Strawberry greenhouse digital twinborn system based on YOLOv10 and decision tree algorithm and management method

Abstract

The invention relates to a strawberry greenhouse digital twinning system based on YOLOv10 and a decision tree algorithm and a management method, and belongs to the technical field of automatic cultivation. The environment detection system comprises an air temperature and humidity sensor, a soil temperature and humidity sensor, a soil nitrogen, phosphorus and potassium content detection sensor and a camera which are arranged in a greenhouse, the camera is used for detecting the growth condition of crops, the environment detection system communicates with the switch through an RS485 bus for data transmission, and the control system comprises a PC terminal. A production control algorithm based on YOLOv10 and a decision-making tree algorithm is arranged in the PC terminal, and the decision-making tree evaluates the health state and demand of strawberry growth in real time according to multi-dimensional data such as soil temperature and humidity, air temperature and humidity, illumination intensity and the like in combination with the characteristics of the strawberry growth stage, so as to determine whether watering and fertilization are needed or not; when the fertilizer needs to be watered, the watering and fertilizing time is calculated by utilizing a designed linear weighted sum formula, so that the effect of accurate watering and fertilizing is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of twin systems, and in particular to a strawberry greenhouse digital twin system and a management method based on YOLOv10 and a decision tree algorithm. Background Art

[0002] Strawberry is a fruit with demanding growing conditions. Traditional strawberry greenhouses usually rely on manual monitoring and management. Although this method is experienced, it has limitations in real-time response and precise control. Intelligent strawberry greenhouses can monitor and record key environmental parameters such as soil temperature and humidity, air temperature and humidity, and light intensity in real time by connecting to various sensors and control systems. These data can not only help farmers grasp the changes in the internal environment of the greenhouse more accurately, but also use data to control watering and other operations. However, whether it is manual or intelligent strawberry greenhouses, these traditional methods can only reduce the burden of strawberry production but cannot guide production. At the same time, the existing intelligent greenhouses cannot observe the growth of crops in real time. Under such conditions, simply using environmental data to control watering and fertilization is often inaccurate. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a digital twin system and management method for a strawberry greenhouse based on YOLOv10 and a decision tree algorithm, and to use the YOLOv10 algorithm to analyze the monitoring video data inside the strawberry greenhouse in real time, so as to quickly and accurately identify the growth status of strawberry plants. Based on the collected environmental data and the results of detecting the growth of strawberries by YOLOv10, the decision tree evaluates the health status and demand of strawberry growth in real time according to multidimensional data such as soil temperature and humidity, air temperature and humidity, and light intensity, combined with the characteristics of the strawberry growth stage, so as to decide whether watering and fertilization are needed. When watering and fertilization are needed, the designed linear weighted sum formula is used to calculate the time of watering and fertilization, so as to achieve the effect of accurate watering and fertilization.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is: A strawberry greenhouse digital twin system based on YOLOv10 and decision tree algorithm is characterized in that it includes an environment detection system and a control system. The environment detection system includes an air temperature and humidity sensor, a soil temperature and humidity sensor, a soil nitrogen, phosphorus and potassium content detection sensor and a camera arranged in the greenhouse. The camera is used to detect the growth of crops. The environment detection system communicates with a switch through an RS485 bus to transmit data. The control system includes a PC end, and a production control algorithm based on YOLOv10 and a decision tree algorithm is arranged in the PC end. The PC end is connected to the switch for communication and transmits a control signal to a relay group through the RS485 bus. The watering motor and the fertilizing motor are controlled by the relay group. The environment detection system connects the data to the PC end and then transmits it to the digital twin system through the UDP protocol.

[0005] The YOLOv10 in the PC mentioned above is used to identify and locate crop plants in the greenhouse environment.

[0006] The above-mentioned YOLOv10 simultaneously identifies multiple crop plants and is used to monitor and evaluate the growth status of the plants.

[0007] The above YOLOv10 is trained using a prepared crop dataset.

[0008] The above decision tree algorithm classifies input data through decision nodes and leaf nodes, and is used for watering and fertilizing according to various environmental and crop characteristics.

[0009] Using the above-mentioned management method of the digital twin system of the strawberry greenhouse based on YOLOv10 and decision tree algorithm, the YOLOv10 steps are as follows: Step 1, collect and organize the data of the growth stage of crops, and the crop growth stage includes 7 conditions: healthy leaves, flowering, fruiting, maturity, powdery mildew, leaf spot and fertilizer deficiency; Step 2: Clean and preprocess the data collected in step 1, remove outliers and missing values, and perform standardization to improve the consistency and reliability of the data to form a data set; Step 3: Divide the data set in step 2 into a training set and a test set according to the set ratio, and use the training set data to train the YOLO Vv10 classification model. During the training process, the model parameters are tuned by the cross-validation method to obtain the best model performance. Step 4. After the training in step 3 is completed, the trained model is evaluated using the test set data to verify the accuracy and generalization ability of the model; Step 5: Use the trained model to predict the growth status of crops based on the current crop environmental data.

