A smart device and method for trapping chinch bugs

Abstract

This invention discloses an intelligent trapping device and method for mirid bugs, relating to the field of mirid bug trapping technology. The intelligent trapping device includes a frame and a walking component, a lifting and rotating component, a collecting component, a trapping component, and a photographing component mounted on the frame. The trapping component is mounted on the collecting component and is connected to it. The collecting component is mounted on the lifting and rotating component. The trapping component is used to trap pests. The collecting component includes an inner collecting box and side baffles. The inner collecting box collects the pests trapped by the trapping component. An insect entry channel is provided at the connection between the inner collecting box and the trapping component. The side baffles are slidably connected to the inner collecting box. When the inner collecting box is separated from the trapping component, the side baffles can close the insect entry channel. The lifting and rotating component can drive the collecting component to lift, rotate, and collect images. This invention can increase the trapping range and prevent pests from escaping.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pasture stink bug trapping, in particular to a pasture stink bug intelligent trapping device and a trapping method. BACKGROUND

[0002] As a common pest in the production process of pasture, stink bugs seriously harm the growth and yield of pasture, leading to a decrease in the quality of pasture, and thus affecting the feeding efficiency of the livestock industry. Traditional stink bug control methods mainly rely on manual capture and pesticide spraying, which have many problems such as high work intensity, low efficiency, serious environmental pollution, and increased pest resistance, and thus are difficult to meet the needs of modern agriculture for green, environmentally friendly, and efficient control.

[0003] In recent years, with the development of intelligent technology, automatic trapping devices for agricultural pests have gradually attracted attention. Some existing trapping devices are mostly fixed or have limited mobility, and have limited trapping range, making it difficult to effectively cover a large area of pasture pests.

[0004] In addition, the existing trapping devices also have deficiencies in the collection and management of pests. The design of the collection container is unreasonable, which easily leads to the escape of pests, increases the risk of secondary pollution, and brings inconvenience to subsequent cleaning work. SUMMARY

[0005] The purpose of the present application is to provide a pasture stink bug intelligent trapping device and a trapping method to solve the problems existing in the prior art and to improve the trapping range and prevent pests from escaping.

[0006] To achieve the above-mentioned purpose, the present application provides the following solutions: The present application provides a pasture stink bug intelligent trapping device, comprising a rack, a walking assembly, a lifting and rotating assembly, a collection assembly, a trapping assembly, and a shooting assembly arranged on the rack. The trapping assembly is arranged on the collection assembly and communicates with the collection assembly. The collection assembly is arranged on the lifting and rotating assembly. The trapping assembly is used for trapping pests. The collection assembly comprises a collection inner box and a side baffle. The collection inner box is used for collecting pests trapped by the trapping assembly. An entry channel is arranged at the connection between the collection inner box and the trapping assembly. The side baffle is in sliding connection with the collection inner box. When the collection inner box is separated from the trapping assembly, the side baffle can close the entry channel. The lifting and rotating assembly can drive the collection assembly to lift and rotate. The shooting assembly is used for collecting images.

[0007] In some specific embodiments, the lifting and rotating assembly includes a rotating platform, a rotating drive structure, and a lifting rod. The rotating platform is rotatably connected to the frame, the rotating drive structure can drive the rotating platform to rotate, the lifting rod is disposed on the rotating platform, and the collecting assembly is disposed on the lifting rod.

[0008] In some specific embodiments, the power output end of the rotary drive structure is provided with a drive gear, and a transmission gear ring is provided on the rotary table, with the drive gear meshing with the transmission gear ring.

[0009] In some specific solutions, the collection component further includes a fixed outer box, which is disposed on the lifting and rotating component. The collection inner box is disposed in the fixed outer box. The fixed outer box has an opening corresponding to the insect inlet channel. A limiting part is provided on the side baffle. When the collection inner box is located in the fixed outer box, the limiting part contacts the fixed outer box, so that the insect inlet channel can communicate with the trapping component.

[0010] In some specific embodiments, the trapping assembly includes a trapping drive structure, a trapping tube, a trapping rod, and a pushing blade. One end of the trapping tube is closed, and the other end of the trapping tube is connected to the insect inlet channel. The trapping tube is provided with an insect inlet hole. The trapping rod and the pushing blade are both located in the trapping tube. The pushing blade is spirally arranged on the trapping rod. The trapping drive structure is used to drive the trapping rod to rotate, and the pushing blade pushes the pests in the trapping rod toward the insect inlet channel.

[0011] In some specific designs, the trapping rod is also equipped with a light strip, and the insect inlet hole gradually decreases in size from the outer wall of the trapping tube to the inner wall of the trapping tube.

