Search

CN-121970739-A - Unmanned aerial vehicle flow dynamic feedback governing system that gives medicine to poor free of charge

CN121970739ACN 121970739 ACN121970739 ACN 121970739ACN-121970739-A

Abstract

The invention belongs to the field of plant protection machinery and control systems, and particularly relates to a dynamic feedback regulation system for pesticide delivery flow of an unmanned aerial vehicle. The invention discloses a dynamic feedback regulation system for the drug delivery flow of an unmanned aerial vehicle. The system comprises a multidimensional agriculture emotion knowing device, a physiological condition analyzing device, a deep learning decision-making processing system, a pesticide application parameter feedback adjusting device and an unmanned aerial vehicle flight state monitoring device, wherein multisource data are acquired through a sensing device, a physiological stage is determined by utilizing the physiological condition analyzing, a target flow demand is calculated by the deep learning system in combination with an attention mechanism, and the adjusting device is driven to realize variable spraying and closed-loop control. According to the invention, through multidimensional fusion decision of agriculture condition, physical condition and flying state, accurate spraying based on actual disaster demand and growth rule of crops is realized, the utilization rate of liquid medicine, deposition efficiency and operation stability are improved, and the risk of missing spraying caused by visual blind spots is eliminated.

Inventors

  • GUO QING
  • CHENG HANG

Assignees

  • 太原科技大学

Dates

Publication Date
20260505
Application Date
20260403

Claims (10)