[0010] The above decision tree algorithm outputs control on whether watering is needed, watering time, whether to fertilize and fertilize time based on the input characteristics of soil temperature, soil moisture, air temperature and humidity, soil nitrogen content, soil phosphorus content, soil potassium content and crop growth stage.

[0011] The above decision tree algorithm steps are as follows: The decision tree model formula is expressed as: ; in: The model input The prediction results, is the number of leaf nodes, It is The area of ​​leaf nodes; It is The predicted value of a leaf node, is the indicator function, when belong hour, is 1, otherwise it is 0; Treat whether watering and fertilization are needed as a binary classification problem: ; ; in, It is the predicted result of whether to water, 0 means no watering, 1 means watering is needed; It is the result of predicting whether fertilization is needed, 0 means no fertilization, 1 means fertilization is needed; is the input feature vector: ; in is the soil temperature, is soil moisture, is the air temperature, Air humidity, Soil nitrogen content, Soil phosphorus content, Potassium content in soil, It is an indicator of the growth stage of crops; The linear weighted sum formula is used to calculate the watering time and fertilization time: The formula for smart watering: ; T is the final calculated watering time, in seconds; , , , They are the weight coefficients of soil moisture, temperature, growth stage and disease status, which are changed according to empirical data; , , , They are the contribution values ​​to watering time, which are adjusted according to environmental data; the contribution value calculation formula is: ; ; ; ； in , The value changes according to the real-time soil temperature. ,and The value changes according to the results of the YOLOv10 algorithm monitoring in the camera, which shows the growth status of the crops; Intelligent fertilization formula: ; ; ； is the amount of fertilizer applied, , , are the weights of nitrogen, phosphorus and potassium, , , They are the demand contribution values ​​of nitrogen, phosphorus and potassium, , is the weight of growth stage and pests and diseases, , are the contribution values ​​of fertilizer requirements for growth stage and disease status respectively.

[0012] The above-mentioned crop is strawberry.

[0013] The strawberry greenhouse digital twin system and management method based on YOLOv10 and decision tree algorithm provided by the present invention have the following beneficial effects: 1. Real-time monitoring and precise control: The present invention uses the YOLOv10 algorithm to perform real-time analysis on the monitoring video or image data inside the strawberry greenhouse, and can quickly and accurately identify the growth status of strawberry plants. Based on environmental data and strawberry growth status, the decision tree algorithm is used for intelligent decision-making, and the linear weighted sum formula is used to calculate the watering and fertilization time, so as to realize precision agriculture and optimize resource utilization.

[0014] 2. Improve production efficiency and quality: Through the automated management system, the present invention reduces the workload of manual monitoring and management and improves production efficiency. The combination of the YOLOv10 algorithm and the decision tree algorithm makes the prediction of strawberry growth status more accurate, thereby improving the yield and quality of strawberries.

[0015] 3. Intelligent management strategy: Based on multi-dimensional environmental data and combined with the growth stage characteristics of strawberries, the decision tree algorithm can generate intelligent management strategies to guide watering and fertilization operations in the actual production process. This intelligent management not only improves the management efficiency of strawberry greenhouses, but also effectively reduces resource waste and environmental impact.

[0016] 4. Real-time display of the twin system: The twin system designed by the present invention adopts U3D software to display the data of the environmental detection sensor in real time through the UI interface. According to the strawberry growth conditions identified by the camera and YOLOv10 algorithm, the twin system can accurately replace the strawberry model to achieve synchronous display of the virtual scene and the actual growth status.

[0017] 5. Multiple control modes: The present invention provides two control modes: automatic mode and manual mode. In automatic mode, the system automatically implements watering and fertilizing operations according to the output results of the algorithm; in manual mode, users can control by clicking buttons in the virtual scene, flexibly responding to different management needs.

[0018] 6. Automatic optimization algorithm: The present invention can not only collect data in real time but also provide the collected data to the decision tree algorithm, and can continuously optimize the decision tree algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 This is a framework diagram of the digital twin system of the strawberry greenhouse of the present invention; Figure 2 This is the control operation flow of the control system of the present invention. DETAILED DESCRIPTION

[0020] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings and embodiments.