[0012] In some specific embodiments, the imaging component includes a multispectral camera, which is used to simultaneously acquire and output visible light and near-infrared images of the pasture area; the intelligent trapping device for pasture mirid bugs also includes an image processing unit, which is used for: Perform radiometric calibration and atmospheric correction on the visible light image and the near-infrared band image, and output the corrected multi-band image; The corrected multi-band image is aligned using image registration technology, and the registered multi-band image is output. Based on the obtained normalized differential vegetation index, pasture areas are extracted from the registered multi-band image, a pasture area mask is output, and the texture changes within the pasture area mask are detected by background subtraction method to generate initial environmental monitoring data. The registered multi-band image is input into a trained convolutional neural network model, which outputs a single-channel probability heatmap. The trained convolutional neural network model adopts an encoder-decoder architecture, wherein the encoder part uses a pre-trained residual network as the backbone and includes multiple downsampling stages, and the decoder part includes multiple upsampling stages. Each upsampling stage is skip-connected to the high-resolution feature map output by the corresponding stage of the encoder, and a channel attention module is embedded in the skip connection. Each pixel value of the single-channel probability heatmap represents the probability that there are mirid bug damage features at the corresponding location. Based on the single-channel probability heatmap, generate mirid bug density levels and hazard area distribution characteristic values; Microclimate parameters are obtained from the registered multi-band images using image derivation technology; the microclimate parameters include leaf surface temperature and light intensity. By using the Kalman filter algorithm to fuse the mirid bug density level, the distribution characteristics of the affected area, the microclimate parameters, and the initial environmental monitoring data, a unified environmental state estimation vector is constructed. The environmental state estimation vector is input into a preset Bayesian network, which outputs a blind bug risk map and an environmental anomaly alarm.

[0013] In some specific embodiments, the intelligent trapping device for pasture mirid bugs further includes a reinforcement learning controller, which is used for: Based on the environmental state estimation vector, the trapping strategy is optimized online using the Q-learning algorithm to obtain the optimal action combination. The optimal action combination includes control commands to adjust the spectral band and pulse frequency of the light strip on the trapping component, control commands to adjust the rotation speed of the trapping rod on the trapping component, and control commands to dynamically plan the lifting height of the electric telescopic pole based on the obtained normalized difference vegetation index change and terrain data.

[0014] In some specific embodiments, the imaging component is also used to acquire and output real-time images of the grassland ahead during the movement of the intelligent trapping device for pasture mirid bugs; the device also includes a path planning and navigation control unit, which is used for: Using visual simultaneous localization and mapping technology, feature point matching and motion estimation are performed on the image of the grass ahead to construct a two-dimensional grid map of the driving area; An incremental dense reconstruction algorithm is used to fuse the depth and color information of the foreground grassland image to generate a two-dimensional elevation raster map; each raster of the two-dimensional elevation raster map stores terrain height, passability probability, and uncertainty value; A fully convolutional neural network is used to perform terrain semantic segmentation on the grassland image in front to obtain semantic segmentation results, and the semantic segmentation results are projected onto the two-dimensional grid map to obtain a semantic cost map; Based on the acquired current location and target waypoint, an improved A search algorithm is run on the semantic cost map to obtain an initial global path. Then, based on the real-time acquired moving obstacle information, the initial global path is locally adjusted to obtain an adjusted global path. The improved A search algorithm introduces the terrain slope factor calculated based on the two-dimensional elevation raster map and the grass density factor quantified based on the normalized difference vegetation index as additional weights of the heuristic function to search for the initial global path with the shortest path and optimal energy. Based on the initial global path and the preset track wheel kinematic model constraints, the linear velocity control command and angular velocity control command are obtained by solving. By extracting regions with stable semantic features from the two-dimensional raster map as natural beacons; Determine whether the natural beacon has been detected; If so, the current pose estimate is calculated based on the natural beacon and the image of the grassland ahead. Based on the image of the grassland ahead, information about sudden obstacles is identified using optical flow and a deep learning object detection network. Based on the information about sudden obstacles, the linear velocity control command and the angular velocity control command are optimized to obtain updated linear velocity and updated angular velocity control commands. The updated linear velocity and updated angular velocity control commands are used to control the movement of the intelligent trapping device for pasture mirid bugs.

[0015] This invention also provides a method for trapping mirid bugs using an intelligent trapping device, comprising the following steps: Step 1: Activate the walking component to move the intelligent trapping device for pasture mirid bugs in the area to be trapped, and use the camera component to monitor the environment in real time and perform road guidance and recognition; Step 2: Activate the lifting and rotating assembly, which will lift the collection and trapping assemblies to the top of the pasture, and the lifting and rotating assembly will then rotate. Step 3: Control the trapping component to attract pests to approach and enter the trapping component, and push the pests inside the trapping component into the collection box. Step 4: When the collection box reaches the set collection capacity or the trapping time ends, the lifting and rotating component stops rotating and the trapping component stops working; Step 5: The operator removes the inner collection box, and the side panel closes the inner collection box, completing the collection and cleaning of pests.

