Crop seed production harvester cleaning system based on target detection

Abstract

The invention relates to a crop seed production harvester cleaning system based on target detection, which comprises more than one group of crop seed production harvester cleaning devices, and each group of crop seed production harvester cleaning device comprises a slide rail accessory, a base, a flange, a nozzle bracket, an electric push rod, a steering engine, a nozzle mechanism, a target detection unit, a control unit and an action execution unit; the target detection unit comprises a visual camera and an image processing assembly, the control unit sends an instruction to the action execution unit according to whether crop seed residues exist or not, and the action execution unit controls all the components to act, including the driving motor I, the electric push rod and the steering engine to move according to a set stroke, so that the nozzle mechanism is driven to blow the crop seed residues. The technical problem that in the prior art, residues in a crop harvester are difficult to automatically clean, and seed production purity and quality are affected is solved.

Description

Technical Field

[0001] The invention belongs to the technical field of agricultural machinery, and in particular relates to a cleaning system of a crop seed harvester based on target detection. Background Art

[0002] In modern agriculture, wheat, rice and other crop harvesters are important equipment for improving harvesting efficiency and reducing labor demand. However, there is currently no device on the market specifically designed to clean up the residues of crop harvesters after harvesting. Especially during the harvesting process of crop seed production, in order to avoid mixing between seeds, it is necessary to ensure that there is no residue inside the harvester. This requirement places higher cleaning standards on harvesting equipment to ensure the purity and quality of seeds.

[0003] Existing crop harvesting equipment mainly focuses on functions such as harvesting, threshing and cleaning, but often ignores the cleaning of internal residues after harvesting. These residues include stems, leaves, husks and seeds that are not completely threshed. If they are not cleaned in time, it is easy to cause the next batch of seeds to be contaminated, affecting the purity and quality of seed production. In addition, the long-term accumulation of residues inside the harvester may also cause equipment blockage, increased wear and tear, and reduced work efficiency.

[0004] At present, the method of cleaning the residue of the harvester mainly relies on manual operation, which is inefficient and incomplete, and cannot meet the high standards of crop seed production. Therefore, there is an urgent need for a device that can automatically, efficiently and thoroughly clean the residue inside the crop seed harvester to improve the quality control level of the seed production process, ensure the purity of the seeds and the long-term stable operation of the harvesting equipment. Summary of the invention

[0005] In view of the deficiencies existing in the related art, the present invention provides a crop seed harvester cleaning system based on target detection, which solves the problem of incomplete cleaning and low efficiency of the crop seed harvester.

[0006] In a possible embodiment, a target detection-based crop seed harvester cleaning system is provided, the cleaning system includes more than one group of crop seed harvester cleaning devices, each group of crop seed harvester cleaning devices includes: a slide rail accessory, which is installed on the slide rail and can slide along the slide rail; a base, which is installed on the slide rail accessory and has a drive motor I installed at the bottom; a flange, one end of which passes through the base to connect to the nozzle bracket, and the other end is connected to the output shaft of the drive motor I; the nozzle bracket rotates under the drive of the drive motor I, the nozzle bracket is provided with a slide groove I, and the servo gear bracket is placed in the slide groove I and can slide along the slide groove I; an electric push rod, which is installed on the nozzle bracket and the push rod is connected to the servo gear bracket; the electric push rod drives the servo gear bracket to slide along the slide groove I; the servo gear is installed on the servo gear bracket, and the output shaft of the servo gear is connected to the nozzle mechanism through a connecting piece, and the servo gear rotates to drive the connecting piece and The nozzle mechanism rotates; the nozzle mechanism has a through opening inside, and the opening is connected to an air source; it also includes: a target detection unit, a control unit and an action execution unit; wherein the target detection unit includes: a visual camera, configured to capture images inside the harvester; and, an image processing component, configured to process the image acquired by the visual camera to determine whether there are any crop seed residues; and, a control unit, the control unit is further configured to send corresponding instructions to the action execution unit based on whether there are any crop seed residues determined, including one or more of the following: a drive motor I rotation instruction, an electric push rod extension instruction, a servo rotation instruction and an electric proportional control valve opening instruction; the action execution unit controls the drive motor I to rotate according to the corresponding instructions sent by the control unit, drives the electric push rod to extend and retract, controls the servo rotation and adjusts the opening of the electric proportional control valve.

[0007] In a possible implementation, the control unit includes: a main controller and a detection component; wherein the detection component includes a flow meter installed in an opening of the nozzle mechanism, and is used to detect the airflow output by the nozzle mechanism.

