Turnover spoon type bud correcting device and method based on machine vision

Abstract

The invention discloses an overturning spoon type bud correcting device and method based on machine vision, the overturning spoon type bud correcting device comprises a controller, a seed taking device, a visual detection system and a bud correcting device, the seed taking device, the visual detection system and the bud correcting device are electrically connected with the controller, the seed taking device is used for picking and conveying garlic seeds, the visual detection system is used for identifying bulbil orientations of the garlic seeds, and the bud correcting device is used for correcting the bulbil orientations of the garlic seeds. The controller is used for controlling the bud correcting device to adjust the posture of the garlic seeds according to the bulbil orientation information of the garlic seeds provided by the visual detection system, so that bud correcting seeding is realized. Through cooperative control of a rotatable profiling spoon body curved surface sliding structure and a visual detection system, accurate recognition of the bulbil orientation of garlic seeds is achieved, and meanwhile the bulbil posture of the garlic seeds is adjusted through the resistance effect of a silica gel soft head in a push-pull type electromagnet; the garlic seeds can fall into the duckbilled inserting and sowing device in a bulbil upward posture, so as to realize bud setting and sowing.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural machinery, and in particular to a flip-spoon type bud-correcting device and method based on machine vision. Background Art

[0002] Garlic, an important global cash crop, is undergoing a critical transition from traditional manual to mechanized cultivation. While traditional manual sowing can precisely control the orientation of bulbils, it is labor-intensive and inefficient, making it difficult to meet the demands of large-scale planting. Existing mechanical bud-correcting equipment relies heavily on physical feature recognition, resulting in a high bud-correcting rate for varieties with regular morphology. However, it lacks adaptability to hybrid varieties with short bud tips and diverse morphologies, leading to a decrease in the uniformity of seedling emergence after sowing and a direct impact on the commercial yield of bulbils. Furthermore, existing mechanized sowing devices generally face the dual challenges of structural complexity and functional limitations. For example, the vibrating seed-dispensing mechanism is prone to blockage of the tube lumen due to friction with the garlic seeds, while the duckbill-type orientation device is highly dependent on the morphology of the bud tip, making it difficult to accommodate the differences in physical properties of different varieties.

[0003] Therefore, how to accurately identify and correct the bulb orientation of garlic seeds to improve the correct sprout rate of garlic seeds is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0004] In order to solve the above technical problems, this application proposes the following technical solutions:

[0005] In the first aspect, an embodiment of the present application provides a flip-spoon type bud-correcting device based on machine vision, comprising: a controller and a seed-taking device electrically connected to the controller, a visual detection system and a bud-correcting device, wherein the seed-taking device is used to pick up and transport garlic seeds, the visual detection system is used to identify the orientation of the garlic seed bulbs, and the controller controls the bud-correcting device to adjust the posture of the garlic seeds according to the garlic seed bulb orientation information provided by the visual detection system, thereby achieving correct bud sowing.

[0006] In a possible implementation, the seed taking device includes a seed box, which is fixed on a seed box bracket. A plurality of seed discharging plates are evenly arranged in the seed box, and a circulating conveying mechanism is arranged around the seed discharging plates.

[0007] In one possible implementation, the circulating transmission mechanism includes a transmission shaft arranged at the bottom of the seed box bracket, a sprocket is provided on the transmission shaft, and transmission wheels are provided at the corners and the middle position of the seed plate. The sprocket and the transmission wheel are connected by a chain transmission, and a groove is provided at the position where the seed plate contacts the chain, and the transmission wheel is arranged in the groove; flip seed spoons are equidistantly provided on the chain.

[0008] In one possible implementation, the seed box includes a seed box bottom plate, and a plurality of seed punching plates are arranged between the seed box bottom plate and the seed box front plate. The seed punching plates divide the seed box into a plurality of seed taking ports corresponding to the number of the circulating conveying mechanisms. A plurality of spring beams are arranged on the seed box bracket, and a spring seat is fixedly arranged on the spring beam. An inclined guide device is fixedly arranged on the spring seat. The inclined guide device is a transparent groove structure, and an anti-drop box is arranged on the top. The width and height of the inclined guide device are adapted to the flip seed taking spoon, and the flip seed taking spoon passes through the inclined guide device when accompanied by the chain transmission.

