End effector with mounting bracket for picking

Abstract

The application relates to the technical field of agricultural robots, in particular to a picking end effector with a mounting rack, which comprises a mounting rack, a multispectral camera module, a bionic flexible gripper, a shearing device and a micro spray head. The mounting rack is connected with a mechanical arm through a magnetic quick-change mechanism and can be quickly disassembled and assembled; the multispectral camera module realizes fruit maturity judgment and pest and disease identification in combination with an improved RTMDet-ECA algorithm; the bionic flexible gripper is made of silica gel material and realizes lossless picking; the shearing device realizes millimeter-level positioning and accurate shearing of fruit stems based on dynamic shearing control; and the micro spray head realizes targeted pesticide application, and the utilization rate of pesticide liquid is increased to 85%. Through modular design, the application integrates the functions of inspection, picking and pesticide application, significantly reduces the labor cost and improves the agricultural production efficiency.

Description

Technical Field

[0001] This utility model belongs to the field of agricultural robot technology, specifically a multi-functional end effector suitable for facility agriculture scenarios. Background Technology

[0002] In modern agricultural production, the rapid development of facility agriculture has placed higher demands on efficient, precise, and integrated operating equipment. However, current fruit and vegetable harvesting operations still face many challenges, especially given the labor shortage and high labor costs, where traditional harvesting methods have significant shortcomings.

[0003] Traditional harvesting methods rely heavily on manual labor, which is not only inefficient but also leads to a high fruit damage rate, typically between 15% and 30%. Furthermore, operational decisions heavily depend on human experience, lacking scientific rigor and consistency, making it difficult to meet the precision and intelligence demands of modern agriculture. While existing agricultural robots alleviate some of these problems, their functions are limited, often only capable of performing one task among inspection, harvesting, or pesticide application, failing to achieve fully autonomous operation and becoming a key bottleneck restricting the development of smart agriculture. To address these issues, existing technologies attempt to improve the performance of agricultural robots by introducing multimodal perception and intelligent algorithms, but significant limitations remain in practical applications. For example, in fruit recognition and localization, current technologies lack the ability to extract features from small targets in complex, occluded environments, resulting in a high false negative rate, and the accuracy of fruit stem localization is difficult to achieve at the millimeter level. Simultaneously, dynamic shearing control technology is not yet mature, easily causing fruit damage during harvesting, especially failing to meet the need for non-destructive harvesting of soft fruits such as strawberries. Furthermore, existing agricultural robots have a low level of intelligence in task planning and allocation, relying excessively on human intervention and failing to achieve full autonomy from "perception" to "decision-making," thus limiting their promotion and application in actual agricultural production. In summary, the current agricultural robotics field urgently needs a multi-functional end effector that integrates inspection, harvesting, and pesticide application to address prominent issues such as high fruit damage rates, reliance on manual decision-making, and limited functionality. Especially in complex agricultural environments, how to achieve high-precision fruit positioning, low-damage harvesting operations, and efficient targeted pesticide application through technological innovation is a core challenge driving the advancement of agricultural robot technology and the development of smart agriculture.

[0004] Based on this background, the present invention aims to provide an end effector with a mounting frame for harvesting, in order to meet the urgent needs of modern agriculture for efficient, precise and integrated operating equipment. Utility Model Content

[0005] This invention addresses the shortcomings of existing agricultural robots in facility agriculture scenarios, which suffer from limited functionality and low operational efficiency in fruit harvesting, pest and disease inspection, and targeted pesticide application. To address these issues, this invention employs the following technical solution: This invention provides an end effector with a mounting frame for harvesting, comprising a mounting frame, a multispectral camera module, a biomimetic flexible gripper, a shearing device, and a micro-spray head. The mounting frame is connected to the robotic arm of the agricultural robot via a magnetic quick-change mechanism, supporting rapid disassembly and maintenance. The multispectral camera module collects surface information of fruits and vegetables and uses deep learning algorithms to determine fruit maturity and identify pests and diseases. The biomimetic flexible gripper is designed with flexible materials, simulating human hand grasping actions to achieve non-destructive harvesting of soft fruits. The shearing device is equipped with a dynamic shearing control system, achieving millimeter-level positioning of fruit stems based on an improved RTMDet-ECA algorithm, and completing precise shearing by combining affine transformation and contour extraction algorithms. The micro-spray head is used for targeted pesticide application, with precise and controllable spray coverage, increasing pesticide utilization to 85%.