[0008] In one possible embodiment, the image processing component is configured to: process the image of the interior of the harvester taken by the visual camera based on the trained model, determine whether there are crop seed residues in the image, and determine the amount of crop seed residues in the image.

[0009] In a possible implementation, a plurality of groups of crop seed harvester cleaning devices are provided and are respectively placed in the silo and the discharge port of the harvester.

[0010] In a possible implementation, the crop seed harvester cleaning device is disposed on both sides of the conveying mechanism in the silo and is symmetrically distributed.

[0011] In a possible implementation, four groups of crop seed harvester cleaning devices are provided; an image processing component processes the image acquired by the visual camera to determine whether there are any crop seed residues, and sends the residue position information to the main controller of the control unit; the main controller is configured to: process the position information according to a set algorithm; wherein the position information includes X-axis and Y-axis coordinates, and the coordinate information of each point is sorted according to the size of the X-axis coordinate. After the sorting is completed, it is divided into four quadrants, and the four cleaning devices clean the residues in one quadrant respectively, control the action of the actuator, and clean the residues.

[0012] In a possible implementation, a method for detecting crop seed residues in a crop seed production and harvesting machine is provided, and a crop seed production and harvesting machine cleaning system based on target detection is applied, comprising the following steps: S10, obtaining a target detection image dataset of crop seed residues in the crop seed production and harvesting machine, and dividing the dataset into a training set and a validation set in a ratio of 9:1; S20, establishing a target detection model framework, the target detection model framework adopts YOLOv8 with a C2f backbone structure and an SPPF module; the backbone network Backbone for extracting features uses a convolutional neural network architecture EfficientNet, and the neck network adopts a high-resolution network HRNet; S30, inputting the image dataset of crop seed residues into a convolutional neural network based on an attention mechanism for training, so as to obtain a target detection model for crop seed residues; S40, using the validation set to test and verify the performance of the trained target detection model framework.

[0013] Based on the above technical scheme, the target detection-based crop seed harvester cleaning system of the present invention includes a slide rail accessory, a base, a flange, a nozzle bracket, an electric push rod, a servo, a nozzle mechanism, a target detection unit, a control unit and an action execution unit, etc. The target detection unit uses a visual camera to capture an image inside the harvester, and the image processing component processes the image to determine whether there are any crop seed residues; the control unit sends corresponding instructions to the action execution unit based on the judgment result of the image processing component, controls the rotation of the drive motor I, the extension and retraction of the electric push rod, the rotation of the servo and the opening of the electric proportional control valve, thereby automatically jet cleaning the residues; automatic detection and cleaning of residues in the crop seed harvester are realized, solving the problem of low efficiency of traditional cleaning methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0015] Figure 1 This is a schematic structural diagram of a crop seed production and harvesting machine cleaning device according to one embodiment of the present application;

[0016] Figure 2 for Figure 1 Stereoscopic image of

[0017] Figure 2a This is a schematic structural diagram of a crop seed harvester cleaning system based on target detection according to an embodiment of the present application;

[0018] Figure 3 This is a schematic structural diagram of a crop seed harvester cleaning system based on target detection according to an embodiment of the present application;

[0019] Figure 4 This is a flow chart of a method for detecting crop seed residues in a crop seed production and harvesting machine according to an embodiment of the present application;

[0020] Figure 4a This is a flowchart of the working process of the target detection unit / vision module of the crop seed harvester cleaning system based on target detection according to one embodiment of the present application;

[0021] Figure 5 A flowchart of the working process of a crop seed harvester cleaning system based on target detection according to an embodiment of the present application;

[0022] Figure 6 This is a schematic diagram of electrical connections of a crop seed harvester cleaning system based on target detection according to an embodiment of the present application;

[0023] Figure 7 This is a schematic diagram of the installation position of a cleaning device for a crop seed production and harvesting machine according to one embodiment of the present application;

[0024] Figure 8 is a bottom view of the base;

[0025] Fig. 9 This is a schematic diagram of the three-dimensional structure of a cleaning device for a crop seed production and harvesting machine according to an embodiment of the present application;

[0026] Fig.10 It is a schematic diagram of the structure of the nozzle mechanism and the steering gear;

[0027] Fig.11 for Fig.10 Front view of.

