Loquat orchard intelligent anti-freezing and rain-sheltering equipment based on multi-sensor fusion and self-adaptive control method

Abstract

The invention provides loquat orchard intelligent anti-freezing and rain-sheltering equipment based on multi-sensor fusion and a self-adaptive control method. The equipment comprises an anti-freezing and rain-sheltering support main body; the lead screws are arranged on two sides of the top of the support body and connected with output shafts of the driving motors; the telescopic driver is slidably arranged on the screw rod; the top storage rod is connected with an output shaft of the telescopic driver; and the film is arranged on the top storage rod. The side wall drivers are arranged on the two sides of the support body. The side accommodating rod is movably arranged on the bracket main body, is connected with an output shaft of the side wall driver and is also provided with a film; the balance weight rod is movably connected with the side face storage rod through a thin film. According to the method, environment data are collected through the equipment sensor array and the infrared thermal imager, a data set is formed by combining edge computing nodes, and model building and temperature gradient updating are supported. And a fuzzy PID algorithm of the control module optimizes control parameters of a driver and a motor according to a weather prediction result, so that the whole process is intelligent. Compared with manual operation, closed-loop response is realized in an emergency scene, and the protection efficiency and reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of anti-freezing and rain-avoidance equipment technology, and in particular to an intelligent anti-freezing and rain-avoidance equipment for loquat orchards based on multi-sensor fusion and an adaptive control method. Background Technology

[0002] Currently, most loquat orchard frost and rain protection equipment uses traditional manual operation, requiring manual on-site adjustments to the support structure and the covering or retraction of the film. For example, farmers need to manually drive the screw to raise and lower the top storage rod, adjust the unfolding angle of the side wall actuators, and fix the counterweight rod with physical locks, based on weather forecasts or experience. Such equipment is usually equipped with only a single type of sensor (such as a temperature or humidity probe), resulting in limited data acquisition dimensions and a lack of real-time data processing capabilities, leading to a lag in environmental response. Although some automated equipment achieves local drive motor control, the lack of coordination logic between actuators prevents dynamic parameter adjustment based on multi-source data, still requiring manual intervention to complete complex operations.

[0003] Existing technologies suffer from low efficiency due to manual operation. In emergency scenarios such as sudden frost or heavy rain, there is a significant time delay from decision-making to execution, which can easily cause the best protection window to be missed. Summary of the Invention

[0004] This invention aims to at least solve the technical problem of low efficiency in manual operation in the prior art, and innovatively proposes an intelligent anti-freezing and rain-avoidance device for loquat orchards and an adaptive control method based on multi-sensor fusion.

[0005] To achieve the above-mentioned objectives of this invention, this invention provides an intelligent anti-freezing and rain-avoidance device for loquat orchards based on multi-sensor fusion, the device comprising: The main body of the frost-proof and rain-proof support frame; A lead screw is located on both sides of the top of the antifreeze and rainproof bracket body and is connected to the output shaft of the drive motor. The telescopic actuator is slidably mounted on the lead screws on both sides of the top of the antifreeze and rainproof bracket body; The top storage rod is connected to the output shaft of the telescopic driver set on the lead screw on both sides of the top of the antifreeze and rainproof bracket body; A thin film is disposed on the top storage rod; Side wall actuators are located on both sides of the main body of the anti-freeze and rainproof bracket; A side storage rod is movably mounted on the main body of the antifreeze and rainproof bracket and connected to the output shaft of the side wall driver. The side storage rod is provided with a thin film. The counterweight rod is movably connected to the side storage rod via a membrane.

[0006] As an optional embodiment of the present invention, the device may further include: The control module contains a control algorithm for controlling the telescopic driver, the side wall driver, and the drive motor. A signal transmitter, connected to the control module, is used to send control signals to the control module.

[0007] On the other hand, the present invention also provides an intelligent anti-freezing and rain-avoidance adaptive control method for loquat orchards based on multi-sensor fusion, applied to the aforementioned equipment, the method comprising: S1. Based on the sensor array and infrared thermal imager deployed on the top of the anti-freezing and rain-sheltered support body, the edge computing node is used to collect real-time environmental data of the loquat orchard to form a multimodal perception dataset. S2. Based on the collected multimodal sensing dataset, construct a three-dimensional environmental model of the loquat orchard and dynamically update the canopy temperature gradient distribution; S3. Based on the three-dimensional environment model of the loquat orchard, a meteorological prediction model is constructed using an LSTM neural network. The model is trained by fusing historical meteorological data with real-time sensor data to generate frost / rainstorm prediction results for the next n hours. S4. Based on the prediction results and dual-mode control logic, the fuzzy PID algorithm is used to optimize the coordinated control parameters of the telescopic driver, side wall driver and drive motor, and generate the action command of the anti-freezing and rain-avoiding equipment. S5. Based on the generated motion commands, the drive motor drives the lead screw to simultaneously start the telescopic driver to adjust the top storage rod, and synchronously controls the side wall driver to adjust the side storage rod to unfold the film, and fixes the counterweight rod by electromagnet.

[0008] As another optional embodiment of the present invention, optionally, in step S2, constructing a three-dimensional environment model of the loquat orchard and dynamically updating the canopy temperature gradient distribution based on the collected multimodal sensing dataset includes: S201. Based on the data from the lidar and vision sensor deployed on the top of the anti-freezing and rain-sheltered bracket body 1, the ICP algorithm is used to perform spatiotemporal alignment of point cloud and image data, and the initial three-dimensional model is generated by semantic segmentation through the U-Net network. S202. Based on the initial three-dimensional model and the canopy surface temperature data collected by the infrared thermal imager, the temperature features are extracted using the dual-light fusion algorithm combined with the attention mechanism module. The transpiration cooling effect is calculated using the three-temperature method to generate a vertical temperature gradient distribution map of the canopy. S203. Based on the vertical temperature gradient distribution map of the tree canopy and combined with real-time meteorological data, the temperature field distribution in the initial three-dimensional model is dynamically adjusted, and the temperature gradient is smoothed by the Kalman filter algorithm to generate a three-dimensional environmental model of the loquat orchard.