[0006] Furthermore, the mounting bracket is fixedly connected to the end flange of the robotic arm via an array of threaded holes and equipped with a magnetic quick-change mechanism. Its magnetic attraction force reaches over 100N in the vertical direction, ensuring the stability of the equipment in complex operating environments. The multispectral camera module integrates visible light, near-infrared, and depth sensors, achieving a fruit ripeness judgment accuracy of ≥95% through multimodal data fusion processing, and controlling the missed detection rate to below 5% even in complex occlusion scenarios. Specifically, the biomimetic flexible gripper is made of silicone material, with micron-level protrusions on its surface to increase friction. It also incorporates a pressure sensor to monitor the gripping force in real time and transmit the feedback signal to the central processing unit. The gripper has a maximum opening diameter of 120mm and a gripping force range of 0.5N to 5N, suitable for harvesting fruits and vegetables of different sizes and hardnesses, such as strawberries and tomatoes.

[0007] Furthermore, the shearing device includes a DC motor, a gear reduction mechanism, a blade assembly, and a dynamic shearing control system. The blade assembly has a cutting edge width of 0.3 mm, a shearing stroke of 15 mm, and a maximum shearing force of 20 N. The dynamic shearing control system is based on an improved RTMDet-ECA algorithm, which enhances the extraction of small target features of the fruit stem by introducing an ECA attention mechanism, and combines affine transformation and contour extraction algorithms to achieve a millimeter-level positioning accuracy of ±0.5 mm for the fruit stem shearing point. During the shearing process, the system dynamically adjusts the blade posture according to the angle and position of the fruit stem to ensure smooth and damage-free shearing action.

[0008] Furthermore, the micro-spray head is driven by piezoelectric ceramics, with a nozzle diameter of 0.2 mm, a spray particle diameter distribution range of 20 μm to 50 μm, and a coverage radius of 50 mm to 100 mm. The spray head has an embedded flow sensor that monitors the pesticide flow rate in real time and feeds the data back to the central processing unit to achieve precise control of the spray volume. The spray angle of the spray head can be adjusted within the range of 30° to 90° to meet the application needs of different plant heights and densities. The working principle of this utility model is as follows: S1, the multispectral camera module acquires fruit and vegetable image information and obtains three-dimensional environmental data through lidar; S2, the central processing unit analyzes multimodal data based on the improved RTMDet-ECA algorithm to determine fruit maturity, fruit stalk position, and pest and disease information; S3, the decision-making system generates the optimal harvesting path based on the agricultural knowledge graph and controls the robotic arm to move to the target position; S4, the biomimetic flexible gripper grasps the fruit, and the shearing device cuts the fruit stalk; S5, the micro-spray head targets the plant with pesticides based on the pest and disease identification results.

[0009] Furthermore, the agricultural knowledge graph is built upon a 40-year agronomic database, covering crop growth cycles, pest and disease occurrence patterns, and the influence of environmental factors. This knowledge graph, combined with a large-scale model bidirectional enhancement technology, forms an intelligent agronomic decision-making engine, enabling 7-day advance warning of pests and diseases, autonomous harvesting priority decisions, and path optimization. The rate of manual intervention in task allocation is reduced to below 20%, significantly improving the autonomous operation capabilities of agricultural robots.

[0010] In particular, this invention achieves an innovative architecture of an integrated "inspection, shearing, and spraying conveyor-type end effector" through modular design. The multispectral camera module, biomimetic flexible gripper, and micro-spray head are all connected to the mounting bracket via standardized interfaces, supporting magnetic quick-change. This design enables a single machine to seamlessly switch between three major functions simultaneously: inspection, non-destructive harvesting, and targeted pesticide application. It breaks through the limitations of traditional agricultural robots with their single function and constructs a fully closed-loop intelligent agricultural operation system encompassing "perception-decision-execution."