[0028] In the figure:

[0029] 1. Slide rail accessories; 101. Slide rail; 2. Base; 21. Drive motor I; 3. Flange; 4. Nozzle bracket; 5. Electric push rod; 6. Servo bracket; 7. Servo; 8. Connector; 9. Nozzle mechanism; 10. Control unit; 41. Slide chute I; 42. Slide chute II; 11. Target detection unit; 111. Visual camera; 112. Image processing component; 12. Action execution unit; 13. Silo; 14. Conveying mechanism. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0031] In the description of the present invention, it should be understood that the terms "center", "lateral", "longitudinal", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0032] The terms "first", "second", and "third" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Thus, a feature defined as "first", "second", and "third" may explicitly or implicitly include one or more of the features.

[0033] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0034] In order to solve the problem of incomplete and inefficient cleaning of crop seed harvesters, the present application provides, on the one hand, a crop seed harvester cleaning system based on target detection and a method for detecting crop seed residues in the crop seed harvester.

[0035] See also Figure 1 , Figure 2 , Figure 2a , Figure 3 and Figure 8-11 In a possible embodiment, the cleaning system includes more than one group of crop seed harvester cleaning devices, each group of crop seed harvester cleaning devices includes a slide rail accessory 1, which is installed on the slide rail 101 and can slide along the slide rail 101; a base 2, which is installed on the slide rail accessory 1, and a driving motor Ⅰ21 is installed at its bottom; a flange 3, one end of which passes through the base 2 to connect to the nozzle bracket 4, and the other end is connected to the output shaft of the driving motor Ⅰ21; the nozzle bracket 4 rotates under the drive of the driving motor Ⅰ21, and the nozzle bracket 4 is provided with a slide groove Ⅰ41, and the steering gear bracket 6 is placed in the slide groove Ⅰ41 and can slide along the slide groove Ⅰ41; an electric push rod 5, which is installed on the nozzle bracket 4, and its push rod is connected to the steering gear bracket 6, and the electric push rod 5 drives the steering gear bracket 6 to move; a steering gear 7, which is installed on the steering gear bracket 6, and the output shaft of the steering gear 7 is connected to the nozzle mechanism 9 through a connecting member 8, and the steering gear 7 rotates to drive the connecting member 8 and the nozzle mechanism 9 to rotate; The nozzle mechanism 9 has a through opening inside, and the opening is connected to the air source; it also includes: a target detection unit 11, a control unit 10 and an action execution unit 12; wherein the target detection unit 11 includes: a visual camera 111, which is configured to capture images inside the harvester; and, an image processing component 112, which is configured to process the image acquired by the visual camera to determine whether there are crop seed residues; and, a control unit 10, the control unit is further configured to send corresponding instructions to the action execution unit 12 based on whether there are crop seed residues determined. The instructions include one or more of the following: a drive motor I rotation instruction, an electric push rod 5 extension and retraction instruction, a servo 7 rotation instruction and a pneumatic unit opening and closing instruction; the action execution unit 12 controls the drive motor I to rotate according to the corresponding instructions sent by the control unit, drives the electric push rod 5 to extend and retract, controls the servo 7 to rotate and controls the opening and closing of the pneumatic unit.

[0036] As a possible implementation, a slide groove II 42 is provided on the nozzle bracket 4 ; the nozzle mechanism 9 has a through opening inside, the opening is connected to the gas source through a pipeline, and a part of the pipeline is placed in the slide groove II 42 .

[0037] In the above scheme, the base 2 is fixed on the slide rail accessory 1, one end of the flange 3 is connected to the nozzle bracket 4, and the other end is connected to the output shaft of the drive motor I, so that the nozzle bracket 4 rotates under the action of the drive motor I. The slide groove I41 on the nozzle bracket 4 is used for the sliding of the servo bracket 6 and the electric push rod 5. The electric push rod 5 pushes the servo bracket 6 to slide in the slide groove I41, thereby adjusting the position of the nozzle mechanism 9. The servo 7 is installed on the servo bracket 6, and is connected to the nozzle mechanism 9 through its output shaft to control the rotation of the nozzle mechanism 9. The target detection unit 11 includes a visual camera 111 and an image processing component 112, which are used to detect the presence of crop seed residues. The control unit 10 receives the processing results of the image processing component 112, and sends corresponding instructions to the action execution unit 12 to control the actions of the drive motor I, the electric push rod 5, the servo 7 and the pneumatic unit to achieve the cleaning function.

[0038] The present invention automatically detects and cleans crop seed residues through target detection technology, thereby improving cleaning efficiency and reducing labor costs. The combined use of the drive motor I21 and the electric push rod 5 enables the nozzle mechanism 9 to flexibly adjust its position to meet different cleaning requirements; the automated operation of the entire system reduces human intervention and improves the stability and reliability of the cleaning process.