[0016] The present invention achieves the following technical effects compared to the prior art: This invention uses a walking component to propel the device, and then a lifting and rotating component to raise, lower, and rotate the collection and trapping components, allowing adjustment of the trapping height and range, thus improving trapping efficiency. Furthermore, once pests enter the collection box, the side baffles close the box when it is removed, preventing pests from escaping and reducing the risk of secondary contamination. In addition, this invention improves the capture efficiency of mirid bugs, reduces manual labor intensity, and achieves green and environmentally friendly intelligent pest control. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of an intelligent trapping device for mirid bugs in some embodiments of the present invention; Figure 2 This is a schematic diagram illustrating the adjustment of the trapping height of the intelligent trapping device for pasture mirid bugs in some embodiments of the present invention. Figure 3 This is a schematic diagram of the collection components in some embodiments of the present invention; Figure 4 This is a schematic diagram of the collection inner box in some embodiments of the present invention; Figure 5 This is a schematic diagram showing the side baffles blocking the inner box on both sides in some embodiments of the present invention. Figure 6 This is a half-sectional schematic diagram of the collection component and the trapping component in some embodiments of the present invention. Figure 7 This is a schematic diagram of the trapping component in some embodiments of the present invention; Figure 8 This is a schematic diagram of the internal structure of the trapping component in some embodiments of the present invention; Figure 9 This is a schematic diagram of a rotation drive structure in some embodiments of the present invention; Figure 10 This is a schematic diagram of a trapping rod in some embodiments of the present invention; Figure 11 This is a schematic diagram of the cross-section of the trapping rod in some embodiments of the present invention; Figure 12 This is a flowchart of the trapping method in some embodiments of the present invention; In the diagram: 1. Frame; 2. Rotary platform; 3. Lifting rod; 4. Collection assembly; 5. Trapping assembly; 6. Rotation drive structure; 7. Walking assembly; 8. Imaging assembly; 41. Secure the outer box; 42. Collect the inner box; 43. Insect entry channel; 421. Side panel; 422. Handle; 51. Trapping tube; 52. Trapping rod; 53. Pushing blade; 54. Second geared motor; 55. LED strip; 56. Wire; 57. Battery; 551. Insect entry hole; 61. Mounting bracket; 62. First geared motor; 63. Drive gear; 64. Transmission gear ring. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] The purpose of this invention is to provide an intelligent trapping device and method for mirid bugs in pasture, so as to solve the problems existing in the prior art, improve the trapping range, and prevent pests from escaping.

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Example 1 like Figures 1 to 11As shown, this embodiment provides an intelligent trapping device for mirid bugs in pasture, including: a frame 1 and a walking component 7, a lifting and rotating component, a collecting component 4, a trapping component 5, and a shooting component 8 mounted on the frame 1. The trapping component 5 is mounted on the collecting component 4 and is connected to the collecting component 4. The collecting component 4 is mounted on the lifting and rotating component. The trapping component 5 is used to trap pests. The collecting component 4 includes a collecting inner box 42 and two side baffles 421. The collecting inner box 42 is used to collect the pests trapped by the trapping component 5. An insect inlet channel 43 is provided at the connection between the collecting inner box 42 and the trapping component 5. The two side baffles 421 are located at both ends of the collecting inner box 42 and are slidably connected to the collecting inner box 42. When the collecting inner box 42 is separated from the trapping component 5, the side baffles 421 can close the insect inlet channel 43. The lifting and rotating component can drive the collecting component 4 to lift and rotate. The shooting component 8 is used to collect images. In this embodiment, the walking component 7 drives the device to move, and then the lifting and rotating component drives the collecting component 4 and the trapping component 5 to rise, fall, and rotate, which can adjust the trapping height and range, thereby improving the trapping ability. Furthermore, when pests enter the collecting inner box 42, the side baffle 421 can close the collecting inner box 42 when it is removed, preventing pests from escaping and reducing the risk of secondary contamination. In addition, this embodiment improves the capture efficiency of mirid bugs, reduces manual labor intensity, and achieves green and environmentally friendly intelligent pest control.

[0023] In some specific embodiments, the lifting and rotating assembly includes a rotating platform 2, a rotating drive structure 6, and a lifting rod 3. The rotating platform 2 is located at the center of the top of the frame 1 and is rotatably connected to the frame 1. The rotating drive structure 6 includes a first reduction motor 62 and is fixed to the rear of the top of the frame 1 by a mounting bracket 61. The rotating drive structure 6 can drive the rotating platform 2 to rotate 360°. The lifting rod 3 is an electric telescopic rod and is located at the center of the top of the rotating platform 2. The collecting assembly 4 is located on the top of the lifting rod 3.

[0024] In some specific embodiments, the power output end of the rotary drive structure 6 is provided with a drive gear 63, and a transmission gear ring 64 is fixedly sleeved on the rotary table 2. The drive gear 63 meshes with the transmission gear ring 64. The rotary drive structure 6 drives the drive gear 63 to rotate, and the drive gear 63 drives the rotary table 2 to rotate through meshing with the transmission gear ring 64, so that the device can rotate when it is not moving, thereby expanding the trapping range.

[0025] In some specific embodiments, the collection component 4 further includes a fixed outer box 41, which is fixedly disposed on the top of the lifting rod 3 of the lifting and rotating component. The collection inner box 42 is disposed in the fixed outer box 41 and is snapped into the fixed outer box 41. A handle 422 is fixedly disposed on the top of the collection inner box 42. The fixed outer box 41 is provided with an opening corresponding to the insect inlet channel 43. A limiting part is provided on the side baffle 421. When the collection inner box 42 is located in the fixed outer box 41, the limiting part contacts the fixed outer box 41, so that the insect inlet channel 43 can communicate with the trapping component 5. The side baffle 421 is slidably connected to the inner collection box 42 in the vertical direction. When the inner collection box 42 is placed downward into the fixed outer box 41, the side baffle 421 opens both ends of the fixed inner box under the action of the limiting part, so that the insect inlet channel 43 is connected to the trapping component 5. When the inner collection box 42 is taken out upward from the fixed outer box 41, the side baffle 421 closes both ends of the inner collection box 42 under the action of gravity, thereby closing the insect inlet channel 43 and preventing the pests in the inner collection box 42 from escaping.