[0011] The technical advantages of this invention are as follows: By combining a multispectral camera module with an improved RTMDet-ECA algorithm, the accuracy rate of fruit maturity judgment reaches over 95%, the success rate of fruit stem cutting point location is increased to 92.67%, and the missed detection rate is less than 5% in complex occlusion scenarios. Through the design of a biomimetic flexible gripper, the strawberry damage rate is significantly reduced from the industry average of 20% to below 5%. Through the targeted application technology of the micro-spray head, pesticide usage is reduced by 50%, and the utilization rate of pesticide solution is increased to 85%. Furthermore, this invention can cover a daily operating area of ​​up to 8 mu (approximately 1.3 hectares) and has a runtime of ≥10 hours, significantly reducing labor costs and improving agricultural production efficiency. Attached Figure Description

[0012] Figure 1 This is an overall schematic diagram of the present invention, showing the layout of the robot's main structure, mounting frame and end effector, as well as its working environment; Figure 2 This is a system architecture diagram of the present invention, which describes in detail the composition and connection relationship of the data acquisition module, central processing unit, decision system, execution system and communication module; Figure 3 This is a schematic diagram of the robotic arm structure of this utility model, including the arrangement of robotic arm joints, depth camera, bionic flexible gripper and monitoring camera; Figure 4 This is a schematic diagram of the modular design of the end effector of this utility model, showing the integration method of the multispectral camera module, the biomimetic flexible gripper, the shearing device and the micro spray head and their magnetic quick-change interface.

[0013] The attached diagram is labeled as follows: 1. Main structure of the robot; 2. Dual-purpose land and rail chassis; 3. Mounting frame; 4. Multispectral camera module; 5. Bionic flexible gripper; 6. Shearing device; 7. Miniature spray head; 8. Magnetic quick-change mechanism; 9. Depth camera; 10. LiDAR; 11. Central processing unit; 12. Decision system; 13. Robotic arm; 14. Monitoring camera. Detailed Implementation

[0014] This utility model provides an end effector with a mounting frame for harvesting, the specific implementation of which is as follows: This utility model provides an end effector with a mounting frame for harvesting, the specific implementation of which is described in conjunction with the attached... Figure 1 To be continued Figure 4 A detailed explanation will be provided.

[0015] like Figure 1 As shown, the end effector includes a mounting frame 3, a multispectral camera module 4, a biomimetic flexible gripper 5, a shearing device 6, and a micro-spray head 7. The mounting frame 3 is connected to the agricultural robot arm 13 via a magnetic quick-change mechanism 8, supporting rapid disassembly and maintenance. The multispectral camera module 4 integrates visible light, near-infrared, and depth sensors to collect information about the surface of fruits and vegetables and uses an improved RTMDet-ECA algorithm to determine fruit maturity and identify pests and diseases. The biomimetic flexible gripper 5 is made of silicone with micron-level protrusions on its surface to increase friction and has a built-in pressure sensor to monitor the clamping force in real time. The shearing device 6 is equipped with a dynamic shearing control system that uses an improved RTMDet-ECA algorithm to achieve millimeter-level positioning of the fruit stalk and combines affine transformation and contour extraction algorithms to complete precise shearing. The micro-spray head 7 is driven by piezoelectric ceramics, with a nozzle diameter of 0.2 mm, a spray particle diameter distribution range of 20 μm to 50 μm, and a coverage radius of 50 mm to 100 mm.

[0016] S1, the multispectral camera module 4 acquires image information of fruits and vegetables, and obtains three-dimensional environmental data through the lidar 10. The multispectral camera module 4, as shown... Figure 4 As shown, the device connects to the mounting bracket 3 via a standardized interface to ensure stability during data acquisition. The visible light sensor in the multispectral camera module 4 captures the color information of the fruit and vegetable surface, the near-infrared sensor detects the sugar content inside the fruit, and the depth sensor acquires the spatial relationship between the fruit and its surrounding environment. The acquired data is transmitted to the central processing unit 11, which analyzes the multimodal data based on an improved RTMDet-ECA algorithm to determine fruit maturity, stem location, and pest and disease information. The improved RTMDet-ECA algorithm enhances the extraction of small target features on the stem by introducing an ECA attention mechanism, improving the accuracy of stem cutting point localization to ±0.5mm. In complex occlusion scenarios, the false negative rate is controlled below 5%.