[0039] In a possible implementation, the control unit 10 includes: a main controller and a detection component; wherein the detection component includes a flow meter installed in an opening of the nozzle mechanism, or other links in the air path, for detecting the air flow rate output by the nozzle mechanism.

[0040] In the above scheme, the flow meter is installed in the opening of the nozzle mechanism 9 to monitor the airflow output by the nozzle mechanism in real time. The main controller adjusts the opening of the electric proportional control valve according to the flow data of the detection component to ensure the stability and appropriateness of the airflow during the cleaning process.

[0041] By using the flow meter, the present invention can monitor and adjust the airflow of the nozzle mechanism 9 in real time to ensure the efficiency and stability of the cleaning process. At the same time, the data feedback of the detection component enables the system to adjust the opening of the electric proportional control valve according to the airflow, perform closed-loop airflow control, and improve the cleaning effect and system reliability.

[0042] In one possible embodiment, the image processing component is configured to: process the image of the interior of the harvester taken by the visual camera based on the trained model, determine whether there are crop seed residues in the image, and determine the amount of crop seed residues in the image.

[0043] In the above scheme, the image processing component 112 processes the image acquired by the visual camera 111 based on the trained target detection model, identifies and locates the crop seed residues in the image, quantifies the amount of residues, and transmits the analysis results to the control unit 10.

[0044] The present invention uses a trained model to perform image processing, thereby improving the accuracy and efficiency of residue detection. By accurately locating and quantifying residues, the system can clean more effectively, improving the cleaning effect and automation.

[0045] In a possible implementation, a plurality of groups of crop seed harvester cleaning devices are provided and are respectively placed in the silo and the discharge port of the harvester.

[0046] In the above scheme, multiple groups of crop seed harvester cleaning devices are respectively installed at the silo and the discharge port in the harvester. Each group of cleaning devices works independently to ensure the cleanliness of the silo and the discharge port.

[0047] By arranging multiple groups of cleaning devices, the present invention can clean different parts in the harvester at the same time, improve the cleaning efficiency and effect, and ensure the smooth progress of the entire harvesting process.

[0048] See also Figure 7 In a possible implementation manner, the crop seed harvester cleaning device is arranged on both sides of the conveying mechanism 14 in the silo 13 and is symmetrically distributed.

[0049] In the above scheme, the cleaning devices are symmetrically arranged on both sides of the conveying mechanism 14, see Figure 7 Positions ①, ②, ③ and ④ in the middle ensure cleaning after delivery.

[0050] By symmetrically arranging the cleaning devices, the present invention can evenly clean the residues on both sides of the conveying mechanism, avoiding cleaning blind areas and improving the cleaning effect.

[0051] In a possible implementation, four groups of crop seed harvester cleaning devices are provided; an image processing component processes the image acquired by the visual camera to determine whether there are any crop seed residues, and sends the residue position information to the main controller of the control unit; the main controller is configured to: process the position information according to a set algorithm; wherein the position information includes X-axis and Y-axis coordinates, and the coordinate information of each point is sorted according to the size of the X-axis coordinate. After the sorting is completed, it is divided into four quadrants, and the four cleaning devices clean the residues in one quadrant respectively, and the actuator action is controlled in ascending order to clean the residues.

[0052] In the above scheme, because the coordinate information transmitted by the image processing unit to the main controller is in character type, data processing cannot be performed, and data type conversion must be performed first to convert the character type into floating point type; after the conversion is completed, the coordinate information is sorted using bubble sort, and after completion, the coordinate information is divided into four quadrants according to the set middle value; the coordinate information belonging to each quadrant is sorted and placed in an array, and the controller extracts the coordinate information in the array in ascending order, and controls the action of the actuator according to the coordinate information.

[0053] In the above scheme, the four cleaning devices work according to the residue position information provided by the image processing component 112. The main controller processes the position information and divides the coordinate information into four quadrants. Each cleaning device is responsible for cleaning residues in one quadrant and performs cleaning operations in ascending order.

[0054] By allocating cleaning tasks to different cleaning devices, the present invention improves cleaning efficiency and accuracy. The algorithm of the main controller ensures the orderliness and efficiency of the cleaning process, avoiding repeated cleaning and omissions.

[0055] In a possible embodiment, there is a space between the connecting piece 8 and the nozzle support 4 for accommodating the nozzle mechanism 9 .

[0056] The spatial design between the connector 8 and the nozzle holder 4 in this embodiment ensures that the nozzle mechanism 9 is accommodated in the space between the connector 8 and the nozzle holder 4, thereby avoiding interference and damage to the device during harvesting operations.