[0009] As another optional embodiment of the present invention, optionally, in step S3, based on the three-dimensional environment model of the loquat orchard, a meteorological prediction model is constructed using an LSTM neural network, and the model is trained by fusing historical meteorological data with real-time sensor data to generate frost / rainstorm prediction results for the next n hours, including: S301. Based on the three-dimensional environment model of the loquat orchard and the multimodal perception dataset, the U-Net network is used to extract the spatial features of the tree canopy, and the LSTM network is combined to capture the n-hour time series patterns to generate multi-scale spatiotemporal feature vectors. S302. Based on multi-scale spatiotemporal feature vectors and real-time temperature data collected by an infrared thermal imager, a prediction model is constructed using a dual-layer LSTM network and an attention mechanism module, and basic prediction results are generated through a fully connected layer. S303. Based on the basic prediction results and the sliding window training strategy, the prediction model parameters are dynamically updated every x minutes. The Kalman filter algorithm is used to fuse historical meteorological data and real-time sensor data to generate a probability distribution map of frost / rainstorm for the next n hours.

[0010] As another optional embodiment of the present invention, optionally, in step S4, based on the prediction results and dual-mode control logic, the coordinated control parameters of the telescopic driver, sidewall driver, and drive motor are optimized using a fuzzy PID algorithm to generate the action command for the anti-freezing and rain-avoiding device, including: S401. Based on the frost / rainstorm prediction probability distribution map in the prediction results, use the fuzzy logic rule base to establish an environmental level classification standard and divide the prediction results into five warning levels: safe, attention, warning, danger, and emergency. S402. Dynamically adjust the PID controller parameters based on the warning level, optimize the Kp / Ki / Kd coefficients through a genetic algorithm to obtain optimized control parameters, and perform coordinated control of the telescopic actuator, side wall actuator and drive motor. S403: Based on the dual-mode control logic switching strategy, PID closed-loop control is used for adjustment in the basic mode, and fuzzy inference engine is enabled for rapid response in the emergency mode. S404. Generate a set of equipment action instructions based on the optimized control parameters, including specific control parameters for drive motor speed, lead screw stroke, telescopic actuator extension amount, and side wall actuator deployment height.

[0011] The beneficial effects of this invention are as follows: This invention effectively solves the core pain point of low efficiency in traditional manual operation through multi-sensor fusion and automated control architecture design. Specifically, the sensor array and infrared thermal imager on the device can collect real-time environmental data of the loquat orchard in all dimensions, and combine it with edge computing nodes to form a multimodal perception dataset, supporting the dynamic construction of a three-dimensional environmental model and the updating of the tree canopy temperature gradient. The fuzzy PID algorithm built into the control module can automatically optimize the coordinated control parameters of the telescopic actuator, side wall actuator, and drive motor based on the frost / rainstorm prediction results generated by the LSTM meteorological prediction model, realizing intelligent operation of the entire process from environmental perception, model building, prediction decision-making to execution control. Compared with manual operation, this device can shorten the response time to the second level, achieving a closed-loop response of "prediction-decision-execution" in sudden frost / rainstorm scenarios, significantly improving protection efficiency and reliability.

[0012] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0013] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a schematic diagram of the structure of the intelligent anti-freezing and rain-avoidance device for loquat orchards based on multi-sensor fusion in Embodiment 1 of the present invention; Figure 2 This is a partial structural schematic diagram of the intelligent anti-freezing and rain-avoidance device for loquat orchards based on multi-sensor fusion in Embodiment 1 of the present invention; Figure 3 This is a partial structural schematic diagram of an intelligent anti-freezing and rain-avoidance device for loquat orchards based on multi-sensor fusion in Embodiment 2 of the present invention; Figure 4 This is a flowchart of an intelligent anti-freezing and rain-avoidance adaptive control method for loquat orchards based on multi-sensor fusion, as described in Embodiment 3 of the present invention.

[0014] 1. Anti-freeze and rainproof bracket body; 2. Lead screw; 3. Telescopic actuator; 4. Top storage rod; 5. Film; 6. Windproof rod; 7. Counterweight rod; 8. Storage shed; 9. Side wall actuator; 10. Side storage rod; 11. Drive motor; 12. Discharge motor; 13. Support rod. Detailed Implementation

[0015] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0016] Example 1 like Figure 1 and 2 As shown, a smart anti-freezing and rain-shelter device for loquat orchards based on multi-sensor fusion includes: Anti-freeze and rainproof bracket main body 1; such as Figure 1 As shown, the main body 1 of the frost-proof and rain-sheltered support frame in this embodiment is a frame structure made of high-strength aluminum alloy, which has good weather resistance and corrosion resistance, and can adapt to the complex and ever-changing environmental conditions of the loquat orchard. The top of the main body 1 is designed with a triangular structure to reduce rainwater accumulation and improve wind resistance. In addition, the bottom of the support body 1 is fixed to the ground with anchor bolts to ensure the stability of the equipment in severe weather.

[0017] Lead screws 2 are located on both sides of the top of the antifreeze and rainproof bracket body 1 and are connected to the output shaft of the drive motor 11; for example Figure 1 As shown, in this embodiment, a slider is also installed on the lead screw 2, and the telescopic driver 3 is installed on the slider; two lead screws 2 are respectively installed on both sides of the top of the antifreeze and rainproof bracket body 1. Each lead screw 2 is equipped with a telescopic driver 3; The telescopic driver 3 is slidably mounted on the lead screws 2 on both sides of the top of the antifreeze and rainproof bracket body 1. In this embodiment, the telescopic driver 3 is a servo motor, which is mounted on the lead screw 2 via a slider and moves synchronously with the slider.

[0018] The top storage rod 4 is connected to the output shaft of the telescopic driver 3 set on the lead screw 2 on both sides of the top of the antifreeze and rainproof bracket body 1; the top storage rod 4 in this embodiment is made of metal material and has a cylindrical structure, with both ends installed on two opposite telescopic drivers 3.

