Greenhouse cucumber irrigation system based on multi-modal information fusion

Abstract

The invention discloses a greenhouse cucumber irrigation system based on multi-modal information fusion, and relates to the technical field of intelligent irrigation, and the system comprises a cucumber water demand strategy module which carries out cucumber growth index monitoring and image instance segmentation reasoning to obtain a cucumber phenotype index and an image reasoning result, and carries out time scale analysis to obtain a cucumber water demand strategy; the irrigation model construction module is used for training the deep learning model according to the historical environment data, the historical cucumber image data and the cucumber water demand strategy to obtain a multi-modal information fusion irrigation model; and the automatic irrigation realization module is used for collecting environment data and cucumber image data of the day and inputting the data into the multi-modal information fusion irrigation model to obtain the irrigation amount of the day, and automatic irrigation is carried out. Through combination of cucumber growth index monitoring, image instance segmentation reasoning, the multi-modal information fusion irrigation model and the automatic irrigation implementation module, the dynamic irrigation requirement of the whole growth period of cucumbers can be met, and the irrigation accuracy is improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent irrigation technology, and in particular to a greenhouse cucumber irrigation system based on multimodal information fusion. Background Art

[0002] Under different irrigation amounts, the growth, dry matter accumulation, quality and nutrient absorption of cucumbers are affected to varying degrees. Too much or too little irrigation will affect crop growth and development, resulting in a decrease in yield. Therefore, it is necessary to study the cucumber irrigation system.

[0003] In the prior art, Chinese patent CN116957141A discloses an intelligent irrigation prediction method and system based on the Transformer model, which predicts whether plants need irrigation through environmental factors so as to start irrigating the plants at the best time. It uses the collected historical information on temperature and humidity, humidity changes during irrigation, light, weather and climate, location and other dimensions, combined with a deep learning framework, to build an intelligent irrigation system based on the Transformer model; and uses machine learning methods to replace manual judgment on whether irrigation is needed.

[0004] However, the above-mentioned existing technology simply relies on sensor data to determine the water demand of green plants, has a single modality, and ignores the different water requirements of green plants in different growth stages. If used for watering cucumbers, it cannot meet the dynamic irrigation needs of cucumbers throughout the entire growth period, and the irrigation accuracy is poor. Summary of the invention

[0005] The present application provides a greenhouse cucumber irrigation system based on multimodal information fusion to solve the problem that the existing technology cannot meet the dynamic irrigation needs of cucumbers throughout their entire growth period and has poor irrigation accuracy.

[0006] On the one hand, the present application provides a greenhouse cucumber irrigation system based on multimodal information fusion, including: a cucumber water requirement strategy module, an irrigation model construction module and an automatic irrigation implementation module.

[0007] The cucumber water requirement strategy module is used to monitor cucumber growth indicators and perform image instance segmentation reasoning to obtain cucumber phenotypic indicators and image reasoning results, and perform time scale analysis based on the cucumber phenotypic indicators and the image reasoning results to obtain the cucumber water requirement strategy.

[0008] The irrigation model construction module is used to train the deep learning model according to the historical environmental data, the historical cucumber image data and the cucumber water demand strategy to obtain a multimodal information fusion irrigation model.

[0009] The automatic watering realization module is used to collect the environmental data of the day and the cucumber image data of the day and input them into the multimodal information fusion watering model to obtain the watering amount of the day, and automatically water according to the watering amount of the day.

[0010] In a possible implementation, the cucumber water requirement strategy module includes: a growth index monitoring unit, an instance segmentation reasoning unit and a time scale analysis unit.

[0011] The growth index monitoring unit is used to monitor the growth index of cucumber and obtain the phenotypic index of cucumber.

[0012] The instance segmentation reasoning unit is used to perform image instance segmentation reasoning to obtain an image reasoning result.

[0013] The cucumber phenotypic index and the image reasoning result correspond in time scale.

[0014] The time scale analysis unit is used to perform time scale analysis according to the cucumber phenotypic index and the image reasoning result to obtain the cucumber water requirement strategy.