[0057] In a possible implementation manner, the nozzle mechanism 9 is hinged to the connecting member 8 , and the rotation direction of the hinge is perpendicular to the rotation direction of the steering gear 7 .

[0058] This embodiment further improves the flexibility and cleaning range of the nozzle mechanism 9 by hingedly connecting the nozzle mechanism 9 to the connecting member 8 and making its rotation direction perpendicular to the rotation direction of the steering gear 7. The steering gear 7 rotates the connecting member 8 to achieve precise adjustment of the nozzle mechanism 9 in different directions, thereby more effectively cleaning the crop seed residues.

[0059] In the above solution, the nozzle mechanism 9 can clean in a wider range, improving the cleaning efficiency and effect. The vertical hinge setting helps to avoid mechanical interference that may occur during the cleaning process, ensuring the smoothness and efficiency of the cleaning process. In addition, the hinge setting also improves the applicability of the device, enabling it to cope with more complex cleaning tasks.

[0060] In a possible implementation, the pneumatic unit includes a pneumatic switch and an electric proportional control valve; the pneumatic switch is controlled by the control unit to be opened and closed, and the electric proportional control valve is controlled by the control unit to be opened to different degrees.

[0061] This embodiment realizes precise control of the air flow by setting the air source regulating valve as an electric proportional regulating valve. The control unit dynamically adjusts the air flow output of the nozzle mechanism by adjusting the opening of the electric proportional regulating valve to meet the air flow requirements under different cleaning conditions.

[0062] The introduction of the electric proportional control valve improves the accuracy and response speed of airflow control. By accurately adjusting the airflow, more efficient and accurate cleaning operations can be achieved, reducing airflow waste and optimizing cleaning effects. In addition, the automatic control of the electric proportional control valve simplifies the operating process and improves the intelligence and reliability of the cleaning device.

[0063] In the above scheme, the electric proportional control valve can adopt different specifications and models to adapt to different airflow control requirements. The control algorithm and strategy of the control unit can also be optimized and adjusted according to the actual cleaning requirements to improve the cleaning effect and efficiency. The air source regulation system can also add a pressure regulating device or a filter to further optimize the airflow quality and cleaning effect.

[0064] On the other hand, the present application provides a method for detecting crop seed residues in a crop seed harvester, using the crop seed harvester cleaning system based on target detection according to any of the above embodiments, see Figure 4 and Figure 4a In a possible implementation, the method comprises the following steps: S10, obtaining a target detection image dataset of crop seed residues in a crop seed harvester, and dividing the dataset into a training set and a validation set at a ratio of 9:1; S20, establishing a target detection model framework, wherein the target detection model framework adopts YOLOv8 with a C2f backbone structure and an SPPF module; the backbone network Backbone for extracting features uses a convolutional neural network architecture EfficientNet, and the neck network adopts a high-resolution network HRNet; S30, inputting the image dataset of crop seed residues into a convolutional neural network based on an attention mechanism for training to obtain a target detection model for crop seed residues; S40, using the validation set to test and verify the performance of the trained target detection model framework.

[0065] In the above scheme, we first obtain an image dataset of crop seed residues in a crop seed harvester, and divide the dataset into a training set and a validation set at a ratio of 9:1. Then, we establish a target detection model framework, using a YOLOv8 model with a C2f backbone structure and SPPF module. Next, we use the training set to train the target detection model, and finally use the validation set to test and validate the model.

[0066] In the above scheme, in step S20, because the detected residue is small, the backbone network Backbone for extracting features uses the convolutional neural network architecture EfficientNet to improve the detection speed and accuracy, and adds more convolutional layers to extract more detailed features. The neck network uses the high-resolution network HRNet to maintain high-resolution feature maps to improve the detection ability of small targets. In addition, the attention mechanism is introduced and the SE module is added to enhance the feature representation of small target areas.

[0067] In step S40, the performance is verified according to performance evaluation indicators, such as precision, recall, AP, mAP, etc., and the model is adjusted according to the verification structure until the model achieves satisfactory detection accuracy and effect. Adjusting the model includes adjusting the number of convolutional layers, convolution kernel size, pooling window size and stride.

[0068] Regarding the above dataset, the dataset should be pictures of the inside of the harvester, and the pictures should include residues, including but not limited to crop seeds, straw, debris, etc. The background should be the seed storage location of the crop seed harvester, including but not limited to the silo, discharge port, etc. In addition, in order to increase the diversity of training data, preprocessing such as cropping, scaling, and rotation should be performed. Regarding model training parameters, the number of training rounds epochs is set to 1000, the early stopping patience value is set to 100, the batch size batch is set to 16, and the image size is 640 pixels * 640 pixels.