[0019] Thin film 5 is disposed on the top storage rod 4; as Figure 1 As shown, one end of the film 5 is fixed to the antifreeze and rainproof bracket body 1, and the other end is rolled up on the top storage rod 4. The film 5 is retracted and extended by rotating the top storage rod 4.

[0020] Side wall drivers 9 are disposed on both sides of the antifreeze and rain shelter bracket body 1; the side wall drivers 9 are installed on the side of the antifreeze and rain shelter bracket body 1 and are close to the inner top.

[0021] The side storage rod 10 is movably mounted on the antifreeze and rainproof bracket body 1 and connected to the output shaft of the side wall driver 9. The side storage rod 10 is provided with a thin film 5; for example... Figure 2As shown, the side storage rod 10 is fixed to the output shaft of the side wall driver 9 and rotates with the rotation of the side wall driver 9. One end of the film 5 is fixed to the side storage rod 10, and the film 5 is retracted and extended by the forward and reverse rotation of the side wall driver 9.

[0022] The counterweight 7 is movably connected to the side storage rod 10 via the membrane 5. The counterweight 7 is fixed to the other side of the membrane 5 and drives the membrane 5 downward by its gravity, but the gravity of the counterweight 7 is less than the power of the side wall actuator 9.

[0023] like Figure 1 and 2 As shown, in this embodiment, when the intelligent anti-frost and rain shelter device for loquat orchards based on multi-sensor fusion is in use, if frost weather is detected within the next 1-2 hours (warning level reaches dangerous level), the control module will immediately trigger the emergency mode: the drive motor 11 starts and drives the lead screw 2 to rotate, causing the slider to drive the telescopic driver 3 to move along the lead screw 2 towards the top of the support body. At the same time, the output shaft of the telescopic driver 3 starts to rotate synchronously, and the moving speed of the slider is equal to the rotation speed of the output shaft of the driver 3 to drive the top storage rod 4 to rotate horizontally, so that the film 5 wrapped on the storage rod 4 is evenly unfolded to cover the top of the loquat tree canopy; synchronously, the side wall driver 9 rotates at a preset angular velocity, driving the side storage rod 10 to release the film 5, and the counterweight rod 7 slides down along the inner side of the support body under the action of gravity until the film 5 is fully unfolded to cover both sides of the tree canopy; when the film 5 is unfolded in place, the control module sends a signal to the electromagnet, the electromagnet is energized to generate a strong magnetic force to attract and fix the counterweight rod 7, preventing the film 5 from being lifted by the wind. The entire process takes no more than 3 minutes, far less than the 30 minutes or more required for manual operation, effectively preventing frost damage to the loquat fruit. If the sensors detect that the environmental risk has been eliminated (such as the temperature rising above freezing), the control module will initiate a reset procedure: the electromagnet is de-energized and releases the counterweight rod 7, the side wall driver 9 rotates in the opposite direction to retract the side film 5, the telescopic driver 3 retracts and drives the top storage rod 4 to move in the opposite direction, the drive motor 11 reverses to reset the slider, and the top film 5 is wound back into the storage rod 4, completing one complete anti-freeze protection cycle.

[0024] As an optional embodiment of the present invention, the device may further include a windbreak bar 6, and the counterweight bar 7 is disposed between the windbreak bar 6 and the antifreeze and rainproof bracket body 1.

[0025] like Figure 1As shown, the windbreak rod 6 is installed vertically and parallel to the column of the anti-freeze and rainproof bracket body 1, and close to the column. The counterweight rod 7 is restrained between the windbreak rod 6 and the anti-freeze and rainproof bracket body 1, which helps to prevent the membrane 5 on the counterweight rod 7 from swinging excessively due to wind force, thereby improving the stability of the equipment under severe weather conditions. The windbreak rod 6 is made of a lightweight but strong material, which can effectively reduce the weight of the overall equipment and has good impact resistance, and can withstand strong winds. In addition, the surface of the windbreak rod 6 is specially treated to have anti-slip and anti-corrosion functions, further extending the service life of the equipment.

[0026] As an optional embodiment of the present invention, optionally, an electromagnet is provided at the bottom of the anti-freeze and rainproof bracket body 1, and the electromagnet is used to fix the counterweight rod 7. Figure 1 As shown, in this embodiment, battery irons are installed on the bottom of the six pillars of the anti-freeze and rainproof bracket body 1. These irons are used to attract and fix the counterweight rods, ensuring that the counterweight rods can be effectively prevented from swaying or shifting due to wind after the film is unfolded. The electromagnet is designed with a high-efficiency and energy-saving coil structure, which generates a strong magnetic attraction when energized and quickly releases the attraction state after power is cut off, thereby achieving rapid response and precise control. In addition, the electromagnet's shell is treated with waterproofing and anti-corrosion, enabling it to adapt to the complex and changing environmental conditions of the loquat orchard and ensuring long-term stable operation.

[0027] As an optional embodiment of the present invention, optionally, a storage shed 8 is provided on the top of the anti-freeze and rainproof bracket body 1. Figure 1 As shown, the storage shed 8 is designed to drive all the telescopic actuators 3 into the storage shed 8 during rain or inclement weather, thereby improving the service life of the telescopic actuators 3.

[0028] As an optional embodiment of the present invention, the device may further include: The control module, equipped with a control algorithm, controls the telescopic driver 3, the sidewall driver 9, and the drive motor 11. In this embodiment, the control module is an FPGA, which receives real-time data from the sensor array in Embodiment 2 and precisely controls the telescopic driver 3, the sidewall driver 9, and the drive motor 11 based on a preset control algorithm. Through high-speed data processing capabilities, combined with fuzzy PID algorithm and dual-mode control logic, the control module dynamically adjusts various operating parameters of the equipment to ensure optimal anti-freezing and rain protection under different environmental conditions. Furthermore, the control module has a self-learning function, continuously optimizing the control strategy based on historical operating data to improve the equipment's intelligence level. To ensure system reliability, the control module also integrates multiple fault detection mechanisms. When an abnormal situation is detected, it can automatically switch to a backup control scheme or issue an alarm signal to remind operators to handle the situation promptly.

[0029] A signal transmitter, connected to the control module, is used to send control signals to the control module.