[0021] Example: Environmental detection system: There are many data that need to be detected in the strawberry greenhouse scene, such as greenhouse air temperature and humidity, soil temperature and humidity, nitrogen, phosphorus and potassium content, and crop growth. The equipment used includes air temperature and humidity sensors, soil temperature and humidity sensors, nitrogen, phosphorus and potassium content detection sensors, and cameras for detecting crop growth. Except for the camera, these data are transmitted using the RS485 bus, and the camera data is transmitted using Ethernet. Connect these data to the host computer and then transmit them to my digital twin system through the UDP protocol.

[0022] Control system: The control required in this design includes watering and fertilizing, and the fertilization uses a high-concentration fertilizer liquid. Watering and fertilization are controlled by turning the motor on and off. For the motor that needs to be controlled, a relay is connected between the device end and the computer host end to achieve it. The virtual scene can send instructions to the relay through the UDP protocol to control the motor to control watering and fertilization. The control system of this system includes two modes: automatic mode and manual mode. Its networking method is as follows Figure 1 as shown in .

[0023] YOLOv10 predicts strawberry growth status: YOLOv10 is an efficient target detection algorithm suitable for identifying and locating strawberry plants in greenhouse environments. The algorithm achieves accurate detection of strawberry growth in images through global visual perception and single forward reasoning. YOLOv10 can simultaneously identify multiple strawberries, provide their location coordinates and bounding box information, and thus help agricultural producers monitor and evaluate the growth status of plants. This design trains the model with a prepared strawberry dataset. First, collect and organize data on the growth stages of strawberries, which include healthy leaves, flowering, fruiting, maturity, powdery mildew, leaf spot, and lack of fertilizer. These data are cleaned and preprocessed to remove outliers and missing values, and are standardized to ensure data consistency and reliability. Next, the dataset is divided into a training set and a test set, usually divided in proportion, such as 80% for training and 20% for testing. Use the training set data to train the YOLOVv10 classification model. During the training process, the model parameters are tuned by cross-validation and other methods to obtain the best model performance. After the training is completed, the trained model is evaluated using the test set data to verify the accuracy and generalization ability of the model. By comparing the actual growth situation with the model prediction results, it is ensured that the model can effectively predict the growth status of strawberries. Finally, the model predicts the growth status of strawberries. Based on the growth status results and environmental data, the decision tree algorithm is used to guide management decisions in the actual production process to improve the yield and quality of strawberries.

[0024] The decision tree algorithm determines whether to water: The decision tree model is a commonly used machine learning algorithm that can classify or regress input data through a series of decision nodes and leaf nodes. In the digital twin strawberry greenhouse in this design, the decision tree model is used to water, fertilize, and implement the corresponding operation time according to a variety of environmental and crop characteristics. The model is based on the following input features: soil temperature, soil moisture, air temperature, air humidity, soil nitrogen content, soil phosphorus content, soil potassium content, strawberry growth stage. The output of the decision tree model includes four main parts: whether watering is needed, watering time, whether to fertilize, and fertilization time.

[0025] The general formula of the decision tree model can be expressed as: ; in: The model input The prediction results, is the number of leaf nodes, It is The area of ​​leaf nodes. It is The predicted value of a leaf node, is the indicator function, when belong hour, is 1 if the value is true, otherwise it is 0.

[0026] The decision tree model in this design is used to predict whether watering and fertilization are needed, which is a binary classification problem: ; ; in, is the predicted result of whether to water (0 means no watering, 1 means watering is needed), It is the predicted result of whether fertilization is needed (0 means no fertilization, 1 means fertilization is needed). is the input feature vector: ; in is the soil temperature, is soil moisture, is the air temperature, Air humidity, Soil nitrogen content, Soil phosphorus content, Potassium content in soil, It is an indicator of the growth stage of strawberry.

[0027] The linear weighted sum formula is used to calculate the watering time and fertilization time: The formula for smart watering: ; T is the watering time (s) and the final calculated watering time result. , , , are the weight coefficients for soil moisture, temperature, growth stage, and disease status, respectively, which can be changed based on empirical data. , , , They are the contribution values ​​to the watering time, which can be adjusted according to environmental data. The contribution value calculation formula of this design is: ; ; ; ； in , The value of changes according to the real-time soil temperature. ,and The value changes according to the results of the yolo10 algorithm monitoring of the camera, which shows the growth status of the strawberry. Intelligent fertilization formula: ; ; ； is the amount of fertilizer applied, , , are the weights of nitrogen, phosphorus and potassium, , , They are the demand contribution values ​​of nitrogen, phosphorus and potassium, , is the weight of growth stage and pests and diseases, , The contribution values ​​of fertilizer requirements for growth stage and disease status are calculated as the difference between the current nitrogen, phosphorus and potassium content and the ideal content.