[0069] By establishing and training the target detection model, the present invention can effectively detect seed residues in a crop seed harvester and improve detection accuracy and efficiency. The training and verification steps of the model ensure the reliability and stability of the detection results.

[0070] The target detection model framework can adopt other backbone structures and modules, such as ResNet, MobileNet, etc., to adapt to different detection requirements and computing resources. The data set division ratio and training algorithm can also be adjusted according to specific circumstances to improve the detection effect and performance of the model.

[0071] A crop seed harvester cleaning system based on target detection according to an embodiment of the present invention is described as follows:

[0072] The main structure of the device is as follows Figure 1 , Figure 2As shown, the device base 2 is installed on the slide rail accessory 1. The slide rail accessory has six openings, four of which are used to fix the device base 2, and the two middle holes are used to fix the slide rail accessory 1 on the slide rail 101. The slide rail accessory 1 allows the device to adjust the initial installation position. A motor / drive motor I / base rotating motor is installed inside. There is a hole reserved for the motor shaft in the middle of the device base 2. When installing, the motor shaft extends out, and the motor D axis is connected to the flange 3 using a tightening screw. The flange is provided with four holes for connecting with the nozzle bracket 4. When the base rotating motor / drive motor I rotates, the nozzle bracket 4 is driven to rotate through the flange 3. A slide groove is provided on both sides of the nozzle bracket, namely the slide groove I41, and the steering gear bracket 6 is installed in the slide groove I41 to facilitate forward and backward telescopic movement. An electric push rod mounting ear is provided at the bottom of the steering gear bracket 6 to facilitate the installation of the electric push rod 5 below the steering gear bracket 6. A mounting ear is also provided at the bottom end of the nozzle bracket 4 to fix the electric push rod 5. The electric push rod 5 telescopes and drives the steering gear bracket 6 to move to adjust the position of the entire nozzle mechanism 9. There are two steering gear mounting holes in the middle of the steering gear bracket for mounting the steering gear 7, namely Figure 3 The output shaft of the angle adjustment servo / servo 7 is connected to the nozzle mechanism 9 through a connector 8. The rotation of the angle adjustment servo / servo 7 drives the nozzle mechanism 9 to rotate through the connector 8. The rear end of the nozzle mechanism 9 is a semicircle with an opening for connecting the air pipe to provide air input. The middle part is an extension pipe so that the nozzle can be closer to the residue.

[0073] Multiple devices can be deployed inside the harvester to avoid blind spots. Taking the silo as an example, the device layout is shown in the attached figure. Figure 7 . Arrange the devices on both sides of the conveying mechanism and arrange them symmetrically, referring to Figure 7 The cleaning device of the present application can be installed at positions ①, ②, ③ and ④ in the silo. In actual use, the installation position can be adjusted according to the needs of the site. Installing the cleaning device slightly in front of both sides of the conveying mechanism can avoid cleaning dead corners, and can take into account the cleaning of the left and right sides, as well as the cleaning of the conveying mechanism groove, to avoid residues in the groove after the conveying is completed.

[0074] Object detection unit / vision part:

[0075] 1. Obtain a target detection image dataset, which is divided into a training set and a validation set at a ratio of 9:1. A single data image should be mainly composed of residues, including but not limited to crop seeds, straw, debris, etc., and the background should be the location where the crop seed harvester is likely to store seeds, including but not limited to the silo and discharge port.

[0076] 2. Establish a target detection model framework, which adopts the C2f backbone structure and YOLOv8 of the SPPF module.

[0077] 3. Use the training set to train the target detection model framework, and use the validation set to test the trained target detection framework to verify the performance.

[0078] The target detection unit / visual part includes a visual camera / camera and an image processing component. The visual camera is used to collect internal images of the crop seed harvester and send the captured photos to the image processing component. Before image processing, the model that has been trained and achieved accuracy needs to be deployed to the image processing component. The image processing component receives the photos sent by the visual camera and processes the image based on the target detection model to determine whether there are residues in the image. If the location of crop seeds, residues, etc. is marked in the image, the location information is sent to the main controller for processing.

[0079] The working process of the target detection unit / vision module is as follows Figure 4 .

[0080] Control part:

[0081] The control part includes the main controller and the detection component. The main controller is used to process the information of the target detection unit / visual part and the detection component, and send signals to the actuator. The detection component is used to collect feedback signals and confirm the working status of the system.