[0015] In a possible implementation, performing time scale analysis based on the cucumber phenotypic indicators and the image reasoning results to obtain the cucumber water requirement strategy includes: performing principal component analysis dimensionality reduction and smoothing filtering on the cucumber phenotypic indicators and the image reasoning results on a time scale to obtain the relationship between the variation range of the cucumber throughout the growth period and the water requirement, that is, the cucumber water requirement strategy.

[0016] In a possible implementation, the smoothing filter includes: moving average smoothing technology, exponential smoothing technology, polynomial regression smoothing technology, LOESS smoothing technology, Savitzky-Golay filtering smoothing technology, Gaussian smoothing technology and median filtering smoothing technology.

[0017] In a possible implementation, the irrigation model construction module includes: a feature extraction model construction unit and a deep learning model construction unit.

[0018] The feature extraction model construction unit is used to use the historical cucumber image data to train the instance segmentation reasoning unit in the cucumber water demand strategy module to obtain a feature extraction model, and input the historical cucumber image data into the feature extraction model to obtain image feature data.

[0019] The deep learning model construction unit is used to align the historical environmental data and the image feature data in time series as input features, and use the cucumber water requirement strategy as the target feature to train the deep learning model to obtain a multimodal information fusion irrigation model.

[0020] In a possible implementation, the instance segmentation reasoning unit is trained using YOLO11s-seg weights.

[0021] In a possible implementation, the deep learning model adopts an iTransformer time series model.

[0022] In a possible implementation, the automatic watering implementation module includes: a daily data collection unit, a cloud server and an automatic watering unit.

[0023] The current day data collection unit is used to collect current day environmental data and current day cucumber image data and send them to the cloud server.

[0024] The cloud server is used to input the environmental data of the day and the cucumber image data of the day into the multimodal information fusion irrigation model, obtain the irrigation amount of the day and send it to the automatic irrigation unit.

[0025] The automatic watering unit is used for automatically watering according to the watering amount of the day.

[0026] In a possible implementation, the daily data collection unit includes: a temperature, humidity and light sensor and a camera.

[0027] In a possible implementation, the automatic irrigation unit includes: a control subunit, a communication subunit, a relay, a water pump, a flow meter, and a soil moisture sensor.

[0028] The greenhouse cucumber irrigation system based on multimodal information fusion in this application has the following advantages:

[0029] By combining cucumber growth index monitoring, image instance segmentation reasoning, multimodal information fusion irrigation model and automatic irrigation implementation module, the dynamic irrigation needs of cucumbers throughout the entire growth period can be met and the irrigation accuracy can be improved.

[0030] The proposed multimodal information fusion irrigation model combines historical environmental data, historical cucumber image data, and cucumber water requirement strategies to train a deep learning model and achieve accurate prediction of irrigation amount.

[0031] The proposed deep learning model adopts the iTransformer time series model, which has the ability to process multimodal sequence data and improves the training effect.

[0032] The proposed instance segmentation reasoning unit is trained with YOLO11s-seg weights, which improves the accuracy of identifying different parts of cucumber plants, i.e., the accuracy of extracting image feature data.

[0033] The proposed automatic watering implementation module includes: daily data collection unit, cloud server and automatic watering unit. The setting of cloud server makes the system highly scalable and maintainable.

[0034] The proposed automatic irrigation unit includes: a control subunit, a communication subunit, a relay, a water pump, a flow meter and a soil moisture sensor. The setting of the flow meter and the soil moisture sensor can monitor the cumulative output water volume and the real-time soil moisture to ensure accurate control of irrigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 A module schematic diagram of a greenhouse cucumber irrigation system based on multimodal information fusion provided in an embodiment of the present application. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0038] like Figure 1 As shown, the embodiment of the present application provides a greenhouse cucumber irrigation system based on multimodal information fusion, including: a cucumber water requirement strategy module, an irrigation model construction module and an automatic irrigation implementation module.

[0039] The cucumber water requirement strategy module is used to monitor cucumber growth indicators and perform image instance segmentation reasoning to obtain cucumber phenotypic indicators and image reasoning results, and perform time scale analysis based on the cucumber phenotypic indicators and the image reasoning results to obtain the cucumber water requirement strategy.