  1. 1. The unmanned aerial vehicle pesticide application flow dynamic feedback regulation system is characterized by comprising a multidimensional pesticide emotion awareness device, a physiological weather analysis device, a deep learning decision processing system, a pesticide application parameter feedback regulation device and an unmanned aerial vehicle flight state monitoring device; The multidimensional agriculture emotion awareness device is used for acquiring multisource perception data of a crop canopy below in real time in the unmanned aerial vehicle operation process, wherein the multisource perception data comprises multispectral characteristic information and high-resolution color image information, and transmitting the multispectral characteristic information and the high-resolution color image information to the deep learning decision processing system; The physiological climate analyzing device is used for storing historical growth files of crops and processing current climate stage characteristics, judging the current physiological stage of the crops by extracting morphological characteristics of the crops, and providing biological constraint parameters for application decisions; the deep learning decision processing system is used for receiving perception data and climate information, carrying out region weighted analysis on crop canopy by using an attention mechanism identification algorithm, identifying the severity level and the distribution region of plant diseases and insect pests, and calculating a target flow demand and a target pressure demand; The pesticide application parameter feedback adjusting device is connected with the deep learning decision processing system and is used for adjusting the power parameters of the spraying assembly according to the target flow demand and the target pressure demand and feeding back the actual flow and the pressure data in real time to form closed-loop control; the unmanned aerial vehicle flight state monitoring device is used for monitoring real-time space coordinates, flight speed and operation height of the unmanned aerial vehicle and feeding back physical parameters to the deep learning decision processing system to correct the pesticide application decision instruction.
  2. 2. The unmanned aerial vehicle drug delivery flow dynamic feedback regulation system of claim 1, wherein the multidimensional agriculture emotion awareness device comprises a spectrum sensing unit and a visual camera unit which are physically configured on an unmanned aerial vehicle belly bottom damping cradle head system; the visual camera unit adopts a high-frame-rate complementary metal oxide semiconductor sensor and is used for acquiring an original image with millimeter-level spatial resolution; The spectrum sensing unit comprises a plurality of narrow-band optical filters and a photosensitive array and is used for acquiring reflectivity characteristics of crops in bands including red light bands, green light bands and near infrared bands; The multidimensional pesticide emotion awareness device is internally further integrated with an ambient light compensation unit, and the ambient light compensation unit monitors the natural illumination intensity of an operation area through a built-in upward observation photosynthetic effective radiation sensor and operates an ambient light compensation logic; The ambient light compensation logic performs brightness correction and color reconstruction on the original image according to the ambient light intensity detected in real time, and eliminates interference of cloud cover, solar altitude angle change and shadow in crop canopy on feature extraction by adjusting the exposure time and the sensitivity gain of the visual camera unit, so that comparability of data acquired at different time periods in time dimension is ensured.
  3. 3. The unmanned aerial vehicle application flow dynamic feedback regulation system according to claim 2, wherein the deep learning decision processing system is internally provided with a feature extraction layer based on an attention mechanism, and the feature extraction layer distributes weights to local areas containing suspected lesions or pests by performing pixel-level processing on color image information; The deep learning decision processing system further comprises a space dimension attention layer and a channel dimension attention layer, wherein the space dimension attention layer locks a disease and pest high-incidence abnormal region in a crop canopy through global pooling and correlation calculation; the channel dimension attention layer is used for screening spectral band information with expressive force on plant diseases and insect pests, and the detection efficiency of the model under background interference is enhanced by giving different characteristic gains to different spectral channels; the deep learning decision processing system performs multi-scale downsampling processing on the high-resolution color image through built-in convolution check, extracts basic visual characteristics comprising edges, textures and color moments, and calculates cross-correlation coefficients among different pixel areas in the image through a self-attention module; when the cross-correlation coefficient of a certain area exceeds a preset abnormal threshold, the system judges that the area is a suspected disease area and starts a local feature amplification program, and secondary confirmation is carried out by calling a sub-image with higher resolution.
  4. 4. The unmanned aerial vehicle dosing flow dynamic feedback regulation system of claim 3, wherein the physiological weather analysis device logically comprises an embedded high-performance storage unit, wherein a full life cycle weather feature vector database of crops is pre-stored in the embedded high-performance storage unit, and the full life cycle weather feature vector database comprises typical visual representations from a seedling emergence period, a tillering period, a jointing period, a booting period, a heading period to a maturity period, and comprises an aspect ratio of leaves, a height distribution probability of canopy layers and a hue saturation characteristic of buds; The physiological weather analysis device is used for continuously tracking the development state of crops by utilizing a time sequence analysis logic, and dividing the growth period of the crops into a plurality of preset stages including a vegetative growth period, a flowering period and a fruiting period according to the identified leaf density, the identified flower cluster number and the identified fruit size; Aiming at different preset stages, the physiological condition analyzing device outputs different application strategy constraint limits; In the fruiting period, the physiological weather analysis device increases the weight of the fog drop deposition density of the fruit part and sets a phytotoxicity sensitivity threshold; in the vegetative growth period, the physiological weather analysis device adjusts the flow weight of the penetrating power of the blades by analyzing the hierarchical structure of the blades.
  5. 5. The unmanned aerial vehicle dosing flow dynamic feedback adjustment system of claim 4, wherein the deep learning decision processing system is configured to construct an end-to-end flow decision model that uses the parsed agronomic characteristics, physiological climate information, and flight state monitoring data as multidimensional input variables; The end-to-end flow decision model performs nonlinear mapping through an internal multi-layer neuron network structure, maps high-dimensional image features into a control signal of a bottom layer, and outputs an application flow instruction and a system pressure instruction; In the operation process, the end-to-end flow decision model adjusts the expected value of the particle size of the fog drops according to the complexity of the three-dimensional structure of the crop canopy; When the thickness of the crown layer exceeds a preset threshold value through a three-dimensional reconstruction algorithm, the flow decision model increases a system pressure instruction value, and the penetration capacity of the liquid medicine is improved by reducing the particle size of the fog drops and increasing the kinetic energy; When the deep learning decision processing system executes a decision, the identified plant diseases and insect pests level is converted into a corresponding pesticide application intensity index, and the pesticide application intensity index and the target flow demand are in positive correlation; meanwhile, the system calculates the required droplet size according to the operation target position determined by the physiological material analysis device, and adjusts the pressure set value in the drug application parameter feedback adjustment device according to the droplet size.