[0028] The present invention first collects multidimensional environmental data (such as soil temperature, soil humidity, air temperature, air humidity, soil nitrogen, phosphorus and potassium content) and strawberry growth status data in a strawberry greenhouse (by taking images with a camera and using the YOLOv10 algorithm to mark the strawberry growth stage), and then cleans, pre-processes and standardizes the data. Next, the data set is divided into a training set and a test set, and the decision tree model is trained using the training set, and the model parameters are tuned by cross-validation to obtain the best performance. After the training is completed, the model is evaluated using the test set to ensure its accuracy and generalization ability. Finally, the trained model is deployed to the system, and the environmental data and growth status data are received in real time. The decision tree algorithm and the linear weighted sum formula are used to generate decisions on whether watering is needed and the watering time is controlled, and whether fertilization is applied and the fertilization time is controlled, and the decision is automatically executed by the control system to achieve accurate watering and fertilization operations, optimize the strawberry growth environment, and improve yield and quality.

[0029] Twin system design: In the present invention, the twin system is designed with U3D software, and the data received by the sensors of various environmental detections are displayed in the twin system through the UI interface, and the monitoring video of the camera is also displayed in real time in the virtual scene. There are 7 strawberry twin models, and the strawberry model is replaced accordingly according to the different strawberry growth conditions identified by the camera plus the YOLOv10 algorithm, so that the growth of strawberries can be accurately viewed in the twin scene. The operation of the control system provides button operation in the twin scene. The system provides automatic mode and manual mode. The automatic model automatically realizes watering and other operations according to the above-mentioned decision tree output results and the linear weighted and formula calculation results. Manual control uses the button in the virtual scene to control watering, etc.

[0030] like Figure 1As shown in the figure, it is the framework diagram of the digital twin system of the strawberry greenhouse, in which the sensors for simply detecting the environment (air temperature and humidity sensor, soil temperature and humidity sensor, nitrogen, phosphorus and potassium content detection sensor) are directly connected to the RS485 hub via the RS485 bus protocol. For the motor to be controlled, a relay is added between the motor and the hub to control the power on and off of the motor. These are all RS485 bus devices that are finally aggregated to a hub, and the hub is connected to an RS485 to Ethernet converter to aggregate to the switch. The camera data goes directly to the switch via the Ethernet bus. In this way, all the data is finally aggregated to the switch, and then the switch is connected to the PC. The PC transmits all the data to the data forwarding server, and then transmits the camera data to the YOLOv10 algorithm to monitor the growth status of the strawberry, and judges whether the results of the growth status monitoring have changed. If there is a change, the change result will be sent to the data forwarding server, and the monitoring results and environmental data are used as the input of the decision tree algorithm to calculate whether to water and apply fertilizer. When the decision tree algorithm has instructions for watering or fertilizing, the environmental data and the strawberry growth status results calculated by the algorithm are used as the input of the linear weighted formula to calculate the time for watering and fertilizing, and finally the calculated time for watering and fertilizing is sent to the data forwarding server. The data forwarding server then packages the strawberry model change instructions, watering and fertilizing instructions, watering and fertilizing time, and hardware data into the required format and transmits them to the twin system through the UDP protocol. The twin system updates the strawberry model in the twin scene and the environmental data of the strawberry greenhouse displayed on the UI in real time based on the server data. At the same time, when there is a control instruction, the twin system sends a control instruction to the hardware.

[0031] like Figure 2 As shown in the figure, it is the control operation process of the control system, which is mainly divided into two modes, one is automatic mode and the other is manual mode. The automatic mode is to determine whether the decision tree algorithm has instructions for watering or fertilizing. When there is an instruction, the strawberry growth status, the result of the decision tree algorithm, and the environmental data are used as the input of the linear weighted formula to calculate the contribution value of each indicator of watering and fertilizing, and then calculate the time of watering and fertilizing, and then send the output result to the data forwarding server. The server then sends the corresponding instructions to the twin system. The twin system then sends the instructions of opening and closing the relay to the data forwarding server at a specific time according to the control instructions and time instructions. The server then sends the data to the PC end, and then sends the instructions to the corresponding relay according to the previous connection method to control the corresponding motor. In manual mode, the client of the twin system directly sends the instructions to be sent to the algorithm server by clicking the button in the twin system and then performs the above operations.