[0009] In one possible implementation, the flip seed spoon includes a rotatable contoured spoon body, one end of the rotatable contoured spoon body is fixedly connected to one end of a torsion spring, the second end of the torsion spring is fixedly connected to the first end of the bottom support, the second end of the bottom support is fixedly connected to the accessory outer chain plate, and the accessory outer chain plate is arranged on the chain.

[0010] In one possible implementation, the bud-correcting device includes a duckbill inserter arranged at the bottom of the inclined guide device, and push-pull electromagnets are provided at both ends of the entrance of the duckbill inserter. A retractable silicone soft head is provided at one end of the push-pull electromagnet, and the silicone soft head is turned toward the flip seed spoon. The control end of the push-pull electromagnet is electrically connected to the controller.

[0011] In one possible implementation, the visual inspection system includes an industrial camera and a visual processing system. The industrial camera is mounted on a seed box bracket, the industrial camera lens is aligned with a preset flip seed spoon point, the industrial camera is electrically connected to the visual processing system, and the visual processing system is electrically connected to a controller.

[0012] In a second aspect, an embodiment of the present application provides a flip-spoon bud correction method based on machine vision, comprising:

[0013] The sprocket drives the flip seed taking spoon to reciprocate after taking the seeds from the seed box. When it flips to a preset inclination angle, the garlic seeds slide along the curved surface of the rotatable contoured spoon body, and slide to the back of the previous flip seed taking spoon through the inclined guide device, completing the relay transfer and pre-positioning of the garlic seeds.

[0014] An industrial camera is used to collect image information of garlic seeds at the preset flip seed spoon position, and the YOLO11 target detection model is used to identify the orientation of the garlic bulbs in multiple consecutive frames of images.

[0015] The BOT-SORT tracking algorithm is used to track the spatial trajectory of garlic seeds in real time during the relay process of the flip seed scoop.

[0016] When the garlic seed moves to the last mechanical positioning point before the duckbill inserter falls, the push-pull electromagnet is controlled according to the recognition result to drive the silicone soft head on the same side of the bulbil to extend a certain distance, guiding the flip seed spoon to flip to the corresponding side, so that the garlic seed slides into the duckbill inserter with the bulbil facing upward to achieve positive bud sowing.

[0017] In one possible implementation, identifying the orientation of garlic bulbs in multiple consecutive frames of images using a YOLO11 target detection model includes:

[0018] Constructing a frame buffer list and a garlic seed orientation determination list, wherein the garlic seed orientation determination list adopts a circular storage structure of end-write and head-read;

[0019] When a single garlic seed enters the visual recognition area, images are collected at a preset time period, and the confidence of the garlic bulb and stem is calculated using the YOLO11 target detection model.

[0020] When the confidence level of scale bud identification in a single frame image is greater than or equal to a preset value, element 1 is added to the frame buffer list; when the confidence level of stem disc identification is greater than or equal to a preset value, element 0 is added to the frame buffer list;

[0021] Count the number of 1s and 0s in the frame buffer list;

[0022] If the number of 1s is greater than the number of 0s, write 1 to the end of the judgment list and clear the frame buffer list;

[0023] If the number of 0s is greater than the number of 1s, write 0 to the end of the decision list and clear the frame buffer list.

[0024] In one possible implementation, when the garlic seed moves to the last mechanical positioning point before the duckbill inserter falls, the push-pull electromagnet is controlled according to the recognition result to drive the silicone soft head on the same side of the bulbil to extend a certain distance, guiding the flip seeding spoon to flip to the corresponding side, so that the garlic seed slides into the duckbill inserter with the bulbil facing upward to achieve positive bud sowing, including:

[0025] When the garlic seed moves to the last mechanical positioning point before the duckbill inserter falls, the first element of the garlic seed direction determination list is extracted. If it is 1, a left push-pull electromagnet trigger signal is sent to the controller to drive the left silicone soft head to extend;

[0026] If it is 0, the right push-pull electromagnet trigger signal is sent to the controller to drive the right silicone soft head to extend;

[0027] The silicone soft head guides the flip seed taking spoon to flip to one side, so that the garlic seeds slide to one side, ensuring that the garlic seed bulbs face upward and fall into the duckbill inserter.