[0040] The irrigation model construction module is used to train the deep learning model according to the historical environmental data, the historical cucumber image data and the cucumber water demand strategy to obtain a multimodal information fusion irrigation model.

[0041] The automatic watering realization module is used to collect the environmental data of the day and the cucumber image data of the day and input them into the multimodal information fusion watering model to obtain the watering amount of the day, and automatically water according to the watering amount of the day.

[0042] Exemplarily, the cucumber water requirement strategy module includes: a growth index monitoring unit, an instance segmentation reasoning unit and a time scale parsing unit.

[0043] The growth index monitoring unit is used to monitor the growth index of cucumber and obtain the phenotypic index of cucumber.

[0044] The instance segmentation reasoning unit is used to perform image instance segmentation reasoning to obtain an image reasoning result.

[0045] The cucumber phenotypic index and the image reasoning result correspond in time scale.

[0046] The time scale analysis unit is used to perform time scale analysis according to the cucumber phenotypic index and the image reasoning result to obtain the cucumber water requirement strategy.

[0047] Specifically, in this embodiment, the growth index monitoring unit first conducts a greenhouse cucumber test under different irrigation levels: different irrigation level test groups are set up in the greenhouse, 125% Etc, 100% Etc, 75% Etc, (Etc represents reference evapotranspiration), and the growth, yield and quality of the greenhouse cucumbers in each test group are monitored.

[0048] Comprehensive evaluation of growth, yield and quality data: Through comprehensive evaluation of the growth indicators (such as plant height and leaf area), yield data (such as single plant yield and total yield) and quality indicators (such as sugar content and vitamin content) of cucumbers in each experimental group, weighted scoring or principal component analysis methods are used to determine the optimal irrigation level.

[0049] In this embodiment, the test results show that the greenhouse cucumber under the irrigation level of 100% Etc has the best comprehensive growth condition. Therefore, the cucumber under this irrigation level is selected as the standard, and the phenotypic index is continuously monitored, including the daily plant height growth, the number of leaves, the development of flowers and fruits, etc., to obtain the cucumber phenotypic index.

[0050] Exemplarily, performing time scale analysis based on the cucumber phenotypic indicators and the image reasoning results to obtain the cucumber water requirement strategy includes: performing principal component analysis dimensionality reduction and smoothing filtering on the cucumber phenotypic indicators and the image reasoning results on a time scale to obtain the relationship between the variation range of the cucumber throughout the growth period and the water requirement, that is, the cucumber water requirement strategy.

[0051] Specifically, in this embodiment, the time scale analysis unit associates the continuously monitored cucumber phenotypic indicators with the water requirement and image reasoning results on the corresponding time scale, and extracts the main influencing factors using principal component analysis. The difference between the principal component analysis values ​​representing the growth of cucumbers on two consecutive days is calculated to reflect the dynamic changes in cucumber growth, and then the difference sequence is smoothed and filtered to obtain the relationship between the change amplitude of the cucumber throughout its growth period and the water requirement, that is, the cucumber water requirement strategy.

[0052] Exemplarily, the smoothing filter includes: moving average smoothing technology, exponential smoothing technology, polynomial regression smoothing technology, LOESS smoothing technology, Savitzky-Golay filtering smoothing technology, Gaussian smoothing technology and median filtering smoothing technology.

[0053] Specifically, in this embodiment, moving average smoothing technology, exponential smoothing technology, polynomial regression smoothing technology, LOESS smoothing technology, Savitzky-Golay filtering smoothing technology, Gaussian smoothing technology and median filtering smoothing technology are used to smooth the difference sequence respectively, and the values ​​calculated by these technologies are averaged as the final change range of the cucumber during the entire growth period.

[0054] In a possible embodiment, phenotypic indicators are monitored for 90 consecutive days, and the plant height, leaf area and irrigation amount of cucumbers are recorded daily. Through principal component analysis, it is found that the contribution rate of the first principal component of plant height and leaf area is more than 80%. The difference in principal component values ​​of two consecutive days is calculated, such as the difference between the nth day and the n-1th day. The difference sequence is filtered by a 7-day moving average smoothing technique to obtain a smooth growth change curve. According to the changing trend of the curve, the relationship between the change amplitude and water requirement of the cucumber during the whole growth period is obtained.