[0030] The signal transmitter in this embodiment is a wireless communication module that supports multiple communication protocols, including Wi-Fi, Bluetooth, and LoRa, to adapt to signal transmission requirements in different scenarios. Through real-time communication with the control module, the signal transmitter ensures rapid command transmission and immediate device response. Its low-power design effectively extends the overall battery life of the device, while also possessing strong anti-interference performance, enabling stable operation in complex electromagnetic environments.

[0031] Example 2 like Figure 3 As shown, in this embodiment, the ventilation method on the side of the antifreeze and rain shelter bracket body 1 is exactly the same as in embodiment 1. The difference in this embodiment lies in the top structure of the antifreeze and rain shelter bracket body 1. In this embodiment, four support rods 13 are installed on the antifreeze and rain shelter bracket body 1. Each support rod 13 has a receiver / discharge motor 12 installed on its top. The output shaft of the receiver / discharge motor 12 is connected to the top storage rod 4. A thin film 5 is rolled on the top storage rod 4, and the end of the thin film 5 is fixed to the top of the antifreeze and rain shelter bracket body 1. The bottom of each support rod 13 is rotatably connected to the antifreeze and rain shelter bracket body 1. The rotation can be achieved by using bearings.

[0032] Specifically, during use, when it is necessary to store the film 5 on the top storage rod 4, the two corresponding retractors 12 are started to rotate towards the top of the antifreeze and rain shelter bracket body 1. As the retractors 12 rotate, the top storage rod 4 starts to rotate synchronously, and the film 5 is rolled into the top storage rod 4. Since the end of the film 5 is fixed to the top of the antifreeze and rain shelter bracket body 1, the retractors 12 move towards the top of the antifreeze and rain shelter bracket body 1 under the pulling force of the film 5. At the same time, the support rod 13 rotates towards the top of the antifreeze and rain shelter bracket body 1, thus realizing the storage of the film 5. When the film 5 needs to be unfolded for frost and rain protection, the four collectors 12 are activated to rotate in the opposite direction. The top storage rod 4 then releases the film 5. Under the action of its own structure and the gravity of the collectors 12, the support rod 13 rotates away from the top of the frost and rain protection bracket body 1, causing the collectors 12 and the top storage rod 4 to unfold outward synchronously until the film 5 is completely laid flat on the top of the bracket body, forming a barrier covering the entire protective area. This effectively blocks the direct contact of the low-temperature airflow and rainwater with the loquat plants, thus achieving the function of frost and rain protection.

[0033] When encountering extreme weather such as strong winds, the retractor 12 can automatically start the retraction program based on the wind speed signal fed back by the sensor, quickly roll the film 5 back onto the top retraction rod 4, and at the same time, the support rod 13 is retracted to the top of the bracket body, reducing the wind-exposed area of ​​the equipment and improving its wind resistance.

[0034] Example 3 like Figure 4 As shown, a multi-sensor fusion-based intelligent anti-freezing and rain-avoidance adaptive control method for loquat orchards is applied to the aforementioned equipment. The method includes: S1. Based on the sensor array and infrared thermal imager deployed on the top of the anti-freezing and rain-sheltered support body 1, the edge computing node is used to collect real-time environmental data of the loquat orchard to form a multimodal perception dataset. In step S1, it should be noted that the sensor array includes temperature and humidity sensors, wind speed sensors, light intensity sensors, and rainfall sensors. These sensors are installed on the top and sides of the frost-proof and rain-sheltered support body 1 to achieve comprehensive monitoring of the loquat orchard environment. An infrared thermal imager is used to capture the temperature distribution on the tree canopy surface, providing crucial data support for subsequent frost warnings. Edge computing nodes connect to the sensor array via a high-speed data interface, enabling real-time processing and feature extraction of the collected raw data, thereby forming a multimodal sensing dataset containing environmental parameters and canopy temperature information. This dataset includes current environmental state information and also records historical data trends.

[0035] S2. Based on the collected multimodal sensing dataset, construct a three-dimensional environmental model of the loquat orchard and dynamically update the canopy temperature gradient distribution; S3. Based on the three-dimensional environment model of the loquat orchard, a meteorological prediction model is constructed using an LSTM neural network. The model is trained by fusing historical meteorological data with real-time sensor data to generate frost / rainstorm prediction results for the next n hours. S4. Based on the prediction results and dual-mode control logic, the fuzzy PID algorithm is used to optimize the coordinated control parameters of the telescopic driver 3, the side wall driver 9 and the drive motor 11, and generate the action command of the anti-freezing and rain-avoiding equipment. S5. Based on the generated motion command, control the drive motor 11 to drive the lead screw 2 and simultaneously start the telescopic driver 3 to adjust the top storage rod 4, and synchronously control the side wall driver 9 to adjust the side storage rod 10 to unfold the film 5, and fix the counterweight rod 7 by electromagnet.

[0036] In this embodiment, a multi-sensor fusion-based intelligent anti-frost and rain-avoidance adaptive control method for loquat orchards is based on the deep integration of multi-sensor fusion and intelligent control algorithms to achieve accurate perception and rapid response to environmental changes in the loquat orchard. Specifically, multi-dimensional data such as temperature, humidity, wind speed, light intensity, and rainfall collected by the sensor array are processed in real time by edge computing nodes to form a high-precision environmental state description. This data is further input into an LSTM neural network meteorological prediction model, which generates frost or rainstorm predictions for the next few hours through the fusion analysis of historical and real-time data. Based on these predictions, the control module dynamically switches operating modes according to dual-mode control logic: in basic mode, a fuzzy PID algorithm is used to finely adjust the drive motor, telescopic actuator, and sidewall actuator to ensure smooth and efficient equipment operation; in emergency mode, a fuzzy inference engine is activated to complete the film deployment and covering operation as quickly as possible, minimizing damage to the loquat tree canopy from sudden weather. The entire control process, from data acquisition to action execution, achieves closed-loop management, significantly improving the automation level and protection capabilities of the equipment.