[0028] Compared with the prior art, the present invention has the following advantages:

[0029] The present application realizes accurate identification of bulbil orientation through coordinated control of the curved sliding structure of the profiling spoon body and the visual detection system, and combines the push-pull electromagnet to drive the silicone soft head on the same side of the bulbil to extend a certain distance so that it can contact one side of the flip seed taking spoon to trigger its flipping function. When the flip seed taking spoon moves to the height of the push-pull electromagnet, its rotatable profiling spoon body will contact the silicone soft head. Under the action of the resistance of the silicone soft head, the flip seed taking spoon deflects a certain angle toward the garlic seed stem disk. The garlic seed changes from its original posture lying flat on the back of the flip seed taking spoon to a posture with the bulbil tilted upward. As the flip seed taking spoon continues to move downward, the deflection angle of the flip seed taking spoon becomes larger and larger. Finally, the garlic seed falls into the duckbill inserter with the bulbil upward under the influence of gravity to achieve correct bud sowing. When the garlic seed slides off the back of the flip seed taking spoon, the push-pull electromagnet pulls back the silicone soft head, and the flip seed taking spoon is reset under the action of the internal torsion spring, thereby continuing the cycle of seed taking-transport-flip-reset.

[0030] Compared with the traditional gravity guide trough solution, this application effectively solves the problem of random sliding caused by the weight difference between the bulbils and the stem disc. Through the collaborative working mechanism of the frame buffer list and the orientation determination list, continuous multi-frame verification is performed from the time the garlic seeds enter the visual recognition area to the time they move out, which can effectively overcome the problem of instantaneous misidentification of single-frame images caused by mechanical vibration and sudden changes in illumination. By setting up visual detection for 4 seed-taking spoon points in advance, the algorithm processing period and the mechanical stable transmission period are matched in time and space, avoiding the motion blur and posture distortion problems that exist in traditional solutions when collecting images during the rapid acceleration stage before falling, further improving the accuracy of bulb orientation correction, and the error in the push-pull electromagnet triggering and mechanical phase synchronization is controlled within 0.3ms. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A schematic diagram of the overall structure of a flip-spoon-type bud-correcting device based on machine vision provided in an embodiment of the present application;

[0032] Figure 2 A schematic structural diagram of a circulating transmission mechanism provided in an embodiment of the present application;

[0033] Figure 3 A schematic diagram of the seeding process provided in an embodiment of the present application;

[0034] Figure 4 A schematic diagram of the structure of the bud-forming device provided in an embodiment of the present application;

[0035] Figure 5 An exploded view of the flip seed scoop provided in an embodiment of the present application;

[0036] Figure 6A schematic flow chart of a flip-spoon bud correction method based on machine vision provided in an embodiment of the present application.

[0037] in, Figure 1-5 The symbols in the figure are: 1-sprocket, 2-industrial camera, 3-spring beam, 4-spring seat, 5-anti-drop box, 6-flip seed spoon, 7-seed box side plate, 8-seed box front plate, 9-seed punching plate, 10-seed box, 11-drive shaft, 12-seed plate, 13-push-pull electromagnet, 14-drive wheel, 15-chain, 16-accessory outer chain plate, 17-rotatable contoured spoon body, 18-tilt guide device, 19-bottom support, 20-duckbill inserter, 21-silicone soft head, 22-torsion spring. DETAILED DESCRIPTION

[0038] The present invention will be described below with reference to the accompanying drawings and specific implementation methods.

[0039] In this embodiment, a flip-spoon type bud-correcting device based on machine vision includes: a controller and a seed-taking device electrically connected to the controller, a visual detection system and a bud-correcting device, wherein the seed-taking device is used to pick up and transport garlic seeds, and the visual detection system is used to identify the orientation of the garlic seed bulbs. The controller controls the bud-correcting device to adjust the posture of the garlic seeds according to the garlic seed bulb orientation information provided by the visual detection system, thereby achieving correct bud sowing.