[0055] Exemplarily, the irrigation model construction module includes: a feature extraction model construction unit and a deep learning model construction unit.

[0056] The feature extraction model construction unit is used to use the historical cucumber image data to train the instance segmentation reasoning unit in the cucumber water demand strategy module to obtain a feature extraction model, and input the historical cucumber image data into the feature extraction model to obtain image feature data.

[0057] The deep learning model construction unit is used to align the historical environmental data and the image feature data in time series as input features, and use the cucumber water requirement strategy as the target feature to train the deep learning model to obtain a multimodal information fusion irrigation model.

[0058] Specifically, in this embodiment, the historical environmental data includes the daily average values ​​of the historical temperature, humidity and light intensity in the greenhouse, and the historical environmental data can reflect the influence of environmental factors on the growth of cucumbers. The historical cucumber image data is obtained by timed shooting with a camera installed in the greenhouse. In this embodiment, the historical cucumber image data is also manually annotated to mark key parts such as leaves, flowers, terminal buds, and fruits.

[0059] Exemplarily, the instance segmentation reasoning unit is trained using YOLO11s-seg weights.

[0060] Specifically, in this embodiment, YOLO11s-seg weight is used as the training weight of the instance segmentation reasoning unit, and the instance segmentation reasoning unit is trained using the manually annotated historical cucumber image data, and the training parameters are set as follows: batch size (batch) is 1, image size (imgsz) is 2048, and the number of training rounds (epoch) is 150. The feature extraction model finally obtained can accurately segment various parts of the cucumber plant, and count the number of pixels of each target (leaves, flowers, fruits, etc.) obtained by segmentation as the feature data of the image, reflecting the growth status of the cucumber.

[0061] Exemplarily, the deep learning model adopts the iTransformer time series model.

[0062] Specifically, the iTransformer time series model has the ability to process multimodal sequence data. In this embodiment, the historical environmental data (including the daily average of historical temperature, humidity and light intensity) and the image feature data (including the number of pixels of leaves, flowers, fruits, etc.) are aligned in time series to form a multimodal input feature matrix as the input feature of the iTransformer time series model, and the cucumber water demand strategy is used as the target feature. The output of the iTransformer time series model is the predicted irrigation amount, that is, the daily irrigation amount. Use an appropriate loss function (such as mean square error) and optimization algorithm (such as Adam optimizer) to iterate training until the model converges.

[0063] In this example, the hyperparameters of the iTransformer time series model are as follows:

[0064] The model dimension is 64, the number of multi-head attention heads is 8, the dropout rate is 0.1, the number of encoder blocks is 2, the number of decoder blocks is 1, the fully connected network dimension is 16, the attention factor is 5, the activation function uses the gelu activation function, the channel independence is 0 (channel dependence), the parameter in the time block is 5, the startup parameter is 6, and distillation is enabled.

[0065] In this embodiment, data from the past year was used during the training of the iTransformer time series model, and a total of 5 groups of cucumbers and 1 group of cucumber average values ​​were continuously monitored, including 1,272 groups of environmental data and corresponding image features.

[0066] Exemplarily, the automatic watering implementation module includes: a daily data collection unit, a cloud server and an automatic watering unit.

[0067] The current day data collection unit is used to collect current day environmental data and current day cucumber image data and send them to the cloud server.

[0068] The cloud server is used to input the environmental data of the day and the cucumber image data of the day into the multimodal information fusion irrigation model, obtain the irrigation amount of the day and send it to the automatic irrigation unit.

[0069] The automatic watering unit is used for automatically watering according to the watering amount of the day.

[0070] Specifically, in this embodiment, the instance segmentation reasoning unit, the time scale analysis unit, the feature extraction model construction unit and the deep learning model construction unit are all arranged on the cloud server, and the growth indicator monitoring unit is communicatively connected with the cloud server.

[0071] Exemplarily, the daily data collection unit includes: a temperature, humidity and light sensor and a camera.

[0072] Exemplarily, the automatic irrigation unit includes: a control subunit, a communication subunit, a relay, a water pump, a flow meter and a soil moisture sensor.