[0017] S2, the central processing unit 11 transmits the analysis results to the decision-making system 12. The decision-making system 12 generates the optimal harvesting path based on the agricultural knowledge graph and controls the robotic arm 13 to move to the target position. The agricultural knowledge graph is built based on a 40-year agronomic database, covering crop growth cycles, pest and disease occurrence patterns, and the influence of environmental factors. The knowledge graph is combined with a large-scale model bidirectional enhancement technology to form an intelligent agronomic decision-making engine, enabling 7-day advance warning of pests and diseases, autonomous decision-making on harvesting priorities, and path optimization. The rate of manual intervention in task allocation is reduced to below 20%, significantly improving the autonomous operation capability of agricultural robots. The robotic arm 13, as shown... Figure 3 As shown, it includes multiple joints to ensure the end effector can move flexibly in three-dimensional space. The end flange of the robotic arm 13 is fixedly connected to the mounting bracket 3 through an array of threaded holes. The magnetic attraction force of the magnetic quick-change mechanism 8 reaches over 100N in the vertical direction, ensuring the stability of the equipment in complex working environments. S3, the bionic flexible gripper 5 grasps the fruit, and the shearing device 6 completes the cutting of the fruit stem. The bionic flexible gripper 5 is as follows: Figure 4 As shown, it is made of silicone with micron-level raised structures on the surface to increase friction. The gripper has a maximum opening diameter of 120mm and a gripping force range of 0.5N to 5N, suitable for harvesting fruits and vegetables of different sizes and hardness, such as strawberries and tomatoes.

[0018] The gripper has a built-in pressure sensor that monitors the gripping force in real time and transmits the feedback signal to the central processing unit 11, ensuring smooth and damage-free gripping. The shearing device 6 includes a DC motor, gear reduction mechanism, blade assembly, and dynamic shearing control system. The blade assembly has a cutting edge width of 0.3mm, a shearing stroke of 15mm, and a maximum shearing force of 20N. The dynamic shearing control system dynamically adjusts the blade posture according to the angle and position of the fruit stalk, ensuring smooth and damage-free shearing. During shearing, the system calculates the position of the fruit stalk shearing point using affine transformation and contour extraction algorithms, achieving a positioning success rate of 92.67%. S4, the micro-spray head 7 targets the plant with pesticides based on pest and disease identification results. The micro-spray head 7, as shown... Figure 4 As shown, a piezoelectric ceramic drive is used, the nozzle diameter is 0.2 mm, the spray particle diameter distribution range is 20 μm to 50 μm, and the coverage radius is 50 mm to 100 mm.

[0019] The spray head has an embedded flow sensor that monitors the pesticide flow rate in real time and feeds the data back to the central processing unit 11 to achieve precise control of the spray volume. The spray angle of the spray head can be adjusted within the range of 30° to 90° to meet the application needs of different plant heights and densities. During the application process, the micro-spray head 7 adjusts the spray coverage area according to the pest and disease identification results, increasing the pesticide utilization rate to 85% and the application accuracy to over 90%. This utility model achieves an innovative architecture of an integrated "inspection, pruning, and spraying" conveyor-type end effector through modular design. The multispectral camera module 4, the biomimetic flexible clamp 5, and the micro-spray head 7 are all connected to the mounting bracket 3 through standardized interfaces, supporting magnetic quick-change.

[0020] This design enables a single machine to seamlessly switch between three major functions: inspection, non-destructive harvesting, and targeted pesticide application. It breaks through the limitations of traditional agricultural robots with their single function and constructs a fully closed-loop agricultural intelligent operation system that integrates perception, decision-making, and execution.

[0021] like Figure 2 As shown, the system architecture includes a data acquisition module, a central processing unit 11, a decision-making system 12, an execution system, and a communication module. The data acquisition module is responsible for collecting multimodal data; the central processing unit 11 is responsible for data processing and algorithm analysis; the decision-making system 12 is responsible for task planning and path optimization; the execution system is responsible for the motion control of the robotic arm 13 and the end effector; and the communication module ensures efficient data transmission between the subsystems. In practical applications, this invention is suitable for facility agriculture scenarios such as greenhouses and orchards.