[0037] As another optional embodiment of the present invention, optionally, in step S2, constructing a three-dimensional environment model of the loquat orchard and dynamically updating the canopy temperature gradient distribution based on the collected multimodal sensing dataset includes: S201. Based on the data from the lidar and vision sensor deployed on the top of the anti-freezing and rain-sheltered bracket body 1, the ICP algorithm is used to perform spatiotemporal alignment of point cloud and image data, and the initial three-dimensional model is generated by semantic segmentation through the U-Net network. In step S201, it is necessary to explain in detail that the lidar acquires high-precision point cloud data of the loquat orchard environment by emitting and receiving laser beams, while the visual sensor captures color image information. The combination of the two provides rich spatial and textural details. The ICP algorithm achieves precise alignment between the point cloud data and image data through multiple iterations of matching, thereby constructing an initial 3D model framework. Subsequently, the U-Net network is introduced for the semantic segmentation task. With its encoder-decoder structure, this network can efficiently extract key features from the image and map these features to the point cloud data, giving the 3D model more explicit semantic information. For example, different regions such as tree canopy, ground, and supports are clearly distinguished in the model.

[0038] S202. Based on the initial three-dimensional model and the canopy surface temperature data collected by the infrared thermal imager, the temperature features are extracted using the dual-light fusion algorithm combined with the attention mechanism module. The transpiration cooling effect is calculated using the three-temperature method to generate a vertical temperature gradient distribution map of the canopy. In step S202, it is necessary to explain in detail that the dual-light fusion algorithm combines temperature data acquired by an infrared thermal imager with spatial information from a 3D model, and uses an attention mechanism module to accurately extract temperature features from different areas of the tree canopy. Specifically, the algorithm first preprocesses the infrared thermal imaging data to remove noise and enhance the temperature contrast of key areas; then, it dynamically adjusts the weights through the attention mechanism module to highlight areas with significant temperature changes on the tree canopy surface, thereby improving the accuracy of feature extraction. Based on this, the three-temperature method is used to calculate the transpiration cooling effect, that is, by analyzing the relationship between the tree canopy surface temperature, ambient air temperature, and soil temperature, the impact of transpiration on the temperature distribution of the tree canopy is evaluated. The final generated vertical temperature gradient distribution map of the tree canopy can intuitively reflect the temperature change pattern from top to bottom of the canopy, providing a scientific basis for the subsequent frost warning and protection strategy formulation. Furthermore, this method can dynamically adjust the temperature gradient distribution map according to real-time updated sensor data, ensuring that the model always remains consistent with the actual environment.

[0039] S203. Based on the vertical temperature gradient distribution map of the tree canopy and combined with real-time meteorological data, the temperature field distribution in the initial three-dimensional model is dynamically adjusted, and the temperature gradient is smoothed by the Kalman filter algorithm to generate a three-dimensional environmental model of the loquat orchard.

[0040] In step S203, it is necessary to explain in detail that the Kalman filter algorithm effectively eliminates noise interference and improves the accuracy of the temperature field distribution by performing real-time analysis and optimization of the data in the vertical temperature gradient distribution map of the tree canopy. The algorithm first constructs a state prediction equation based on the spatial information in the initial 3D model and real-time temperature data collected by an infrared thermal imager to estimate the current temperature value of each region of the tree canopy. Subsequently, it dynamically adjusts the temperature field distribution by combining environmental meteorological data (such as wind speed and humidity) provided by the sensor array. During this process, the Kalman filter continuously updates the optimal estimate of the temperature gradient through recursive calculation, thereby achieving accurate capture of temperature change trends. Furthermore, to ensure the stability and reliability of the model, the algorithm also introduces an adaptive weight adjustment mechanism to automatically adjust the filtering parameters according to changes in environmental conditions. The final generated 3D environmental model of the loquat orchard not only contains high-precision spatial structural information but also reflects the dynamic changes of the tree canopy temperature field in real time.

[0041] As another optional embodiment of the present invention, optionally, in step S3, based on the three-dimensional environment model of the loquat orchard, a meteorological prediction model is constructed using an LSTM neural network, and the model is trained by fusing historical meteorological data with real-time sensor data to generate frost / rainstorm prediction results for the next n hours, including: S301. Based on the three-dimensional environment model of the loquat orchard and the multimodal perception dataset, the U-Net network is used to extract the spatial features of the tree canopy, and the LSTM network is combined to capture the n-hour time series patterns to generate multi-scale spatiotemporal feature vectors. In step S301, it is necessary to explain in detail that the U-Net network first performs fine segmentation of the canopy region in the 3D environment model of the loquat orchard, extracting the spatial structural features of the canopy, such as its height, volume, and distribution density. These features are further optimized through an encoder-decoder structure to form a high-resolution spatial feature map. Simultaneously, the LSTM network performs time-series analysis on historical meteorological data and real-time sensor data from the multimodal sensing dataset, capturing the patterns of environmental parameter changes over time, such as temperature fluctuations, wind speed trends, and humidity variations. Subsequently, the spatial features extracted by U-Net are fused with the time-series features captured by LSTM to generate a multi-scale spatiotemporal feature vector. This fusion method not only reflects the static characteristics of the canopy spatial structure but also captures the dynamic changes in environmental parameters.

[0042] S302. Based on multi-scale spatiotemporal feature vectors and real-time temperature data collected by an infrared thermal imager, a prediction model is constructed using a dual-layer LSTM network and an attention mechanism module, and basic prediction results are generated through a fully connected layer. In step S302, it is necessary to explain in detail that the first layer of the dual-layer LSTM network is mainly used to process multi-scale spatiotemporal feature vectors and extract the complex correlation between environmental parameters and canopy spatial structure. The second layer further combines real-time temperature data collected by infrared thermal imager and dynamically weights key features through the attention mechanism module to highlight factors that have an important impact on frost or rainstorm prediction. For example, in the case of a sudden drop in temperature or a sharp increase in humidity, the attention mechanism can automatically enhance the weight of these abnormal changes in the model, thereby improving the accuracy of prediction. The fully connected layer then integrates the features processed by the dual-layer LSTM network and outputs the basic prediction result. This result represents the probability of frost or rainstorm occurring in the next n (72) hours in probabilistic form, and provides the trend curve of the relevant environmental parameters, laying the foundation for the formulation of subsequent control strategies.