[0082] See also Figure 5, when the system is started, the main controller waits for the target detection unit / visual part to send the residue position information to the main controller. After receiving the position information, the main controller processes the position information according to the set algorithm. First, the position information should contain the X-axis and Y-axis coordinates. The coordinate information of each point is sorted according to the size of the X-axis coordinate. After the sorting is completed, it is divided into four quadrants. The four cleaning devices are responsible for cleaning the residue in one quadrant respectively, and then the actuator is controlled in ascending order to align with the residue. The main controller first drives the base motor to rotate, adjusts the angle of the overall nozzle mechanism, collects data through the encoder of the motor for feedback, and confirms the current angle of the nozzle mechanism. Then the main controller sends a pulse to the electric push rod to adjust the extension length of the nozzle mechanism to approach the residue. Then the main controller sends a PWM pulse to control the angle of the servo, and fine-tunes the angle of the nozzle mechanism so that the nozzle mechanism is aligned with the residue at the best angle. After the nozzle mechanism is adjusted, the main controller outputs a signal, drives the relay through the GPIO port, and the relay drives the solenoid valve. The air path is connected, and the airflow sprays the residue out of the harvester. In addition, the device can adjust the airflow size according to the amount of residue, determine the amount of residue through the visual camera, send the information to the main controller, the main controller outputs a voltage signal, adjusts the valve opening, and collects the working flow through the flow meter and feeds it back to the main controller, which adjusts and controls it. After the cleaning work is completed, the steering gear is driven to rotate, so that the nozzle mechanism rotates backward and is in a folded state to avoid damage when the harvester is working.

[0083] As an embodiment, the electrical connection of the crop seed harvester cleaning device is as follows: Figure 6 The control unit uses the STM32F407VET6 chip, and controls the rotation of the steering gear 7, the extension and retraction of the electric push rod, the rotation of the motor / servo motor, and the electromagnetic valve through the relay to connect the gas source, and adjusts the opening of the electric proportional control valve to adjust the gas flow of the gas source according to the image processing analysis of the visual camera to determine whether there are crop residues or the amount of crop residues determined, combined with the air flow detected by the flow meter.

[0084] The crop seed harvester cleaning system based on target detection of the present invention has the following beneficial effects:

[0085] 1. The device is installed inside the harvester. When the harvester is harvesting, the cleaning device does not work. In order to prevent the cleaning device from affecting the harvesting operation, the device sets the nozzle mechanism to be rotatable and adds a storage function. The nozzle mechanism is folded and stored in the space between the connecting piece and the nozzle bracket. While increasing the degree of freedom to make cleaning residues more accurate and movement more efficient, the nozzle mechanism can be folded when the harvester is harvesting, avoiding affecting the harvesting operation and protecting the nozzle mechanism from damage.

[0086] 2. This solution sets up a nozzle mechanism with a movable platform. The platform is driven by an electric push rod. The adjustable distance nozzle mechanism can adjust the front and rear distance, which can avoid the problem that distant residues cannot be cleaned up due to a single-length nozzle mechanism, or simply lengthening the nozzle mechanism to create a dead angle, resulting in the inability to clean nearby residues.

[0087] 3. This solution introduces target detection, and uses a visual camera to take photos to identify the location of the residue. The location information of the residue is sent to the controller. The controller drives the execution part, and the actuator moves and adjusts directly according to the location information of the target detection, which greatly improves the cleaning efficiency, avoids invalid movement and rotation, and saves cleaning time.

[0088] 4. The installation setting of this solution is adjustable. Through the slide rail accessories, the initial installation position of the entire device can be adjusted, and in conjunction with the electric push rod, the device can be adjusted with two degrees of freedom on the X-axis and Y-axis, greatly improving the versatility and ease of use of the device.

[0089] 5. This solution can automatically adjust the airflow size according to the amount of residue collected visually, making the device more energy-efficient.

[0090] 6. The device can be installed in a distributed manner inside the harvester to improve cleaning efficiency and avoid cleaning dead corners.

[0091] 7. The base of the device is adjustable, and can be used with the electric push rod to achieve movement with two degrees of freedom on the X-axis and Y-axis.

[0092] Finally, it should be noted that: the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0093] The above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention is described in detail with reference to the preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present invention, which should be included in the scope of the technical solution for protection of the present invention.