[0026] In some specific embodiments, a trapping component 5 is provided at each end of the collection component 4. The trapping component 5 includes a trapping drive structure, a trapping tube 51, a trapping rod 52, and a pusher blade 53. One end of the trapping tube 51 is closed, and the other end of the trapping tube 51 is connected to the insect inlet channel 43. A plurality of insect inlet holes 551 are evenly provided on the trapping tube 51. The trapping rod 52 and the pusher blade 53 are both located in the trapping tube 51. The trapping rod 52 is coaxially arranged with the trapping tube 51. The pusher blade 53 is spirally fixed on the trapping rod 52. The trapping drive structure includes a second reduction motor 54. The trapping drive structure is installed at one end of the trapping tube 51 and is located outside the trapping tube 51. The trapping drive structure is used to drive the trapping rod 52 to rotate. The pusher blade 53 pushes the pests in the trapping rod 52 toward the insect inlet channel 43 and finally into the collection inner box 42.

[0027] In some specific embodiments, a light strip 55 is also wrapped around the surface of the trapping rod 52. The light strip 55 is located in the gap of the pusher blade 53. The light strip 55 is preferably an LED light strip. The light strip 55 is connected to the battery 57 through the wire 56 to realize the power supply of the light strip 55. The battery 57 is embedded in the end face of the trapping rod 52. The insect inlet hole 551 gradually decreases from the outer wall of the trapping tube 51 to the inner wall of the trapping tube 51. That is, the insect inlet hole 551 is a conical hole with a larger outer diameter and a smaller inner diameter, which is conducive to the entry of pests and makes it difficult for them to escape.

[0028] In some specific embodiments, a walking component 7 is provided on each side of the frame 1. The walking component 7 is a tracked walking structure, and the track wheels of the tracked walking structure are provided on the frame 1. In this embodiment, the tracked walking structure ensures the stable movement capability of the device on the pasture.

[0029] In some specific embodiments, the shooting component 8 is a camera, which is located at the front center of the top of the frame 1, and is used for environmental monitoring and path recognition.

[0030] The intelligent mirid bug trapping device in this embodiment uses a tracked walking structure to drive the frame 1 to move stably on the pasture. Combined with the top rotating platform 2, it achieves 360° rotation, greatly expanding the trapping range and improving the capture efficiency of mirid bugs. The lifting rod 3 drives the collection component 4 and the trapping components 5 on both sides to the top of the pasture, accurately locking onto the mirid bug activity area and improving the targeting and effectiveness of the trapping. The trapping component 5 features a cone-shaped insect inlet 551 with a larger outer diameter and a smaller inner diameter, effectively preventing mirid bugs from escaping. Combined with the coordinated work of the spiral pushing blade 53 and the trapping rod 52, it achieves continuous pushing and concentrated collection of mirid bugs, ensuring a highly efficient and stable trapping process. The collection inner box 42 is equipped with an automatically sliding side baffle 421, which automatically blocks the insect inlet channel 43 when the inner box is pulled out, preventing mirid bugs from escaping and facilitating safe cleaning and management. Furthermore, the camera module 8 mounted on the top front of the device not only monitors the environment but also assists in path recognition and navigation, enabling the device to move precisely in complex pasture environments and ensuring the continuity and stability of the trapping operation. The overall structure is reasonable and compact, with a high degree of automation and simple operation, greatly reducing the intensity of manual labor.

[0031] This embodiment, through multiple technological innovations and rational design, achieves intelligent, efficient trapping and convenient management of mirid bugs in forage grasses, demonstrating significant practical value and promising prospects for widespread application. This device not only effectively reduces the damage caused by mirid bugs to forage grasses, increasing forage yield and quality, but also provides a new technological means for intelligent pest control in agriculture, promoting the development of modern agriculture towards intelligence and precision.

[0032] In some specific embodiments, a camera component 8 is installed at the center of the front end of the top of the frame 1 for environmental monitoring and path recognition.

[0033] In terms of environmental monitoring, the imaging component 8 includes a multispectral camera that integrates multispectral imaging and real-time image analysis algorithms to simultaneously acquire and output visible light and near-infrared images of the pasture area, thereby enabling dynamic monitoring of pasture growth status, mirid bug aggregation density, and surrounding environmental factors. The device also includes an image processing unit that uses a trained convolutional neural network model to analyze the color, texture, and morphological characteristics of the captured pasture leaves, automatically identifying the mirid bug damage level and activity hotspots. Simultaneously, it can monitor microclimate parameters such as temperature, humidity, and light intensity, and through data fusion technology, correlate environmental information with trapping status, autonomously adjusting the trapping strategy to achieve precise trapping and intelligent decision-making based on environmental feedback.

[0034] Specifically, in terms of multispectral imaging and real-time image analysis, the imaging component 8 integrates a multispectral sensor to simultaneously acquire images in the visible and near-infrared bands, and runs real-time image analysis algorithms through a processor. The image processing unit performs the following operations: First, radiometric calibration and atmospheric correction are performed on the visible light image and the near-infrared band image to eliminate illumination changes and atmospheric interference, and the corrected multi-band image is output.