[0040] Figure 1 This is a schematic diagram of the overall structure of a flip-spoon type bud-correcting device based on machine vision provided in an embodiment of the present application, see Figure 1 The seed taking device in this embodiment includes a seed box 10, which is fixed on a seed box bracket. A plurality of seeding plates 12 are evenly arranged in the seed box 10, and a circulating transmission mechanism is arranged around the seeding plates 12. Figure 2 In this embodiment, the circulating transmission mechanism includes a transmission shaft 11 arranged at the bottom of the seed box bracket, a sprocket 1 is provided on the transmission shaft 11, and transmission wheels 14 are provided at the corners and the middle position of the seed plate 12. The sprocket 1 and the transmission wheel 14 are connected by a chain 15. A groove is provided at the position where the seed plate 12 contacts the chain 15, and the transmission wheel 14 is provided in the groove. Accessory outer chain pieces 16 are equidistantly provided on the chain 15, and a flip seed spoon 6 is fixed on each accessory outer chain piece 16.

[0041] In this embodiment, the seed box 10 includes a seed box bottom plate, and a plurality of seed punching plates 9 are arranged between the seed box bottom plate and the seed box front plate 8. The seed punching plates 9 divide the seed box 10 into a plurality of seed taking ports corresponding to the number of the circulating transmission mechanism. A plurality of spring beams 3 are arranged on the seed box bracket, and a spring seat 4 is fixedly arranged on the spring beam 3. An inclined guide device 18 is fixedly arranged on the spring seat 4. Figure 3In this embodiment, in order to accurately obtain the direction of the garlic bulb, the inclined guide device 18 is set to a transparent groove structure, and an anti-drop box 5 is provided on the top. The width and height of the inclined guide device 18 are adapted to the flip seed spoon 6. The flip seed spoon 6 passes through the inclined guide device 18 when driven by the chain 15.

[0042] See also Figure 4 The flip seeding spoon 6 in this embodiment includes a rotatable contoured spoon body 17, one end of which is fixedly connected to one end of a torsion spring 22, a second end of which is fixedly connected to a first end of a bottom support 19, a second end of which is fixedly connected to an accessory outer chain piece 16, and an accessory outer chain piece 16 is arranged on a chain 15. The restoring torque of the torsion spring 22 is 0.5-1.2 N·m, so that the spoon body is reset within 300ms after the external force disappears.

[0043] See also Figure 5 In this embodiment, the bud-correcting device includes a duckbill inserter 20 disposed at the bottom of the inclined guide device 18. Both ends of the entrance of the duckbill inserter 20 are provided with push-pull electromagnets 13. One end of the push-pull electromagnet 13 is provided with a retractable silicone soft head 21. The silicone soft head 21 faces the flip seed scoop 6. The control end of the push-pull electromagnet 3 is electrically connected to the controller. In this embodiment, the visual inspection system includes an industrial camera 2 and a visual processing system. The industrial camera mount 2 is mounted on the seed box bracket. The lens of the industrial camera 2 is aligned with the preset flip seed scoop position. The industrial camera 2 is electrically connected to the visual processing system, which is in turn electrically connected to the controller.

[0044] Corresponding to the above-mentioned embodiment of a flip-spoon-type bud correction device based on machine vision, the present application also provides an embodiment of a flip-spoon-type bud correction method based on machine vision.

[0045] Figure 6 A schematic diagram of a process flow of a flip-spoon bud correction method based on machine vision provided in an embodiment of the present application, see Figure 6 In this embodiment, a flip-spoon-type bud correction method based on machine vision includes:

[0046] S101, the sprocket drives the flip seed taking spoon to reciprocate after taking the seeds from the seed box. When it flips to a preset inclination angle, the garlic seeds slide along the curved surface of the rotatable contoured spoon body, and slide to the back of the previous flip seed taking spoon through the inclined guide device to complete the relay transfer and pre-positioning of the garlic seeds.