[0043] Furthermore, to further improve the robustness of the prediction model, this embodiment also introduces an adaptive learning rate optimization algorithm. This algorithm dynamically adjusts the learning rate based on changes in the loss function during training, ensuring that the model can converge quickly under different data distribution conditions. Simultaneously, to avoid overfitting, the model employs Dropout regularization, randomly discarding some neurons during the training phase to enhance the model's generalization ability. The final prediction results not only possess high accuracy but also adapt to the complex and ever-changing environmental conditions of the loquat orchard.

[0044] S303. Based on the basic prediction results and the sliding window training strategy, the prediction model parameters are dynamically updated every x minutes. The Kalman filter algorithm is used to fuse historical meteorological data and real-time sensor data to generate a probability distribution map of frost / rainstorm for the next n hours.

[0045] In step S303, it is necessary to explain in detail that the sliding window training strategy periodically optimizes the basic prediction results by setting a fixed-length time window. Within each time window, the model dynamically adjusts its internal parameters based on the latest real-time sensor data and historical meteorological data to ensure that the prediction results keep up with the trend of environmental changes. The Kalman filter algorithm plays a key role in this process. It further improves the accuracy and stability of the prediction model by fusing and analyzing historical meteorological data and real-time sensor data. Specifically, the Kalman filter first uses a state-space model to estimate core environmental parameters such as temperature, humidity, and wind speed, and continuously corrects the prediction error by combining real-time data. Then, it generates a probability distribution map of frost or rainstorm in the next n hours through recursive calculation. This distribution map not only intuitively shows the probability of extreme weather events occurring in different time periods, but also provides detailed environmental parameter change curves, laying a scientific basis for the subsequent generation of action commands for anti-freezing and rain shelter equipment. In addition, in order to cope with the complex and ever-changing environmental conditions of the loquat orchard, this embodiment also introduces an adaptive weight allocation mechanism, which dynamically adjusts the weight of each sensor data according to the importance of the current environmental characteristics, thereby further improving the adaptability and reliability of the prediction model.

[0046] As another optional embodiment of the present invention, optionally, in step S4, based on the prediction results and dual-mode control logic, the coordinated control parameters of the telescopic driver 3, the side wall driver 9, and the drive motor 11 are optimized using a fuzzy PID algorithm to generate the anti-freezing and rain-avoidance device action command, including: S401. Based on the frost / rainstorm prediction probability distribution map in the prediction results, use the fuzzy logic rule base to establish an environmental level classification standard and divide the prediction results into five warning levels: safe, attention, warning, danger, and emergency. In step S401, it is necessary to explain in detail that the fuzzy logic rule base analyzes key environmental parameters in the probability distribution map of frost or heavy rain, and combines this with the temperature field distribution and canopy status information in the three-dimensional environmental model of the loquat orchard to formulate a scientific standard for classifying early warning levels. Specifically, the safety level corresponds to scenarios with low probability extreme weather events and stable environmental parameters; the attention level is applicable to situations where there are slight abnormal fluctuations in the prediction results but have not yet reached the danger threshold; the warning level is for medium probability extreme weather events, indicating that some protective measures need to be activated; the danger level corresponds to high probability extreme weather events, requiring the equipment to enter a full protection mode; and the emergency level is used in extreme situations, such as sudden frost or heavy rain, requiring immediate and rapid response actions. Each early warning level is associated with a set of preset action instruction templates.

[0047] S402. Based on the warning level, dynamically adjust the PID controller parameters, optimize the Kp / Ki / Kd coefficients through a genetic algorithm to obtain the optimized control parameters, and perform coordinated control of the telescopic actuator 3, the side wall actuator 9 and the drive motor 11. In step S402, it is necessary to explain in detail that the genetic algorithm dynamically optimizes the Kp, Ki, and Kd coefficients of the PID controller by simulating natural selection and genetic mechanisms. During the optimization process, a fitness function is first set according to the warning level, which comprehensively considers key indicators such as control accuracy, response speed, and system stability. Subsequently, the algorithm generates an initial population and iterates continuously through selection, crossover, and mutation operations to find the optimal parameter combination. For example, under dangerous or emergency warning levels, the algorithm prioritizes improving response speed and appropriately reduces the stability requirements to ensure that the equipment can quickly complete protective actions; while under safe or caution levels, it focuses more on balancing control accuracy and energy consumption. The optimized Kp, Ki, and Kd coefficients are input into the fuzzy PID controller to adjust the cooperative working mode of the telescopic actuator 3, the side wall actuator 9, and the drive motor 11. This parameter optimization method based on the genetic algorithm not only improves the adaptive capability of the control system but also significantly enhances the operating efficiency of the equipment in complex environments.

[0048] Furthermore, to further improve control accuracy, this embodiment introduces a multi-objective optimization strategy. This strategy, based on the genetic algorithm, adds a comprehensive evaluation of multiple performance indicators, such as the uniformity of film unfolding, the load distribution of the drive motor, and the overall energy consumption level. In this way, the control system can meet protection requirements while minimizing resource consumption and extending equipment lifespan. The resulting coordinated control parameters are not only highly flexible but can also be dynamically adjusted according to real-time environmental changes, thus providing reliable assurance for the operation commands of the anti-freezing and rain-avoidance equipment.