[0035] Then, image registration technology is used to align the corrected multi-band images, and the registered multi-band images are output. Grassland regions are extracted from the registered multi-band images based on the Normalized Difference Vegetation Index (NDVI), and a grassland region mask is output. In the threshold-based segmentation algorithm, the image segmentation threshold is set specifically for the NDVI image, with a threshold of 0.25. This value is an optimized parameter determined through statistical analysis of a large number of grass multispectral image samples, effectively distinguishing between vegetated and non-vegetated areas. In the NDVI image, areas with pixel values ​​greater than 0.25 are identified as grass, while areas below this threshold are considered soil, shadows, or other non-vegetated backgrounds. This threshold is chosen based on the high reflectivity of grass in the near-infrared band and the low reflectivity in the red band, ensuring robustness of segmentation in typical grassy environments.

[0036] Next, texture changes within the mask of the pasture area are detected using background subtraction to generate initial environmental monitoring data. Simultaneously, the registered multi-band image is input into a trained convolutional neural network (CNN) model. This trained CNN model is an improved CNN model based on an encoder-decoder architecture, with an overall structure containing symmetrical encoding and decoding paths, totaling approximately 62 trainable layers. The encoder part uses a pre-trained residual network (such as ResNet34) as its backbone, comprising five downsampling stages, consisting of convolutional layers, batch normalization layers, ReLU activation layers, and residual blocks, used to progressively extract multi-scale deep semantic features from the input image. The decoder part comprises five upsampling stages, gradually restoring spatial resolution through five upsampling operations. Each upsampling stage is skip-connected to the high-resolution feature map of the corresponding stage in the encoder to fuse contextual information, and a channel attention module is embedded in the skip connections to weightedly select features important for mirid bug identification. The network ends with a 1x1 convolutional layer and a sigmoid activation function, outputting a single-channel probability heatmap with the same resolution as the input image. Each pixel value represents the probability of mirid bug damage at that location. Based on this probability heatmap, mirid bug density levels and damage area distribution feature values ​​are generated and output.

[0037] Furthermore, microclimate parameters, including leaf surface temperature and light intensity, are obtained from the registered multi-band images using image derivation technology. Specifically, leaf surface temperature is retrieved from thermal infrared images using infrared thermometry principles, and light intensity is estimated by combining it with visible light image brightness. A trained regression model is then used to infer the ambient temperature and humidity.

[0038] In the process of environmental parameter monitoring and data fusion, the Kalman filter algorithm is used to fuse the mirid bug density level, the characteristic values ​​of the distribution of hazardous areas, microclimate parameters, and the initial environmental monitoring data to construct and output a unified environmental state estimation vector. Then, using the environmental state estimation vector as input, a Bayesian network is used to model the correlation between environmental factors and the probability of mirid bug occurrence, dynamically update prior knowledge, and output mirid bug risk maps and environmental anomaly alerts in real time, providing multi-source information support for intelligent decision-making.

[0039] The trapping status refers to the set of parameters directly related to the trapping operation collected in real time by the intelligent trapping device for mirid bugs during operation. Specifically, this includes the working mode of the trapping components (rotation speed of the trapping rod, emission spectrum and flashing frequency of the LED light strip, rotation status of the spiral pusher blades) and the mechanical execution status of the device (extension height of the electric telescopic rod, rotation angle and speed of the rotary drive component, and movement trajectory of the tracked wheels). This status data is acquired in real time through built-in sensors (such as photoelectric sensors, Hall effect sensors, and ammeters) and integrated with environmental monitoring information (mirid bug density heatmap, temperature and humidity, and light intensity) to assess the current trapping efficiency and dynamically optimize the trapping strategy for adaptive control.

[0040] The adaptive trapping strategy adjustment mechanism is based on the aforementioned monitoring results, and the device also includes a reinforcement learning controller. The reinforcement learning controller takes the environmental state estimation vector (including the mirid bug density level) as input and uses a Q-learning algorithm to optimize the trapping strategy online. The state space of the reinforcement learning controller is defined as a continuous multi-dimensional vector, specifically including an environmental state vector (temperature, humidity, light intensity), a pest state vector (mirid bug density level and hazard area distribution characteristics based on real-time output from a convolutional neural network), and a device state vector (current trapping height), totaling 12 features. The action space is designed as a discrete combination space, including adjustment options for four controllable parameters: the spectral mode of the LED light strip (divided into three levels: ultraviolet, blue-green, and full spectrum), the flashing frequency (three levels: low, medium, and high), the rotation speed of the trapping stick (three levels: slow, medium, and fast), and the trapping operation cycle duration (three levels: short, medium, and long), forming a total of 81 discrete action combinations. The reward function employs a weighted multi-objective design, specifically in the form R = α·(capture increment) - β·(energy cost) - β·(time penalty) + δ·(exploration reward). Here, the capture increment is measured by the change in the number of blind bugs entering the box per unit time, estimated through the fusion of the imaging component and the sensor. The energy cost is calculated based on the total power consumption of each actuator. The time penalty encourages quickly reaching an effective capture state, and the exploration reward encourages exploration of untried actions during the initial training phase. The controller outputs an optimal action combination based on the reward function. This optimal action combination includes: control commands to adjust the spectral band and pulse frequency of the light strip (specifically, a light-emitting diode light strip) on the trapping component 5; control commands to adjust the rotation speed of the trapping rod on the trapping component 5; and control commands to dynamically plan the lifting height of the electric telescopic pole based on changes in the normalized difference vegetation index and terrain data. When the density of mirid bugs exceeds a threshold, the controller adjusts the LED light strip to a specific spectral band and modulates the pulse frequency. At the same time, it adjusts the rotation speed of the trapping rod to match the phototactic behavior pattern of the mirid bugs. Combined with the temperature and humidity prediction model, it automatically extends the operation cycle under suitable conditions. Meanwhile, the system dynamically plans the lifting height of the electric telescopic rod based on the changes in the normalized difference vegetation index and terrain data. This ensures that the trapping components accurately locate mirid bug habitat hotspots, achieving fully autonomous, adaptive, and efficient closed-loop control for trapping, thereby achieving the highest capture efficiency per unit of energy consumption.

