predict the harvest time to some extent if there is standard data of the cultivated crops and average weather data, but prediction is performed by artificial intelligence. Depending on the method, it is possible to predict the harvest time in consideration of the case where the number of days for ripening varies due to factors other than the planned integrated temperature 27 .

(Calculation of Corrected Harvest Number)
FIG. 4 is a block diagram showing a configuration for predicting the number of harvests of each strawberry 2 by the fruit harvest number prediction unit 4. As shown in FIG. The fruit harvest number prediction unit 4 corrects the harvest number set from the temporal image 20 of the strawberry 2 stored in the individual fruit management database 10 . That is, as described above, the fruit yield prediction unit 4 determines the flowering date information 25 of the strawberries 2 from the temporal images 20 captured by the imaging camera 114 . After the date of flowering, the date of fruiting of the strawberry 2 is predicted from the scheduled cumulative temperature 27 in the greenhouse 100, and the expected number of harvests 29a is set. In order to set the expected number of harvests 29a, the fruiting rate is required as the variety data 31b. According to this method, "if the environment is controlled and cultivated in the greenhouse 100, it is possible to predict to some extent the number of days to maturity from the date of flowering to the date of harvest." However, the set harvest date may change due to climate change, the occurrence of diseases and pests, the working conditions of honeybees, etc. related to the growth of the strawberries 2, and the number of harvests may change accordingly. In such a situation, the corrected harvest number 29b can be calculated by performing the harvest number prediction 30b in consideration of the harvest number error (Y) using the artificial intelligence described above.

Artificial intelligence can detect changes in the number of days to ripen due to factors other than the scheduled integrated temperature 27 without overlooking subtle changes in the time-lapse images 20 of the strawberries 2 that are taken periodically. Then, by accumulating this variation, the final harvest numbers can be accurately corrected. In addition, when cultivating outside the greenhouse 100, it is generally possible to predict the number of harvests to some extent if there is standard data of the crops to be cultivated and average weather data, but prediction is performed by artificial intelligence. Depending on the method, it is possible to predict the number of harvests in consideration of the case where the number of days for maturation varies due to factors other than the planned integrated temperature 27 .

(Fruit size prediction system)
FIG. 5 is a block diagram showing a configuration for predicting the size of each strawberry 2 by the fruit size prediction unit 6. As shown in FIG. The fruit size prediction unit 6 predicts the size information 13 of the strawberry 2 from the volume of the flower bed 19 a calculated by the flower bed volume calculation unit 5 based on the temporal image 20 of each strawberry 2 . This is because it is said that "the flower bed volume 14 and the weight of harvested fruits are in a proportional relationship." However, the set harvest date and harvest amount may fluctuate due to climate change, the occurrence of diseases and insect pests, and the working conditions of honeybees, etc., which are related to the growth of the strawberries 2, and the fruit size may fluctuate accordingly. In such a situation, the corrected size information 13 of the strawberry 2 can be calculated by performing the fruit size prediction 30c by artificial intelligence.

As shown in FIG. 5, the fruit size prediction unit 6 compares the past fruit harvest size data 32 stored in the individual fruit management database 10 of the individual strawberry 2 with the time-lapse image 20 of the strawberry 2. can be predicted by modifying the size information 13 of . For example, regarding the ripening of the strawberry 2, if the size information 13 at a certain point in time is large or delayed from the past data, the size information corrected in consideration of the error (Z) of the size information 13 is is expected.

The fruit size prediction unit 6 uses artificial intelligence deep learning to planarly organize the inside of the greenhouse 100 with respect to the size of the strawberries 2 from the time-lapse images 20 of all the strawberries 2 stored in the individual fruit management database 10. be able to. For example, deep learning is used to analyze how much the ripening of strawberries at which coordinate position or height position is earlier or later when viewed from past data. As a result, it is possible to ascertain whether the cause of the error (Z) in the size information 13 described above is climate change, the occurrence of disease or pests, the working condition of honeybees, or other reasons. .

In the case of strawberries, the sale destination differs depending on the size information 13 thereof. Therefore, it is important in terms of sales strategy to know in advance how much and what size strawberries 2 can be harvested. In this way, the size of the strawberry 2 can be predicted to some extent from the flowerbed volume 14, and the size distribution of crops other than the strawberry 2 can be estimated to some extent. It is possible to predict the size considering the case where the number of days to maturity fluctuates due to other factors.

(Quality control of fruit by automatic driving robot)
The harvest prediction system 1 further includes a fruit quality checking section 9 . That is, in principle, the automatic traveling robot 113 has the feature of regularly photographing all the strawberries 2 in the greenhouse 100 without omission. Moreover, it has the characteristic that the quality of the strawberry 2 can be managed individually. Taking advantage of these characteristics, when an abnormality in the strawberry 2 that has been relied on visually until now, for example, when the strawberry 2 does not grow due to disease, or when a defective product such as a deformed fruit occurs in the strawberry 2, the strawberry 2 suddenly Strawberries 2 that cannot be shipped due to mutation, etc., can be eliminated in advance by artificial intelligence.

In addition, the harvest prediction system 1 detects not only the strawberry 2 but also the leaves of the strawberry 2 to detect whether signs of a disease such as powdery mildew have occurred. can teach. Then, when the occurrence of disease or the occurrence of defective products is detected, the automatic traveling robot 113 sends the identification number 18 of the strawberry 2 and the type of problem to the monitor of the automatic traveling robot 113 or the centralized monitor for quality control. can be displayed. For the quality control of these strawberries 2, labor costs can be reduced and check omissions can be minimized by substituting the work that has been performed visually until now with the automatic traveling robot 113.