[0047] In this embodiment, the flip seed taking spoon rotates with the sprocket. When the flip seed taking spoon passes the sprocket central axis at an inclination angle of 15°-25°, the garlic seeds in the flip seed taking spoon slide along the curved surface of the rotatable contoured spoon body under the action of gravity, and cooperate with the limiting effect of the inclined guide device to slide along the groove in the center of the inclined guide device to the back of the previous flip seed taking spoon, thereby realizing the relay transmission and physical pre-positioning of the garlic seeds.

[0048] S102: Using an industrial camera to collect image information of garlic seeds at a preset flip seed spoon position and using a YOLO11 target detection model to identify the orientation of the garlic seed bulbs in multiple consecutive frames of images.

[0049] In this embodiment, visual inspection of garlic seeds requires only identifying two distinct features: the bulbils and the stem disc. An industrial camera captures images of the garlic seeds at the fourth flip seed scoop point in front of the duckbill inserter entrance. When the garlic seeds are in the stable sliding phase in the middle section of the inclined guide, valid images are continuously identified at a 20ms cycle. At this point, the garlic seeds are in a state of dynamic equilibrium due to the restraining action of the grooves within the inclined guide, allowing for the clearest images of the bulbils. Compared to forward vertical projection detection, this reduces algorithm complexity, conserves computing resources, and improves detection accuracy.

[0050] After acquiring the corresponding images, the YOLO11 object detection model is used to identify the orientation of garlic bulbs in multiple consecutive frames. In this embodiment, the YOLO11 object detection model replaces the standard convolutional layer with a depthwise separable convolution and incorporates an attention mechanism in the feature fusion module. The model achieves an inference speed of ≤50ms / frame and a recognition accuracy of ≥98%. A garlic seed dataset is created by calibrating garlic seed image acquisition and data augmentation. The YOLO11 model trained on this dataset can quickly identify garlic seeds in images and accurately distinguish their orientation. When the confidence level of bud identification in a single-frame image is greater than or equal to 85%, element 1 is added to the pre-built frame buffer list; when the confidence level of stem-disk identification is greater than or equal to 85%, element 0 is added to the pre-built frame buffer list. When the garlic seed moves out of the recognition area, the number of 1s and 0s in the frame buffer list is counted. If the number of 1s is greater than the number of 0s, 1 is written to the end of the garlic seed orientation determination list and the frame buffer list is cleared. If the number of 0s is greater than the number of 1s, 0 is written to the end of the garlic seed orientation determination list and the frame buffer list is cleared. The garlic seed orientation determination list is initialized to 4 elements, all of which are 0. A circular storage structure of end-write and head-read is adopted. Each time a new element is written, all existing elements are shifted forward one position, and the original head-end element is removed.

[0051] S103, using a BOT-SORT tracking algorithm, real-time tracking is performed on the spatial trajectory of the garlic seeds during the relay transmission of the flipping seed scoop.

[0052] In this embodiment, after the orientation of the garlic seeds is identified by the YOLO11 target detection model, the BoT-SORT tracking algorithm is used to track the spatial trajectory of the garlic seeds in the process of flipping the seed spoon relay in real time. The BoT-SORT algorithm predicts the motion state of the garlic seeds through Kalman filtering, and adopts a dual-weight matching mechanism of appearance similarity and motion consistency to align the garlic seed detection frame in the continuous frame image of the industrial camera with the sprocket speed parameter in time and space to establish an individual garlic seed tracking sequence. The BoT-SORT tracking algorithm records the garlic seeds entering the detection area and assigns a unique ID. It cooperates with the area detection algorithm to identify the state of the garlic seeds in the identification area to prevent the redundant garlic seeds outside the field of view of the industrial camera from affecting the recognition results.

[0053] S104, when the garlic seed moves to the last mechanical positioning point before the duckbill inserter falls, the push-pull electromagnet is controlled according to the recognition result to drive the silicone soft head on the same side of the bulbil to extend a certain distance, guiding the flip seeding spoon to flip to the corresponding side, so that the garlic seed slides into the duckbill inserter with the bulbil facing upward to achieve positive bud sowing.