[0041] Regarding path recognition and navigation control, the imaging component 8 is also used to acquire and output real-time images of the grass ahead during device movement. The device also includes a path planning and navigation control unit, which performs the following operations: First, based on visual simultaneous localization and mapping (FSMR) technology, feature point matching and motion estimation are performed on the foreground grassland image to construct and output a two-dimensional grid map of the driving area. During map construction, the camera module 8 processes the video stream in real time using ORB feature point detection and descriptor extraction, calculates the essential matrix between adjacent frames using multi-view geometry principles, estimates camera motion, optimizes the three-dimensional coordinates of feature points and camera pose using bundle adjustment, and simultaneously constructs a sparse map containing feature point clouds and keyframes.

[0042] Then, an incremental dense reconstruction algorithm is used to fuse the depth and color information of the grassland image in front to generate a two-dimensional elevation grid map and output it. Each grid of the two-dimensional elevation grid map stores the terrain height, passability probability and uncertainty value, and uses the bag-of-words model to perform loop closure detection to correct the accumulated error, and finally forms a real-time dynamic map that can be used for navigation.

[0043] Next, a fully convolutional neural network is used to perform terrain semantic segmentation on the foreground grassland image, outputting the semantic segmentation result, which is the terrain category probability for each pixel. This network is pre-trained on a multispectral dataset containing pasture, bare soil, rocks, water bodies, and artificial obstacles, extracting multi-level features and preserving spatial details through an encoder-decoder structure. During inference, the network performs pixel-level classification on the RGB-NIR images captured by the camera, outputting the terrain category probability for each pixel. In the post-processing stage, a conditional random field is used to optimize the segmentation boundary, and the semantic results are projected onto a raster map constructed using SLAM to generate a semantic cost map, which is then output. Each raster is assigned terrain attributes and passage cost weights.

[0044] During path planning and navigation control, an improved A-search algorithm is run on the semantic cost map, using the current position and target waypoint as input. This improved A-search algorithm introduces terrain slope factors calculated from the two-dimensional elevation grid map and grass density factors quantified using the normalized difference vegetation index as additional weights in the heuristic function to search for and output the initial global path that is shortest and most energy-optimal. Then, based on real-time detected moving obstacle information, the initial global path is locally adjusted to obtain the adjusted global path. Based on the initial global path and preset track wheel kinematic model constraints, linear and angular velocity control commands that satisfy the track wheel kinematic model are generated by solving a constrained optimization problem to drive the frame movement.

[0045] During autonomous obstacle avoidance, regions with stable semantic features are extracted from the constructed 2D grid map as natural beacons. When the imaging component observes these natural beacons again, it determines whether the natural beacons have been detected. If so, based on the natural beacons and the real-time detected foreground grassland image, the deviation between the current position and the 2D grid map is calculated by solving the PnP problem, and extended Kalman filtering is used to fuse multi-source data to obtain the optimal current pose estimate. The autonomous obstacle avoidance module detects moving objects in the foreground using optical flow and identifies sudden obstacle information using a deep learning object detection network. It optimizes the linear velocity control command and the angular velocity control command to obtain updated linear velocity and angular velocity control commands. The optimization operation involves updating the local cost map in the semantic cost map in real time and triggering path replanning. At the same time, it controls the differential speed of the track wheels to achieve in-situ turning or deceleration to bypass obstacles, outputting updated linear velocity and angular velocity control commands to ensure that the device can safely and stably perform full-coverage patrol missions in complex terrain.

[0046] Example 2 like Figure 12 As shown, this embodiment provides a method for trapping mirid bugs using the intelligent trapping device of Embodiment 1, including the following steps: Step 1: Activate the walking component 7 to make the intelligent trapping device for pasture grass bugs move slowly back and forth in the area to be trapped. The camera component 8 monitors the environment in real time and performs road guidance and identification to ensure that the device runs stably along the predetermined path. The walking component 7 stops working after the intelligent trapping device for pasture grass bugs moves into place. Step 2: Activate the lifting and rotating assembly to raise the collection assembly 4 and the trapping assembly 5 to the top of the pasture to facilitate the trapping of mirid bugs. The lifting and rotating assembly will then rotate. Step 3: The trapping drive structure drives the trapping rod 52 to rotate, and at the same time the light strip 55 is lit to attract mirid bugs to approach. The mirid bugs are induced to enter the trapping tube 51 through the evenly distributed conical insect entry holes 551 with a larger outer diameter and a smaller inner diameter. Once inside the trapping tube 51, the mirid bugs are pushed to the insect entry channel 43 by the rotation of the pusher leaf 53, and then enter the collection inner box 42. Step 4: When the collection box 42 reaches the set collection capacity or the trapping time ends, the rotary drive structure 6 and the trapping drive structure stop working, and the lifting rod 3 retracts the collection component 4 to a state that is easy to remove. Step 5: The operator removes the inner collection box 42, and the side baffle 421 closes the inner collection box 42, effectively preventing the mirid bugs from escaping, thus completing the collection and cleaning of the mirid bugs.