Claims

1. A crop seed harvester cleaning system based on target detection, characterized in that: The cleaning system comprises more than one set of crop seed harvester cleaning devices, and each set of crop seed harvester cleaning devices comprises: A slide rail accessory (1) is mounted on the slide rail (101) and can slide along the slide rail (101); A base (2) is mounted on the slide rail accessory (1), and a driving motor I (21) is mounted at the bottom thereof; A flange (3), one end of which passes through the base (2) and is connected to the nozzle bracket (4), and the other end of which is connected to the output shaft of the driving motor I (21); The nozzle bracket (4) rotates under the drive of the driving motor I, and a slide groove I (41) is provided on the nozzle bracket (4), and the steering gear bracket (6) is placed in the slide groove I (41) and can slide along the slide groove I (41); An electric push rod (5) is mounted on the nozzle bracket (4), and the push rod is connected to the steering gear bracket (6). The electric push rod (5) drives the steering gear bracket (6) to move; A steering gear (7) is mounted on the steering gear bracket (6); an output shaft of the steering gear (7) is connected to the nozzle mechanism (9) via a connecting piece (8); the steering gear (7) rotates to drive the connecting piece (8) and the nozzle mechanism (9) to rotate; The nozzle mechanism (9) has a through opening inside, and the opening is connected to a gas source; It also includes: a target detection unit (11), a control unit (10) and an action execution unit (12); wherein, The target detection unit (11) comprises: a visual camera (111) configured to capture an image of the interior of the harvester; and an image processing component (112) configured to process the image captured by the visual camera to determine whether there are crop seed residues; The control unit (10) is further configured to send corresponding instructions to the action execution unit (12) according to whether there are crop seed residues, including one or more of the following: a driving motor I rotation instruction, an electric push rod (5) extension instruction, a steering gear (7) rotation instruction and a pneumatic unit opening and closing instruction; and, The action execution unit (12) controls the driving motor I to rotate according to the corresponding instructions sent by the control unit, drives the electric push rod (5) to extend and retract, controls the steering gear (7) to rotate, and controls the opening and closing of the pneumatic unit.

2. The crop seed production and harvesting machine cleaning system according to claim 1, characterized in that: The control unit (10) comprises: a main controller and a detection component; wherein the detection component comprises a flow meter installed in an opening of the nozzle mechanism and used for detecting the airflow rate output by the nozzle mechanism.

3. The crop seed harvester cleaning system according to claim 2, characterized in that: The image processing component is configured to: process the image of the interior of the harvester taken by the visual camera based on the trained model, determine whether there are crop seed residues in the image, and determine the amount of crop seed residues in the image.

4. The crop seed production and harvesting machine cleaning system according to claim 3, characterized in that: A plurality of groups of crop seed harvester cleaning devices are arranged, which are respectively placed in the hopper and the discharge port of the harvester.

5. The crop seed production and harvesting machine cleaning system according to claim 4, characterized in that: The cleaning devices of the crop seed harvester are arranged on both sides of the conveying mechanism in the silo and are symmetrically distributed.

6. The crop seed production and harvesting machine cleaning system according to claim 5, characterized in that: 4 sets of crop seed harvester cleaning devices are provided; an image processing component processes the image acquired by the visual camera to determine whether there are crop seed residues, and sends the residue position information to the main controller of the control unit; The main controller is configured as follows: the position information is processed according to a set algorithm; the position information includes X-axis and Y-axis coordinates, and the coordinate information of each point is sorted according to the size of the X-axis coordinate. After the sorting is completed, it is divided into four quadrants. Four cleaning devices clean the residues in one quadrant respectively, control the action of the actuator, and clean the residues.

7. A method for detecting crop seed residues in a crop seed production and harvesting machine, using the crop seed production and harvesting machine cleaning system based on target detection according to any one of claims 1 to 6, characterized in that: The steps include: S10, obtaining a target detection image dataset of crop seed residues in a crop seed production and harvesting machine, and dividing the dataset into a training set and a validation set in a ratio of 9:1; S20, establish the target detection model framework, which uses the C2f backbone structure and YOLOv8 with SPPF module; the backbone network Backbone for extracting features uses the convolutional neural network architecture EfficientNet, and the neck network uses the high-resolution network HRNet; S30, inputting the image dataset of crop seed residues into a convolutional neural network based on an attention mechanism for training, so as to obtain a target detection model for crop seed residues; S40, using the validation set to test and verify the performance of the trained target detection model framework.

8. The method for detecting crop seed residues in a crop seed production and harvesting machine according to claim 7, characterized in that: In step S30 model training, the number of training rounds epochs is set to 1000, the early stopping patience value is set to 100, the batch size batch is set to 16, and the image size is 640 pixels * 640 pixels.

Images