[0054] In this embodiment, when the garlic seed moves to the last mechanical positioning point before the duckbill inserter falls, the first element of the garlic seed orientation determination list is extracted. If it is 1, a left push-pull electromagnet trigger signal is sent to the controller to drive the left silicone soft head to extend a certain distance;

[0055] If it is 0, the right push-pull electromagnet trigger signal is sent to the controller to drive the right silicone soft head to extend a certain distance;

[0056] The silicone soft head guides the flip seed spoon to flip to one side, so that the garlic seeds slide to one side, ensuring that the garlic seed bulbs fall into the duckbill inserter with the bulbs facing upward, and realizing precise planting operation with a synchronization error of ≤0.3ms.

[0057] Furthermore, a communication middleware was constructed using a Python script, using the MC protocol to establish a persistent connection with the push-pull electromagnet controller. The middleware receives YOLO11 detection results and BoT-SORT tracking data in real time and converts garlic seed orientation correction instructions into hexadecimal control words recognizable by the controller. In this embodiment, the Python script has a built-in exception handling mechanism. If a push-pull electromagnet response delay of >50ms is detected, a heartbeat packet check is automatically triggered and the communication connection is reset, ensuring a transmission reliability of ≥99.9% for the push-pull electromagnet control signal.

[0058] In the embodiments of this application, "multiple" refers to two or more. "And / or" describes the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can mean that A exists alone, A and B exist simultaneously, or B exists alone. A and B can be singular or plural. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0059] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0060] The above description is merely a specific embodiment of the present application. Any person skilled in the art may easily conceive of variations or substitutions within the technical scope disclosed in this application, and such variations or substitutions shall be within the scope of protection of this application. The scope of protection of this application shall be subject to the scope of protection of the claims.

Claims

1. A flip-spoon bud-correcting device based on machine vision, characterized in that: include: A controller and a seed-picking device, a visual inspection system and a bud-correcting device electrically connected to the controller. The seed-picking device is used to pick up and transport garlic seeds. The visual inspection system is used to identify the orientation of the garlic buds. The controller controls the bud-correcting device to adjust the posture of the garlic seeds according to the garlic bud orientation information provided by the visual inspection system, thereby achieving correct bud sowing.

2. The flip-spoon type bud-correcting device based on machine vision according to claim 1 is characterized in that: The seed taking device comprises a seed box, which is fixed on a seed box bracket. A plurality of seed discharging plates are evenly arranged in the seed box, and a circulating transmission mechanism is arranged around the seed discharging plates.

3. The flip-spoon type bud-correcting device based on machine vision according to claim 2 is characterized in that: The circulating transmission mechanism includes a transmission shaft arranged at the bottom of the seed box bracket, a sprocket is arranged on the transmission shaft, and transmission wheels are arranged at the corners and the middle position of the seed plate. The sprocket and the transmission wheel are connected by chain transmission. A groove is provided at the position where the seed plate contacts the chain, and the transmission wheel is arranged in the groove. Flip seed spoons are equidistantly arranged on the chain.

4. The flip-spoon type bud-correcting device based on machine vision according to claim 2 is characterized in that: The seed box includes a seed box bottom plate, and a plurality of seed punching plates are arranged between the seed box bottom plate and the seed box front plate. The seed punching plates divide the seed box into a plurality of seed taking ports corresponding to the number of the circulating transmission mechanism. A plurality of spring beams are arranged on the seed box bracket, and a spring seat is fixedly arranged on the spring beam. An inclined guide device is fixedly arranged on the spring seat. The inclined guide device is a transparent groove structure, and an anti-drop box is arranged on the top. The width and height of the inclined guide device are adapted to the flip seed taking spoon, and the flip seed taking spoon passes through the inclined guide device when accompanied by the chain transmission.

5. The flip-spoon type bud-correcting device based on machine vision according to claim 4 is characterized in that: The flip seed spoon includes a rotatable contoured spoon body, one end of the rotatable contoured spoon body is fixedly connected to one end of a torsion spring, the second end of the torsion spring is fixedly connected to the first end of the bottom support, the second end of the bottom support is fixedly connected to the accessory outer chain piece, and the accessory outer chain piece is arranged on the chain.