[0049] S403: Based on the dual-mode control logic switching strategy, PID closed-loop control is used for adjustment in the basic mode, and fuzzy inference engine is enabled for rapid response in the emergency mode. In step S403, it is necessary to explain in detail that the dual-mode control logic switching strategy intelligently determines whether the current scenario requires switching from the basic mode to the emergency mode by monitoring the environmental status and equipment operating parameters in real time. In the basic mode, PID closed-loop control, with optimized Kp, Ki, and Kd parameters as its core, ensures smooth and efficient equipment operation through precise adjustment of the telescopic actuator 3, side wall actuator 9, and drive motor 11. This mode is suitable for scenarios with relatively mild environmental changes and prediction results at the safe or warning level, effectively balancing control accuracy and energy consumption. In the emergency mode, the fuzzy inference engine is activated, quickly generating action commands based on the fuzzy logic rule base, prioritizing response speed. For example, in the event of a sudden frost or rainstorm warning, the fuzzy inference engine will directly trigger the film deployment and covering operation based on the judgment result of the danger or emergency level, minimizing the damage of extreme weather to the loquat tree canopy. The switching between the two modes is achieved by a dynamic threshold judgment mechanism, which comprehensively considers the rate of change of environmental parameters, the warning level, and the current status of the equipment, ensuring a seamless and lag-free switching process. In addition, to improve the robustness of the system, this embodiment also introduces a redundancy verification module, which performs secondary verification of the control commands after each mode switch to avoid unnecessary actions caused by misjudgment, thereby further improving the reliability and safety of equipment operation.

[0050] S404. Generate a set of equipment action instructions based on the optimized control parameters, including specific control parameters for the speed of the drive motor 11, the stroke of the lead screw 2, the extension amount of the telescopic driver 3, and the unfolding height of the side wall driver 9.

[0051] In step S404, it is necessary to explain in detail that the generation process of the equipment action instruction set comprehensively considers the optimized control parameters and real-time environmental data to ensure that each instruction accurately matches the current protection requirements. Specifically, the rotational speed of the drive motor 11 is dynamically adjusted by a fuzzy PID controller according to the uniformity requirements of the film deployment to ensure the smoothness and efficiency of the covering operation. The stroke of the lead screw 2 is determined by analyzing the temperature gradient distribution map of the tree canopy in the vertical direction using a Kalman filter algorithm to ensure that it can provide the best protection effect within different height ranges. The extension amount of the telescopic actuator 3 is calculated based on the warning level and the characteristics of the tree canopy space, prioritizing the coverage needs of dangerous areas while avoiding resource waste caused by excessive operation. The deployment height of the side wall actuator 9 is optimized by combining wind speed and humidity information from real-time meteorological data to minimize the interference of the external environment on the equipment operation while ensuring the protection effect. The final generated action instruction set not only includes specific control parameters but also includes execution priorities and time sequence arrangements, providing comprehensive guidance for the efficient operation of the anti-freezing and rain-avoidance equipment. In addition, to cope with the complex and ever-changing environment of the loquat orchard, this embodiment also introduces a dynamic feedback correction mechanism, which collects sensor data in real time during the execution of instructions and fine-tunes the action instructions, thereby further improving the response accuracy and reliability of the equipment.

[0052] The method further includes: collecting the carbon dioxide concentration inside and outside the antifreeze and rain shelter bracket body 1; when the carbon dioxide concentration inside the antifreeze and rain shelter bracket body 1 is equal to or less than the carbon dioxide concentration outside the antifreeze and rain shelter bracket body 1, activating the side wall driver 9 to open the film 5 on the side wall of the antifreeze and rain shelter bracket body 1.

[0053] It should be noted that this carbon dioxide concentration regulation mechanism aims to balance the needs of loquat orchard frost and rain protection with the photosynthetic needs of the plants, avoiding excessively low carbon dioxide concentrations within the support structure due to long-term closed protection, which could negatively impact loquat growth. The specific implementation process includes: First, high-precision carbon dioxide sensors are deployed on the inner and outer sides of the frost and rain protection support structure 1 to collect and transmit concentration data to the central controller in real time; second, the central controller calculates the difference between the internal and external concentration values, sets a dynamic threshold (e.g., triggering an action when the internal concentration is more than 5% lower than the external concentration), and determines whether to perform the membrane opening operation based on the current warning level. The side wall actuator 9 is only allowed to be activated when the warning level is "safe" or "caution"; third, based on the magnitude of the concentration difference, the side wall actuator 9 expands the membrane 5 to a preset ratio (e.g., 30% expansion when the difference is 5%-10%, and 50% expansion when the difference is greater than 10%) to achieve controllable gas exchange.

[0054] It should also be noted that when the carbon dioxide concentration inside the greenhouse is lower than the carbon dioxide concentration outside the greenhouse, there are two situations: (1) when the temperature outside the greenhouse is lower than the preset temperature, the carbon dioxide generator inside the greenhouse is turned on; (2) when the temperature outside the greenhouse is higher than the preset temperature, the top film 5 and the surrounding film 5 of the greenhouse are turned on to achieve fine adjustment of the carbon dioxide concentration inside the greenhouse.

[0055] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A smart anti-freezing and rain-sheltered device for loquat orchards based on multi-sensor fusion, characterized in that: The device includes: Anti-freezing and rainproof bracket body (1); The lead screw (2) is located on both sides of the top of the antifreeze and rainproof bracket body (1) and is connected to the output shaft of the drive motor (11); The telescopic driver (3) is slidably mounted on the lead screws (2) on both sides of the top of the antifreeze and rain shelter bracket body (1); The top storage rod (4) is connected to the output shaft of the telescopic driver (3) provided on the screw (2) on both sides of the top of the antifreeze and rain shelter bracket body (1); A thin film (5) is disposed on the top storage rod (4); Side wall actuators (9) are disposed on both sides of the antifreeze and rainproof bracket body (1); The side storage rod (10) is movably mounted on the antifreeze and rainproof bracket body (1) and connected to the output shaft of the side wall driver (9). The side storage rod (10) is provided with a thin film (5). The counterweight rod (7) is movably connected to the side storage rod (10) via the membrane (5).

2. The intelligent anti-freezing and rain-sheltered device for loquat orchards based on multi-sensor fusion as described in claim 1, characterized in that, The device also includes a windbreak bar (6), and the counterweight bar (7) is disposed between the windbreak bar (6) and the antifreeze and rainproof bracket body (1).

3. The intelligent anti-freezing and rain-sheltered device for loquat orchards based on multi-sensor fusion as described in claim 1 or 2, characterized in that, An electromagnet is provided at the bottom of the main body (1) of the antifreeze and rainproof bracket, and the electromagnet is used to fix the counterweight rod (7).