  6. 6. The unmanned aerial vehicle dispensing flow dynamic feedback adjustment system of claim 5, wherein the dispensing parameter feedback adjustment device comprises a variable frequency pump control subsystem, a pressure adaptive adjustment subsystem, a flow sensing unit, and a pressure sensing unit; the variable frequency pump control subsystem adopts a pulse width modulation technology, adjusts the rotating speed of the driving motor according to the received target flow demand, and realizes millisecond-level adjustment of the output flow of the drug delivery pump; The pressure self-adaptive adjusting subsystem maintains the stability of the spraying pressure by adjusting the opening of the electromagnetic valve of the spray head end and the proportion of the bypass reflux valve, and ensures that the particle size of the mist drops does not have abnormal fluctuation in the variable spraying process; the flow sensing unit and the pressure sensing unit respectively capture an actual flow value and an actual pressure value in the spraying link in real time, and transmit the data back to the deep learning decision processing system through the high-speed industrial bus; When the absolute value of the deviation between the actual flow and the target flow exceeds a preset allowable threshold, the system starts a deviation correction program, and the duty ratio of the variable frequency pump is corrected in real time through a proportional-integral-derivative algorithm until the deviation returns to the allowable range; The variable frequency pump control subsystem is also provided with pre-pressurizing logic, and the pump set is started in advance within a certain microsecond time before the spraying instruction is issued.
  7. 7. The unmanned aerial vehicle dosing flow dynamic feedback adjustment system of claim 6, wherein the unmanned aerial vehicle flight status monitoring device is integrated with a differential global positioning system module, an inertial measurement unit, and a lidar altimeter module; The laser radar height measurement module is used for acquiring the real-time distance between the unmanned aerial vehicle and the top end of the crop canopy so as to correct the imaging scale of the vision sensor in real time; The speed vector output by the unmanned aerial vehicle flight state monitoring device in real time is introduced into the motion compensation logic of the deep learning decision processing system; The motion compensation logic predicts an actual drift track of the fogdrops after leaving the spray head by using a preset fluid dynamics prediction model according to a forward speed vector of the unmanned aerial vehicle and the current undershoot wind field intensity; the system adjusts the time point of spraying opening and the spraying angle in advance according to the predicted drift displacement, and counteracts the deviation of the spraying position caused by flying movement and lateral wind power; the motion compensation logic further considers the influence of the pitch angle and the roll angle of the unmanned aerial vehicle on the spraying coverage range in the gesture adjustment process, calculates the included angle offset between the central axis of the spray head and the ground vertical line by acquiring the three-axis gesture angular speed output by the inertial measurement unit, and performs geometric compensation calculation on the target flow when the offset exceeds a preset angle threshold.
  8. 8. The unmanned aerial vehicle dosing flow dynamic feedback regulation system of claim 7, wherein the deep learning decision processing system has a blind spot prediction function and a self-diagnostic unit; When the multidimensional agriculture emotion awareness device is limited by visual angles or is blocked by branches and leaves, and certain shielding areas cannot be completely observed, the deep learning decision processing system builds a three-dimensional probability occupation map of crops by utilizing a time sequence reasoning module based on a long-short-period memory network based on the principle of consistency of pest and disease distribution density and crop growth of surrounding known areas, estimates the disaster tolerance degree of a blind spot area by a preset probability reasoning algorithm, and generates corresponding compensation application instructions; the self-diagnosis unit continuously monitors heartbeat signals of all modules and packet loss rate of a data link, when detecting that a spectrum sensing unit of the multidimensional pesticide emotion sensing device has hardware faults, the system automatically switches to a single-source identification mode based on a visual camera unit, and a standby lightweight decision model is called to ensure continuity of operation tasks; The system is internally provided with a flow optimization module based on reinforcement learning, and the module records the actual sedimentation effect after operation and feeds the actual sedimentation effect back to a flow decision model as a reward signal.
  9. 9. The unmanned aerial vehicle dosing flow dynamic feedback regulation system of claim 8, wherein a unified data communication bus is established within the system, the data communication bus being based on a high bandwidth, low latency fieldbus protocol for coordinating data interactions between devices; the system runs a synchronization algorithm based on an accurate time protocol, and each device shares a nanosecond global clock through a bus; when the multidimensional agriculture condition sensing device collects images, a current time stamp is automatically stamped in the data packet; After calculating a flow instruction, the deep learning decision processing system calculates a space coordinate corresponding to the instruction according to a time stamp and a real-time speed corresponding to the instruction; After receiving the instruction, the pesticide application parameter feedback adjusting device triggers the executing mechanism at the moment that the unmanned aerial vehicle reaches the space coordinate through an internal delay line buffer queue so as to eliminate the space hysteresis effect caused by time consumption of calculation; The system adopts a dual-redundancy controller local area network bus architecture, two sets of mutually independent data buses run simultaneously and monitor the integrity of the data streams in real time, and when a communication fault occurs in a main bus, the system is switched to a standby bus in a seamless way within microsecond time.
  10. 10. The unmanned aerial vehicle dispensing flow dynamic feedback regulation system of claim 9, wherein the dispensing parameter feedback regulation device further comprises a liquid medicine residual quantity monitoring unit, wherein the liquid medicine residual quantity monitoring unit adopts a non-contact ultrasonic liquid level sensor and is matched with an inclination sensor to dynamically compensate liquid level data; the deep learning decision processing system extracts real allowance information from the original liquid level signal by utilizing a liquid shaking suppression algorithm according to the acceleration and gesture data output by the flight state monitoring device; When the liquid level is lower than a preset warning threshold value, the system automatically triggers an energy-saving operation mode, optimizes a drug application path or reduces flow in a non-target area so as to ensure that the residual liquid medicine covers a core target area; When the deep learning decision processing system processes agricultural condition characteristics, crop images acquired by different time sequences are spliced and fused with the characteristics by utilizing a spatial weighting aggregation logic, and a dynamic agricultural condition prescription chart of an operation area is constructed, wherein the agricultural condition prescription chart comprises crop drug demand levels and physical constraint instructions corresponding to different geographic coordinate points; And the pesticide application parameter feedback adjusting device retrieves the corresponding pesticide demand level from the pesticide condition prescription graph according to the real-time longitude and latitude coordinates of the unmanned aerial vehicle and executes flow adjustment.