The configuration, shape, size, and layout of the greenhouse-grown fruit yield prediction system 1 described in the above embodiment are merely schematic representations to the extent that the present invention can be understood and implemented. Therefore, the present invention is not limited to the described embodiments, but can be modified in various forms without departing from the scope of the technical concept indicated in the claims.

1 Harvest prediction system (of greenhouse-grown fruit) 2 Strawberry 3 Fruit harvest date prediction unit 4 Fruit harvest number prediction unit 5 Flowerbed volume calculation unit 6 Fruit size prediction unit 7 Identification number assignment unit 8 Feature quantity Extraction part 9 Fruit quality confirmation part 10 Individual fruit management database 12 Flowering number information 13 Size information 14 Flower bed volume 15 Individual feature quantity 17 History information 18 Identification number 19a Flower bed 19b Pistil 19c Stamen, 19d Seed, 19e Sepal, 19f Fruit stalk, 19g Petal, 20 Time-lapse image, 23 Fruit, 25 Flowering date information, 26 Cultivation plan, 27 Planned integrated temperature, 28a Expected harvest date, 28b Corrected harvest date, 29a Forecast Harvest number, 29b Corrected harvest number, 30a Harvest date prediction by artificial intelligence, 30b Harvest number prediction by artificial intelligence, 30c Fruit size prediction by artificial intelligence, 31a Variety data (accumulated temperature), 31b Variety data (fruiting rate), 32 Past fruit harvest size data, 100 greenhouses, 101 pillars, 102 beams, 103 skylights with insect nets, 104 stirring fans, 105 automatic light-shielding curtains, 106 automatic heat-retaining curtains, 107 transparent covering materials, 108 cultivation beds, 109 cultivated fruits (strawberry), 110 (automatic traveling robot) traveling path, 111 transmitting/receiving antenna, 112 image monitor, 113 automatic traveling robot, 114 imaging camera, 115 warning sound generator, 116 drainage port, 117 drip watering tube, 118 culture medium, N1 that Information on the number of days from the flowering date of the fruit to the date of bearing fruit, N2 Information on the number of days from the date of bearing fruit to the harvest date, X error in number of days, Y error in number of harvests, Z error in size.

Claims (9)

Photographing a fruit -bearing plant cultivated in a greenhouse where the indoor environment is controlled with an imaging camera,
a fruit harvest date prediction unit that determines the flowering date of the plant from the time-lapse images captured by the imaging camera and sets the harvesting date from the planned accumulated temperature in the greenhouse after the flowering date;
a fruit harvest number prediction unit that predicts the harvest number of the fruit by accumulating the flowering number information of the plant ;
a flower bed volume calculation unit that calculates the flower bed volume of the plant imaged by the imaging camera;
a fruit size prediction unit that predicts the size of the fruit from the volume of the flower bed;
with
The yield prediction system for greenhouse-grown fruit, wherein the plant is strawberry. 2. The harvest prediction system for greenhouse-grown fruit according to claim 1 , wherein an identification number assigning unit assigns an identification number to each fruit, and an individual feature amount of the fruit is extracted by analyzing the time-lapse image of the fruit. and a feature quantity extraction unit associated with the identification number. 3. The harvest prediction system for greenhouse-grown fruit according to claim 2 , wherein the individual feature amount of each fruit photographed by the feature amount extraction unit is stored as an individual feature amount in an individual fruit management database together with the identification number. A harvest prediction system for fruits grown in a facility, characterized by: 4. The harvest prediction system for greenhouse-grown fruit according to claim 1 , wherein the feature quantity extraction unit stores the individual feature quantity of the fruit photographed by an imaging camera in an individual fruit management database. A yield prediction system for greenhouse-grown fruit, characterized in that the fruit is identified by comparing with individual feature values of the fruit. 5. The yield prediction system for greenhouse-grown fruit according to any one of claims 1 to 4 , wherein the individual fruit management database stores history information of temporal images of flowerbeds of each fruit. A harvest prediction system for fruits grown in facilities. 6. The harvest prediction system for the greenhouse-grown fruit according to any one of claims 1 to 5 , wherein the fruit harvest date prediction unit includes a harvest date set from the time-lapse image of the fruit stored in the individual fruit management database. A yield prediction system for greenhouse-grown fruit, characterized by correcting the 7. The yield prediction system for greenhouse-grown fruit according to any one of claims 1 to 6 , wherein the fruit size prediction unit includes past big data related to the size of the fruit stored in an individual fruit management database, A yield prediction system for greenhouse-grown fruit, characterized by predicting the size of the fruit based on comparison with time-lapse images of the fruit. 8. The yield prediction system for greenhouse-grown fruit according to claim 1 , wherein the imaging camera is mounted on an automatic traveling robot and is adjusted to a height that allows close-up photography along a cultivation bed A harvest prediction system for greenhouse-grown fruit, characterized in that the fruits run side by side in a greenhouse. The harvest prediction system for greenhouse-grown fruit according to any one of claims 1 to 8 , further comprising a fruit quality confirmation unit, wherein an abnormality occurs in the fruit from an image of the fruit captured by an automatic traveling robot A harvest prediction system for greenhouse-grown fruit, wherein an alarm is issued as a defective product when a defective product is found.

Images