[0073] Specifically, in this embodiment, the temperature, humidity and light sensors, the camera, the relay, the water pump, the flow meter and the soil moisture sensor are all electrically connected to the control subunit, and the control subunit is communicatively connected to the cloud server through the communication subunit.

[0074] Specifically, in this embodiment, the temperature, humidity and light sensor adopts a temperature, humidity and light intensity sensor with an RS485 interface to monitor the environmental conditions in the greenhouse, that is, the environmental data of the day, in real time, and send it to the control subunit (this embodiment adopts the STM32 controller). The cucumber image data of the day collected by the camera is also sent to the control subunit. The control subunit sends the environmental data of the day and the cucumber image data of the day to the cloud server through the communication subunit (this embodiment adopts a 4G module, and the 4G module adopts the MQTT protocol).

[0075] After receiving the environmental data and cucumber image data of the day, the cloud server calls the feature extraction model and the multimodal information fusion irrigation model to extract features and output the irrigation amount of the day, and sends the irrigation amount of the day to the control subunit.

[0076] The control subunit controls the relay to turn on the water pump for irrigation. At the same time, the cumulative output water volume and the real-time soil moisture (water content) are monitored through the flow meter (this embodiment uses a water meter with an RS485 interface) and the soil moisture sensor. When the soil moisture reaches 70%, the control subunit controls the relay to turn off the water pump. When the soil moisture drops to 60%, the control subunit controls the relay to turn on the water pump again. The above process is repeated until the cumulative output water volume reaches the irrigation volume for the day.

[0077] In a possible embodiment, if a water pump failure or water shortage occurs when the control subunit controls the relay to turn on the water pump for irrigation, the control subunit also sends an alarm message to the cloud server through the communication subunit and notifies the management personnel to handle it.

[0078] In a possible embodiment, a monitoring platform is built on the cloud server to display environmental data, irrigation volume, soil moisture and other information in real time.

[0079] In a possible embodiment, the cloud server stores historical data in a database to facilitate subsequent data mining and model optimization.

[0080] In a possible embodiment, the greenhouse cucumber irrigation system of the present application is used to perform actual irrigation verification: one day the cloud server receives the environmental data of the day with a temperature of 25°C, a humidity of 60%, and a light intensity of 20,000 Lux, as well as the corresponding cucumber image of the day. After calculation by the feature extraction model and the multimodal information fusion irrigation model, it is concluded that the amount of water required for the day is 50 liters. After receiving the instruction, the STM32 controller starts the water pump for irrigation. The soil moisture sensor shows that the soil moisture gradually increases from 55%. When it reaches 70%, the control subunit controls the relay to turn off the water pump. Subsequently, as the cucumber plants absorb water, the soil moisture drops to 60%. The control subunit controls the relay to turn on the water pump again until the cumulative output water volume monitored by the water meter reaches 50 liters.

[0081] The embodiment of the present application combines cucumber growth index monitoring, image instance segmentation reasoning, multimodal information fusion irrigation model and automatic irrigation implementation module to meet the dynamic irrigation needs of cucumbers throughout the entire growth period and improve the irrigation accuracy.

[0082] The proposed multimodal information fusion irrigation model combines historical environmental data, historical cucumber image data, and cucumber water requirement strategies to train a deep learning model and achieve accurate prediction of irrigation amount.

[0083] The proposed deep learning model adopts the iTransformer time series model, which has the ability to process multimodal sequence data and improves the training effect.

[0084] The proposed instance segmentation reasoning unit is trained with YOLO11s-seg weights, which improves the accuracy of identifying different parts of cucumber plants, i.e., the accuracy of extracting image feature data.

[0085] The proposed automatic watering implementation module includes: daily data collection unit, cloud server and automatic watering unit. The setting of cloud server makes the system highly scalable and maintainable.

[0086] The proposed automatic irrigation unit includes: a control subunit, a communication subunit, a relay, a water pump, a flow meter and a soil moisture sensor. The setting of the flow meter and the soil moisture sensor can monitor the cumulative output water volume and the real-time soil moisture to ensure accurate control of the irrigation.