6. The flip-spoon type bud-correcting device based on machine vision according to claim 1 is characterized in that: The bud-correcting device includes a duckbill inserter arranged at the bottom of the inclined guide device, and push-pull electromagnets are provided at both ends of the entrance of the duckbill inserter. A retractable silicone soft head is provided at one end of the push-pull electromagnet, and the silicone soft head is turned toward the flip seed spoon. The control end of the push-pull electromagnet is electrically connected to the controller.

7. The flip-spoon type bud-correcting device based on machine vision according to claim 2 is characterized in that: The visual inspection system includes an industrial camera and a visual processing system. The industrial camera is mounted on a seed box bracket. The industrial camera lens is aligned with a preset flip seed spoon point. The industrial camera is electrically connected to the visual processing system, and the visual processing system is electrically connected to a controller.

8. A flip-spoon bud correction method based on machine vision, characterized in that: The method using the machine vision-based flip-spoon bud-correcting device according to any one of claims 1 to 7 comprises: The sprocket drives the flip seed taking spoon to reciprocate after taking the seeds from the seed box. When it flips to a preset inclination angle, the garlic seeds slide along the curved surface of the rotatable contoured spoon body, and slide to the back of the previous flip seed taking spoon through the inclined guide device, completing the relay transfer and pre-positioning of the garlic seeds. An industrial camera is used to collect image information of garlic seeds at the preset flip seed spoon position, and the YOLO11 target detection model is used to identify the orientation of the garlic bulbs in multiple consecutive frames of images. The BOT-SORT tracking algorithm is used to track the spatial trajectory of garlic seeds in real time during the relay process of the flip seed scoop. When the garlic seed moves to the last mechanical positioning point before the duckbill inserter falls, the push-pull electromagnet is controlled according to the recognition result to drive the silicone soft head on the same side of the bulbil to extend a certain distance, guiding the flip seed spoon to flip to the corresponding side, so that the garlic seed slides into the duckbill inserter with the bulbil facing upward to achieve positive bud sowing.

9. The method of turning spoon type bud correction based on machine vision according to claim 8 is characterized in that: The identification of the orientation of garlic bulbs in the continuous multi-frame images by using the YOLO11 target detection model includes: Constructing a frame buffer list and a garlic seed orientation determination list, wherein the garlic seed orientation determination list adopts a circular storage structure of end-write and head-read; When a single garlic seed enters the visual recognition area, images are collected at a preset time period, and the confidence of the garlic bulb and stem is calculated using the YOLO11 target detection model. When the confidence level of scale bud identification in a single frame image is greater than or equal to a preset value, element 1 is added to the frame buffer list; when the confidence level of stem disc identification is greater than or equal to a preset value, element 0 is added to the frame buffer list; Count the number of 1s and 0s in the frame buffer list; If the number of 1s is greater than the number of 0s, write 1 to the end of the judgment list and clear the frame buffer list; If the number of 0s is greater than the number of 1s, write 0 to the end of the decision list and clear the frame buffer list.

10. The method of turning spoon type bud correction based on machine vision according to claim 8, characterized in that: When the garlic seed moves to the last mechanical positioning point before the duckbill inserter falls, the push-pull electromagnet is controlled according to the recognition result to drive the silicone soft head on the same side of the bulbil to extend a certain distance, guiding the flip seeding spoon to flip to the corresponding side, so that the garlic seed slides into the duckbill inserter with the bulbil facing upward to achieve positive bud sowing, including: When the garlic seed moves to the last mechanical positioning point before the duckbill inserter falls, the first element of the garlic seed direction determination list is extracted. If it is 1, a left push-pull electromagnet trigger signal is sent to the controller to drive the left silicone soft head to extend; If it is 0, the right push-pull electromagnet trigger signal is sent to the controller to drive the right silicone soft head to extend; The silicone soft head guides the flip seed taking spoon to flip to one side, so that the garlic seeds slide to one side, ensuring that the garlic seed bulbs face upward and fall into the duckbill inserter.

Images