Claims

1. A digital twin system for strawberry greenhouse based on YOLOv10 and decision tree algorithm, characterized in that: It includes an environmental detection system and a control system. The environmental detection system includes air temperature and humidity sensors, soil temperature and humidity sensors, soil nitrogen, phosphorus and potassium content detection sensors and cameras installed in the greenhouse. The camera is used to detect the growth of crops. The environmental detection system communicates with the switch through the RS485 bus for data transmission. The control system includes a PC terminal, which has a production control algorithm based on YOLOv10 and decision tree algorithm. The PC terminal is connected to the switch and transmits the control signal to the relay group through the RS485 bus. The watering motor and fertilizing motor are controlled by the relay group. The environmental detection system connects the data to the PC terminal and then transmits it to the digital twin system through the UDP protocol.

2. A strawberry greenhouse digital twin system based on YOLOv10 and decision tree algorithm according to claim 1, characterized in that, The YOLOv10 in the PC is used to identify and locate crop plants in a greenhouse environment.

3. A strawberry greenhouse digital twin system based on YOLOv10 and decision tree algorithm according to claim 2, characterized in that, The YOLOv10 simultaneously identifies multiple crop plants and is used to monitor and evaluate the growth status of the plants.

4. A strawberry greenhouse digital twin system based on YOLOv10 and decision tree algorithm according to claim 3, characterized in that, The YOLOv10 model is trained using a prepared crop dataset.

5. A strawberry greenhouse digital twin system based on YOLOv10 and decision tree algorithm according to claim 4, characterized in that, The decision tree algorithm classifies input data through decision nodes and leaf nodes, and is used for watering and fertilizing according to various environmental and crop characteristics.

6. A management method for a strawberry greenhouse digital twin system based on YOLOv10 and a decision tree algorithm according to claim 5, characterized in that: The steps of YOLOv10 are as follows: Step 1: Collect and organize data on the growth stages of crops. The growth stages of crops include healthy leaves, flowering, fruiting, maturity, powdery mildew, leaf spot and fertilizer deficiency. Step 2: Clean and preprocess the data collected in step 1, remove outliers and missing values, and perform standardization to improve the consistency and reliability of the data to form a data set; Step 3: Divide the data set in step 2 into a training set and a test set according to the set ratio, and use the training set data to train the YOLO Vv10 classification model. During the training process, the model parameters are tuned by the cross-validation method to obtain the best model performance. Step 4. After the training in step 3 is completed, the trained model is evaluated using the test set data to verify the accuracy and generalization ability of the model; Step 5: Use the trained model to predict the growth status of crops based on the current crop environmental data.

7. A management method for a strawberry greenhouse digital twin system based on YOLOv10 and decision tree algorithm according to claim 6, characterized in that, The decision tree algorithm outputs control on whether watering is needed, watering time, whether fertilization is needed and fertilization time based on soil temperature, soil moisture, air temperature and air humidity, soil nitrogen content, soil phosphorus content, soil potassium content and crop growth stage input characteristics.

8. A management method for a strawberry greenhouse digital twin system based on YOLOv10 and decision tree algorithm according to claim 7, characterized in that, The decision tree algorithm steps are as follows: The decision tree model formula is expressed as: ； in: The model input The prediction results, is the number of leaf nodes, It is The area of ​​leaf nodes; It is The predicted value of a leaf node, is the indicator function, when belong hour, is 1, otherwise it is 0; Treat whether watering and fertilization are needed as a binary classification problem: ； ； in, It is the predicted result of whether to water, 0 means no watering, 1 means watering is needed; It is the result of predicting whether fertilization is needed, 0 means no fertilization, 1 means fertilization is needed; is the input feature vector: ； in is the soil temperature, is soil moisture, is the air temperature, Air humidity, Soil nitrogen content, Soil phosphorus content, Potassium content in soil, It is an indicator of the growth stage of crops; The linear weighted sum formula is used to calculate the watering time and fertilization time: The formula for smart watering: ； T is the final calculated watering time, in seconds; , , , They are the weight coefficients of soil moisture, temperature, growth stage and disease status, which are changed according to empirical data; , , , They are the contribution values ​​to watering time, which are adjusted according to environmental data; the contribution value calculation formula is: ； ； ； ； in , The value changes according to the real-time soil temperature. ,and The value changes according to the results of the YOLOv10 algorithm monitoring in the camera, which shows the growth status of the crops; Intelligent fertilization formula: ； ； ； is the amount of fertilizer applied, , , are the weights of nitrogen, phosphorus and potassium, , , They are the demand contribution values ​​of nitrogen, phosphorus and potassium, , is the weight of growth stage and pests and diseases, , are the contribution values ​​of fertilizer requirements for growth stage and disease status respectively.

9. A management method for a strawberry greenhouse digital twin system based on YOLOv10 and decision tree algorithm according to claim 8, characterized in that, The crop is strawberry.

Images