[0047] In the description of this invention, it should be understood that the terms "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this invention, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. Furthermore, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0048] In the description of this invention, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0049] If this invention discloses or relates to components or structural parts that are fixedly connected to each other, then, unless otherwise stated, a fixed connection can be understood as: a detachable fixed connection (e.g., using bolts or screws) or a non-detachable fixed connection (e.g., riveting, welding). Of course, a fixed connection can also be replaced by an integral structure (e.g., manufactured in one piece using a casting process) (except where it is obviously impossible to use an integral molding process).

[0050] In addition, unless otherwise stated, the terms used in any of the technical solutions disclosed in this invention to indicate positional relationships or shapes include states or shapes that are similar to, close to, or approximate with those states or shapes.

[0051] Any component provided by this invention can be assembled from multiple individual components or can be a single component manufactured by a one-piece molding process.

[0052] It should be noted that the structures, proportions, sizes, etc., depicted in the accompanying drawings of this specification are only used to complement the content disclosed in the specification, so as to enable those skilled in the art to understand and read them, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0053] It should also be noted that in the embodiments of this application, the same reference numerals are used to denote the same component or the same part.

[0054] Any adaptive changes made according to actual needs are within the scope of protection of this invention.

[0055] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A smart trapping device for mirid bugs in pasture, characterized in that: include: The frame (1) and a walking assembly (7), a lifting and rotating assembly, a collecting assembly (4), a trapping assembly (5), and a shooting assembly (8) are mounted on the frame (1). The trapping assembly (5) is mounted on the collecting assembly (4) and is connected to the collecting assembly (4). The collecting assembly (4) is mounted on the lifting and rotating assembly. The trapping assembly (5) is used to trap pests. The collecting assembly (4) includes a collecting inner box (42) and a side baffle (421). 42) Used to collect the pests trapped by the trapping component (5), the collection inner box (42) is provided with an insect inlet channel (43) at the connection between it and the trapping component (5), the side baffle (421) is slidably connected to the collection inner box (42), when the collection inner box (42) is separated from the trapping component (5), the side baffle (421) can close the insect inlet channel (43), the lifting and rotating component can drive the collection component (4) to lift and rotate, and the shooting component (8) is used to collect images.

2. The intelligent trapping device for mirid bugs in pastures according to claim 1, characterized in that: The lifting and rotating assembly includes a rotating platform (2), a rotating drive structure (6), and a lifting rod (3). The rotating platform (2) is rotatably connected to the frame (1). The rotating drive structure (6) can drive the rotating platform (2) to rotate. The lifting rod (3) is disposed on the rotating platform (2). The collecting assembly (4) is disposed on the lifting rod (3).

3. The intelligent trapping device for mirid bugs in pastures according to claim 2, characterized in that: The power output end of the rotary drive structure (6) is provided with a drive gear (63), and the rotary table (2) is provided with a transmission gear ring (64), and the drive gear (63) meshes with the transmission gear ring (64).

4. The intelligent trapping device for mirid bugs in pastures according to claim 1, characterized in that: The collection component (4) further includes a fixed outer box (41), which is disposed on the lifting and rotating component. The collection inner box (42) is disposed in the fixed outer box (41). The fixed outer box (41) has an opening corresponding to the insect inlet channel (43). The side baffle (421) has a limiting part. When the collection inner box (42) is located in the fixed outer box (41), the limiting part contacts the fixed outer box (41), so that the insect inlet channel (43) can communicate with the trapping component (5).

5. The intelligent trapping device for mirid bugs in pastures according to claim 1, characterized in that: The trapping assembly (5) includes a trapping drive structure, a trapping tube (51), a trapping rod (52), and a pusher blade (53). One end of the trapping tube (51) is closed, and the other end of the trapping tube (51) is connected to the insect inlet channel (43). An insect inlet hole (551) is provided on the trapping tube (51). The trapping rod (52) and the pusher blade (53) are both located in the trapping tube (51). The pusher blade (53) is spirally arranged on the trapping rod (52). The trapping drive structure is used to drive the trapping rod (52) to rotate. The pusher blade (53) pushes the pests in the trapping rod (52) towards the insect inlet channel (43).

6. The intelligent trapping device for mirid bugs in pastures according to claim 5, characterized in that: The trapping rod (52) is also equipped with a light strip (55), and the insect inlet (551) gradually decreases in size from the outer wall of the trapping tube (51) to the inner wall of the trapping tube (51).