Description

Unmanned aerial vehicle flow dynamic feedback governing system that gives medicine to poor free of charge Technical Field The invention belongs to the technical field of plant protection machinery and control systems, and particularly relates to a dynamic feedback regulation system for pesticide delivery flow of an unmanned aerial vehicle. Background Along with the rapid evolution of modern agricultural technology, unmanned aerial vehicle plant protection operation has become the important technical support of realizing accurate agriculture and green prevention and control of plant diseases and insect pests. Traditional unmanned aerial vehicle sprinkler system mainly relies on to predetermine flow or simple flight parameter proportion adjustment, aims at replacing artifical spraying with improvement operating efficiency through automation equipment. In complex farmland and orchard environments, the diversity of crop growth states and the heterogeneity of pest and disease distribution put forward high requirements on the precision of pesticide application, and the pesticide utilization rate is improved and the agricultural ecological environment is protected through real-time sensing of pesticide conditions and dynamic adjustment of spraying strategies. Variable spraying control technology based on visual perception and physiological state feedback is a key for realizing accurate plant protection. The technology collects spectral information or image data of crops through an onboard sensor, combines biological characteristics of plant growth and development, and aims to construct a closed-loop control link from perception to execution. The ideal regulating system can deeply analyze the three-dimensional structure and physiological characteristics of crop canopy, and adaptively regulate the flow, pressure and droplet physical characteristics of the spraying system according to the severity of plant diseases and insect pests recognized in real time and the drug receiving requirements of different parts of crops, so as to ensure the accurate deposition and permeation of the liquid medicine in a target area. The prior art still faces the problem of sensing and decision-making dislocation in practical application, and the existing visual feature extraction model is difficult to accurately capture fine features of early stage disease spots or hidden pests in complex environments such as orchards and the like with strong shadow interference or branch and leaf shielding, so that the target recognition accuracy is insufficient and visual blind spots exist. Meanwhile, the traditional feedback regulation logic excessively depends on macroscopic physical parameters such as flying speed, height and the like, lacks deep fusion of the climatic information of different growth stages of crops, cannot dynamically optimize the particle size of fogdrops according to the structural characteristics of different organs such as fruits, stems and the like, and causes a serious decision gap between the application demands and the execution actions. In addition, the existing flow control algorithm is difficult to cope with the influence caused by the nonlinear state change of crop canopy, so that the pesticide application stability and the operation coverage rate in a complex operation scene are difficult to balance. Disclosure of Invention The invention aims to provide a dynamic feedback regulating system for the drug delivery flow of an unmanned aerial vehicle, which can solve the problems in the background technology. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: The utility model provides an unmanned aerial vehicle flow dynamic feedback governing system that gives medicine to poor free of charge, includes multidimensional agriculture emotion and knows device, physiology thing weather analysis device, degree of depth study decision-making processing system, parameter feedback governing device and unmanned aerial vehicle flight state monitoring devices, wherein: The multi-dimensional agricultural emotion sensing device is used for acquiring multi-source sensing data of a crop canopy below in real time in the unmanned aerial vehicle operation process, the multi-source sensing data comprises multi-spectral characteristic information and high-resolution color image information, and the multi-dimensional agricultural emotion sensing device captures visual characteristics of the crop surface through preset acquisition frequency and transmits the visual characteristics to the deep learning decision processing system; The physiological weather analysis device is used for storing and processing the historical growth archives and the current weather stage characteristics of the crops, judging the current specific physiological stage of the crops by extracting the morphological characteristics of the crops and comparing the morphological characteristics with a preset weather library, and p