[0022] During operation, the agricultural robot moves to the target area via a dual-purpose land-rail chassis 2. A multispectral camera module 4 collects information about the fruits and vegetables and transmits it to the central processing unit 11. The central processing unit 11 analyzes the data based on an improved RTMDet-ECA algorithm to determine fruit maturity, stem position, and pest and disease information. The decision-making system 12 generates the optimal harvesting path based on an agricultural knowledge graph and controls the robotic arm 13 to move to the target location. A biomimetic flexible gripper 5 grasps the fruit, a shearing device 6 cuts the stem, and a micro-sprayer head 7 applies targeted pesticides to the plants based on pest and disease identification results. The entire process requires no human intervention, covering an area of ​​up to 8 acres per day, with a runtime of ≥10 hours, significantly reducing labor costs and improving agricultural production efficiency.

[0023] The technical effects of this invention are as follows: By combining the multispectral camera module 4 with the improved RTMDet-ECA algorithm, the accuracy rate of fruit maturity judgment reaches over 95%, the success rate of fruit stem cutting point positioning is improved to 92.67%, and the missed detection rate is less than 5% in complex occlusion scenarios. Through the design of the biomimetic flexible gripper 5, the strawberry damage rate is significantly reduced from the industry average of 20% to below 5%. Through the targeted application technology of the micro-spray head 7, pesticide usage is reduced by 50%, and the pesticide utilization rate is increased to 85%. Furthermore, this invention can cover a daily operating area of ​​up to 8 mu (approximately 1.3 hectares) and has a runtime of ≥10 hours, significantly reducing labor costs and improving agricultural production efficiency.

Claims

1. A picking end effector with a belt mount, characterized by The device includes a mounting frame (3), a multispectral camera module (4), a biomimetic flexible gripper (5), a shearing device (6), and a micro spray head (7). The mounting frame (3) is connected to the agricultural robot arm (13) via a magnetic quick-change mechanism (8). The multispectral camera module (4) is used to collect information about the surface of fruits and vegetables. The biomimetic flexible gripper (5) is made of flexible material and simulates grasping action. The shearing device (6) is equipped with a dynamic shearing control system to achieve fruit stem shearing. The micro spray head (7) is used for targeted application of pesticides.

2. The end effector with a belt mount for picking as claimed in claim 1, wherein The magnetic quick-change mechanism (8) is fixedly connected to the end flange of the robotic arm (13) through a threaded hole array, and the mounting bracket (3) is connected to the robotic arm (13) through the magnetic quick-change mechanism (8). The magnetic attraction force of the magnetic quick-change mechanism (8) reaches more than 100N in the vertical direction.

3. The end effector for picking with a belt mount according to claim 2, characterized in that The mounting bracket (3) is connected to the multispectral camera module (4), the bionic flexible clamp (5), the shearing device (6) and the micro spray head (7) through a standardized interface, supporting quick disassembly and maintenance.

4. The end effector for picking with a belt mount according to claim 1, characterized in that The multispectral camera module (4) integrates visible light, near-infrared and depth sensors to collect information on the surface of fruits and vegetables and combine algorithms to determine the maturity of fruits and identify pests and diseases.

5. The end effector for picking with a belt mount according to claim 4, characterized in that The multispectral camera module (4) acquires three-dimensional environmental data through lidar (10) and transmits the data to the central processing unit (11).

6. The end effector with mounting frame for harvesting as described in claim 1, characterized in that... The biomimetic flexible gripper (5) is made of silicone material, and its surface is provided with micron-level protrusions to increase friction. A built-in pressure sensor monitors the clamping force in real time.

7. The end effector for picking with a belt mount according to claim 6, characterized in that The maximum opening diameter of the biomimetic flexible gripper (5) is 120 mm, and the clamping force ranges from 0.5 N to 5 N.

8. The end effector for picking with a belt mount according to claim 1, characterized in that The shearing device (6) includes a DC motor, a gear reduction mechanism, a blade assembly and a dynamic shearing control system. The blade assembly has a cutting edge width of 0.3 mm, a shearing stroke of 15 mm, and a maximum shearing force of 20 N.

9. The end effector for picking with a belt mount according to claim 1, characterized in that The micro spray head (7) is driven by piezoelectric ceramics, with a nozzle diameter of 0.2 mm, a spray particle diameter distribution range of 20 μm to 50 μm, a coverage radius of 50 mm to 100 mm, and a spray angle that can be adjusted within the range of 30° to 90°.

Images