4. The intelligent anti-freezing and rain-sheltered device for loquat orchards based on multi-sensor fusion as described in claim 1, characterized in that, The main body (1) of the antifreeze and rain shelter is equipped with a storage shed (8) on top.

5. The intelligent anti-freezing and rain-sheltered device for loquat orchards based on multi-sensor fusion as described in claim 1, characterized in that, The device also includes: The control module contains a control algorithm for controlling the telescopic driver (3), the side wall driver (9), and the drive motor (11). A signal transmitter, connected to the control module, is used to send control signals to the control module.

6. A method for intelligent anti-freezing and rain-avoidance adaptive control of loquat orchards based on multi-sensor fusion, characterized in that, Applied to the device as described in any one of claims 1 to 2, the method comprises: S1. Based on the sensor array and infrared thermal imager deployed on the top of the anti-freezing and rain-sheltered support body (1), the edge computing node is used to collect real-time environmental data of the loquat orchard to form a multimodal perception dataset. S2. Based on the collected multimodal sensing dataset, construct a three-dimensional environmental model of the loquat orchard and dynamically update the canopy temperature gradient distribution; S3. Based on the three-dimensional environment model of the loquat orchard, a meteorological prediction model is constructed using an LSTM neural network. The model is trained by fusing historical meteorological data with real-time sensor data to generate frost / rainstorm prediction results for the next n hours. S4. Based on the prediction results and dual-mode control logic, the fuzzy PID algorithm is used to optimize the coordinated control parameters of the telescopic driver (3), the side wall driver (9) and the drive motor (11) to generate the action command of the anti-freezing and rain-avoiding equipment. S5. Based on the generated action command, control the drive motor (11) to drive the lead screw (2) and simultaneously start the telescopic driver (3) to adjust the top storage rod (4), synchronously control the side wall driver (9) to adjust the side storage rod (10) to unfold the film (5), and fix the counterweight rod (7) by electromagnet.

7. The intelligent anti-freezing and rain-avoidance adaptive control method for loquat orchards based on multi-sensor fusion as described in claim 6, characterized in that, In step S2, based on the collected multimodal sensing dataset, a three-dimensional environmental model of the loquat orchard is constructed and the canopy temperature gradient distribution is dynamically updated, including: S201. Based on the data of the lidar and vision sensor deployed on the top of the anti-freezing and rainproof bracket (1), the point cloud and image data are spatiotemporally aligned using the ICP algorithm, and the initial three-dimensional model is generated by semantic segmentation through the U-Net network. S202. Based on the initial three-dimensional model and the canopy surface temperature data collected by the infrared thermal imager, the temperature features are extracted using the dual-light fusion algorithm combined with the attention mechanism module. The transpiration cooling effect is calculated using the three-temperature method to generate a vertical temperature gradient distribution map of the canopy. S203. Based on the vertical temperature gradient distribution map of the tree canopy and combined with real-time meteorological data, the temperature field distribution in the initial three-dimensional model is dynamically adjusted, and the temperature gradient is smoothed by the Kalman filter algorithm to generate a three-dimensional environmental model of the loquat orchard.

8. The intelligent anti-freezing and rain-avoidance adaptive control method for loquat orchards based on multi-sensor fusion as described in claim 6, characterized in that, In step S3, based on the three-dimensional environment model of the loquat orchard, a meteorological prediction model is constructed using an LSTM neural network. This model is trained by fusing historical meteorological data with real-time sensor data to generate frost / rainstorm prediction results for the next n hours, including: S301. Based on the three-dimensional environment model of the loquat orchard and the multimodal perception dataset, the U-Net network is used to extract the spatial features of the tree canopy, and the LSTM network is combined to capture the n-hour time series patterns to generate multi-scale spatiotemporal feature vectors. S302. Based on multi-scale spatiotemporal feature vectors and real-time temperature data collected by an infrared thermal imager, a prediction model is constructed using a dual-layer LSTM network and an attention mechanism module, and basic prediction results are generated through a fully connected layer. S303. Based on the basic prediction results and the sliding window training strategy, the prediction model parameters are dynamically updated every x minutes. The Kalman filter algorithm is used to fuse historical meteorological data and real-time sensor data to generate a probability distribution map of frost / rainstorm for the next n hours.

9. The intelligent anti-freezing and rain-avoidance adaptive control method for loquat orchards based on multi-sensor fusion as described in claim 6, characterized in that, In step S4, based on the prediction results and dual-mode control logic, the coordinated control parameters of the telescopic actuator (3), side wall actuator (9), and drive motor (11) are optimized using a fuzzy PID algorithm to generate the action commands for the anti-freezing and rain-avoiding equipment, including: S401. Based on the frost / rainstorm prediction probability distribution map in the prediction results, use the fuzzy logic rule base to establish an environmental level classification standard and divide the prediction results into five warning levels: safe, attention, warning, danger, and emergency. S402. Based on the warning level, dynamically adjust the PID controller parameters, optimize the Kp / Ki / Kd coefficients through a genetic algorithm, obtain the optimized control parameters, and perform coordinated control of the telescopic driver (3), the side wall driver (9), and the drive motor (11). S403: Based on the dual-mode control logic switching strategy, PID closed-loop control is used for adjustment in the basic mode, and fuzzy inference engine is enabled for rapid response in the emergency mode. S404. Generate a set of device action instructions based on the optimized control parameters, including specific control parameters for the drive motor (11) speed, lead screw (2) stroke, telescopic driver (3) telescopic amount, and side wall driver (9) unfolding height.

10. The intelligent anti-freezing and rain-avoidance adaptive control method for loquat orchards based on multi-sensor fusion as described in claim 6, characterized in that, The method further includes: collecting the carbon dioxide concentration inside and outside the antifreeze and rain shelter bracket body (1); when the carbon dioxide concentration inside the antifreeze and rain shelter bracket body (1) is equal to or less than the carbon dioxide concentration outside the antifreeze and rain shelter bracket body (1), activating the side wall driver (9) to open the film (5) on the side wall of the antifreeze and rain shelter bracket body (1).

Images