7. The intelligent trapping device for mirid bugs in pastures according to claim 6, characterized in that: The imaging component (8) includes a multispectral camera, which is used to simultaneously acquire and output visible light and near-infrared images of the pasture area; the intelligent trapping device for pasture mirid bugs also includes an image processing unit, which is used for: Perform radiometric calibration and atmospheric correction on the visible light image and the near-infrared band image, and output the corrected multi-band image; The corrected multi-band image is aligned using image registration technology, and the registered multi-band image is output. Based on the obtained normalized differential vegetation index, pasture areas are extracted from the registered multi-band image, a pasture area mask is output, and the texture changes within the pasture area mask are detected by background subtraction method to generate initial environmental monitoring data. The registered multi-band image is input into a trained convolutional neural network model, which outputs a single-channel probability heatmap. The trained convolutional neural network model adopts an encoder-decoder architecture, wherein the encoder part uses a pre-trained residual network as the backbone and includes multiple downsampling stages, and the decoder part includes multiple upsampling stages. Each upsampling stage is skip-connected to the high-resolution feature map output by the corresponding stage of the encoder, and a channel attention module is embedded in the skip connection. Each pixel value of the single-channel probability heatmap represents the probability that there are mirid bug damage features at the corresponding location. Based on the single-channel probability heatmap, generate mirid bug density levels and hazard area distribution characteristic values; Microclimate parameters are obtained from the registered multi-band images using image derivation technology; the microclimate parameters include leaf surface temperature and light intensity. By using the Kalman filter algorithm to fuse the mirid bug density level, the distribution characteristics of the affected area, the microclimate parameters, and the initial environmental monitoring data, a unified environmental state estimation vector is constructed. The environmental state estimation vector is input into a preset Bayesian network, which outputs a blind bug risk map and an environmental anomaly alarm.

8. The intelligent trapping device for mirid bugs in pastures according to claim 7, characterized in that: The intelligent trapping device for mirid bugs in pasture also includes a reinforcement learning controller, which is used for: Based on the environmental state estimation vector, the trapping strategy is optimized online using the Q-learning algorithm to obtain the optimal action combination. The optimal action combination includes control commands to adjust the spectral band and pulse frequency of the light strip (55) on the trapping component (5), control commands to adjust the rotation speed of the trapping rod (52) on the trapping component (5), and control commands to dynamically plan the lifting height of the electric telescopic pole based on the obtained normalized difference vegetation index change and terrain data.

9. The intelligent trapping device for mirid bugs in pastures according to claim 8, characterized in that: The imaging component (8) is also used to acquire and output real-time images of the grassland in front of the device during the movement of the intelligent trapping device for mirid bugs; the device also includes a path planning and navigation control unit, which is used for: Using visual simultaneous localization and mapping technology, feature point matching and motion estimation are performed on the image of the grass ahead to construct a two-dimensional grid map of the driving area; An incremental dense reconstruction algorithm is used to fuse the depth and color information of the foreground grassland image to generate a two-dimensional elevation raster map; each raster of the two-dimensional elevation raster map stores terrain height, passability probability, and uncertainty value; A fully convolutional neural network is used to perform terrain semantic segmentation on the grassland image in front to obtain semantic segmentation results, and the semantic segmentation results are projected onto the two-dimensional grid map to obtain a semantic cost map; Based on the acquired current location and target waypoint, an improved A search algorithm is run on the semantic cost map to obtain an initial global path. Then, based on the real-time acquired moving obstacle information, the initial global path is locally adjusted to obtain an adjusted global path. The improved A search algorithm introduces the terrain slope factor calculated based on the two-dimensional elevation raster map and the grass density factor quantified based on the normalized difference vegetation index as additional weights of the heuristic function to search for the initial global path with the shortest path and optimal energy. Based on the initial global path and the preset track wheel kinematic model constraints, the linear velocity control command and angular velocity control command are obtained by solving. By extracting regions with stable semantic features from the two-dimensional raster map as natural beacons; Determine whether the natural beacon has been detected; If so, the current pose estimate is calculated based on the natural beacon and the image of the grassland ahead. Based on the image of the grassland ahead, information about sudden obstacles is identified using optical flow and a deep learning object detection network. Based on the information about sudden obstacles, the linear velocity control command and the angular velocity control command are optimized to obtain updated linear velocity and updated angular velocity control commands. The updated linear velocity and updated angular velocity control commands are used to control the movement of the intelligent trapping device for pasture mirid bugs.

10. A method for trapping mirid bugs using the intelligent trapping device of any one of claims 1-9, characterized in that: Includes the following steps: Step 1: Activate the walking component (7) to move the intelligent trapping device for pasture mirid bugs in the area to be trapped, and use the shooting component (8) to monitor the environment in real time and perform road guidance identification; Step 2: Activate the lifting and rotating assembly to lift the collection assembly (4) and the trapping assembly (5) to the top of the pasture, and the lifting and rotating assembly will rotate. Step 3: Control the trapping component (5) to attract pests to approach and enter the trapping component (5), and push the pests that have entered the trapping component (5) into the collection box (42); Step 4: When the collection box (42) reaches the set collection capacity or the trapping time ends, the lifting and rotating component stops rotating and the trapping component (5) stops working; Step 5: The operator removes the inner collection box (42), and the side panel (421) closes the inner collection box (42), completing the collection and cleaning of pests.

Images