[0087] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0088] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. Greenhouse cucumber irrigation system based on multimodal information fusion, characterized in that: include: Cucumber water demand strategy module, irrigation model construction module and automatic irrigation implementation module; The cucumber water requirement strategy module is used to monitor cucumber growth indicators and perform image instance segmentation reasoning to obtain cucumber phenotypic indicators and image reasoning results, and perform time scale analysis based on the cucumber phenotypic indicators and the image reasoning results to obtain the cucumber water requirement strategy; The irrigation model construction module is used to train the deep learning model according to the historical environmental data, the historical cucumber image data and the cucumber water demand strategy to obtain a multimodal information fusion irrigation model; The automatic watering realization module is used to collect the environmental data of the day and the cucumber image data of the day and input them into the multimodal information fusion watering model to obtain the watering amount of the day, and automatically water according to the watering amount of the day.

2. The greenhouse cucumber irrigation system based on multimodal information fusion according to claim 1 is characterized in that: The cucumber water demand strategy module includes: a growth index monitoring unit, an instance segmentation reasoning unit and a time scale analysis unit; The growth index monitoring unit is used to monitor the growth index of cucumber and obtain the phenotypic index of cucumber; The instance segmentation reasoning unit is used to perform image instance segmentation reasoning to obtain an image reasoning result; The cucumber phenotypic index and the image reasoning result correspond to each other in time scale; The time scale analysis unit is used to perform time scale analysis according to the cucumber phenotypic index and the image reasoning result to obtain the cucumber water requirement strategy.

3. The greenhouse cucumber irrigation system based on multimodal information fusion according to claim 2 is characterized in that: The method of performing time scale analysis based on the cucumber phenotypic indicators and the image reasoning results to obtain the cucumber water requirement strategy includes: performing principal component analysis dimensionality reduction and smoothing filtering on the cucumber phenotypic indicators and the image reasoning results on a time scale to obtain the relationship between the variation range of the cucumber throughout the growth period and the water requirement, that is, the cucumber water requirement strategy.

4. The greenhouse cucumber irrigation system based on multimodal information fusion according to claim 3 is characterized in that: The smoothing filter includes: moving average smoothing technology, exponential smoothing technology, polynomial regression smoothing technology, LOESS smoothing technology, Savitzky-Golay filtering smoothing technology, Gaussian smoothing technology and median filtering smoothing technology.

5. The greenhouse cucumber irrigation system based on multimodal information fusion according to claim 2 is characterized in that: The irrigation model construction module includes: a feature extraction model construction unit and a deep learning model construction unit; The feature extraction model construction unit is used to train the instance segmentation reasoning unit in the cucumber water demand strategy module using the historical cucumber image data to obtain a feature extraction model, and input the historical cucumber image data into the feature extraction model to obtain image feature data; The deep learning model construction unit is used to align the historical environmental data and the image feature data in time series as input features, and use the cucumber water requirement strategy as the target feature to train the deep learning model to obtain a multimodal information fusion irrigation model.

6. The greenhouse cucumber irrigation system based on multimodal information fusion according to claim 5 is characterized in that: The instance segmentation reasoning unit is trained using YOLO11s-seg weights.

7. The greenhouse cucumber irrigation system based on multimodal information fusion according to claim 5 is characterized in that: The deep learning model adopts the iTransformer time series model.

8. The greenhouse cucumber irrigation system based on multimodal information fusion according to claim 1 is characterized in that: The automatic watering realization module includes: a daily data collection unit, a cloud server and an automatic watering unit; The daily data collection unit is used to collect the daily environmental data and the daily cucumber image data and send them to the cloud server; The cloud server is used to input the environmental data of the day and the cucumber image data of the day into the multimodal information fusion irrigation model, obtain the irrigation amount of the day and send it to the automatic irrigation unit; The automatic watering unit is used for automatically watering according to the watering amount of the day.

9. The greenhouse cucumber irrigation system based on multimodal information fusion according to claim 8, characterized in that: The daily data collection unit includes: a temperature, humidity and light sensor and a camera.

10. The greenhouse cucumber irrigation system based on multimodal information fusion according to claim 8, characterized in that: The automatic irrigation unit comprises: a control subunit, a communication subunit, a relay, a water pump, a flow meter and a soil moisture sensor.

Images