Full-automatic dead chicken picking robot and control method
Through the fully automatic dead chicken picking robot, using RGBD depth camera and thermal infrared camera combined with visual industrial control computer, the automatic identification and picking of dead chickens in H-type cage chicken houses can be realized, which solves the problem of lack of dead chicken picking function in the existing technology and improves the picking accuracy and unmanned management level.
Patent Information
- Application Number
- CN202511157021.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-19
- Publication Date
- 2025-09-30
AI Technical Summary
In high-density farming environments, existing chicken house inspection robots lack the ability to automatically pick up dead chickens, especially in H-type cage chicken farming models. This makes manual inspections unreliable and increases the risk of human-poultry diseases, hindering the intelligent development of the chicken farming industry.
A fully automatic dead chicken picking robot was designed. It is equipped with an RGBD depth camera, a thermal infrared camera and a lidar. Combined with a visual industrial computer and a motion industrial computer, it can automatically identify and grab dead chickens. The picking task is completed by components such as a robotic arm, a synchronous belt linear slide and a mechanical gripper. It is also equipped with a cage door opening and closing mechanism and a device to prevent live chickens from running out, realizing unmanned management.
It realizes the automatic identification and picking of dead chickens in H-type cage chicken houses, improves the picking precision and accuracy, reduces manual intervention, reduces the risk of human-poultry diseases, and supports unmanned management throughout the process.
Smart Images

Figure CN120715873A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of poultry equipment, and in particular to a fully automatic dead chicken picking robot and a control method. Background Art
[0002] With the rapid expansion and intensive development of chicken farming, high-density farming has become a mainstream trend. However, this high-density farming process is also leading to an increasing incidence of chicken mortality. Failure to promptly detect dead chickens can increase the risk of infection among chickens in the surrounding area. Currently, inspections for abnormal chicken conditions rely primarily on manual inspections and removal. However, this manual approach relies on the inspector's experience, and prolonged inspections reduce accuracy and reliability, increasing the risk of human-avian diseases. To mitigate the unpredictability of manual inspections and save labor costs, there is an urgent need for inspection robots that can fully automatically remove dead chickens in high-density farming environments. However, most chicken house inspection robots currently available on the domestic market are limited to inspections, primarily monitoring environmental parameters (such as temperature, humidity, and ammonia concentration) within the chicken house and the general condition of the birds. They lack the ability to automatically remove dead chickens. Even if dead chickens are discovered during inspections, the data must be uploaded to the cloud for manual processing. Especially for the widely used H-shaped cage chicken farming model, due to its complex structure and unique spatial layout, there is currently no fully automatic dead chicken collection robot that can inspect and collect these cages. This situation seriously restricts the development of intelligent chicken farming industry. Therefore, the development of a fully automatic dead chicken collection robot that can meet the needs of H-shaped cage chicken farming has important practical significance and broad market prospects. Summary of the Invention
[0003] The purpose of this application is to provide a fully automatic dead chicken picking robot and control method, which can realize automatic picking up of dead chickens.
[0004] To achieve the above objectives, this application provides the following solutions.
[0005] In the first aspect, the present application provides a fully automatic dead chicken picking robot, comprising: a data acquisition mechanism, a dead chicken picking mechanism, a cage door opening and closing mechanism, a slider flange and a control mechanism; the control mechanism comprises a visual industrial computer and a motion industrial computer; the data acquisition mechanism comprises an RGBD depth camera group, a thermal infrared camera and a laser radar; the dead chicken picking mechanism, the cage door opening and closing mechanism, and the thermal infrared camera are arranged on the slider flange; the RGBD depth camera group and the thermal infrared camera are used to collect image information and transmit it to the visual industrial computer; the visual industrial computer performs dead chicken detection and cage door detection based on the image information and sends control instructions to the motion industrial computer based on the detection results; the motion industrial computer controls the dead chicken picking mechanism according to the control instructions to complete the grabbing of the dead chicken; the motion industrial computer also performs navigation and posture correction based on the point cloud data of the laser radar and controls the cage door opening and closing mechanism to complete the opening and closing of the cage door.
[0006] In one embodiment, the dead chicken picking mechanism includes: a robotic arm, a synchronous belt linear slide, a mechanical gripper and a turbine push rod motor; the slider flange is arranged on the synchronous belt linear slide; the turbine push rod motor is connected to the mechanical gripper; the turbine push rod motor and the mechanical gripper are both arranged on the end flange of the robotic arm; the motion industrial computer controls the turbine push rod motor to drive the mechanical gripper to complete the grabbing of the dead chicken according to the control instructions.
[0007] In one embodiment, the cage door opening and closing mechanism includes: a cage door pulling mechanism and a push rod motor connected to the cage door pulling mechanism; the push rod motor is connected to a motion industrial computer; the cage door pulling mechanism and the push rod motor are both arranged on the slider flange.
[0008] In one embodiment, the cage door opening and closing mechanism further includes: a device for preventing live chickens from running out; the device for preventing live chickens from running out is connected to a motion industrial computer; and the device for preventing live chickens from running out is arranged on a joint flange of a robotic arm.
[0009] In one embodiment, the fully automatic dead chicken picking robot also includes a dead chicken processing and analysis mechanism; the dead chicken processing and analysis mechanism includes a colorimetric detection system, a dead chicken tissue sampling mechanism, a dead chicken pushing mechanism and a dead chicken collection device; the colorimetric detection system, the dead chicken tissue sampling mechanism, the dead chicken pushing mechanism and the dead chicken collection device are all connected to the motion industrial computer.
[0010] In the second aspect, the present application provides a control method for a fully automatic dead chicken picking robot, which is applied to the fully automatic dead chicken picking robot, and the method includes: acquiring point cloud data and image information; performing navigation control according to the point cloud data; performing dead chicken detection according to the image information to obtain a detection result; if the detection result is yes, identifying the cage door position according to the image information and adjusting the posture of the fully automatic dead chicken picking robot according to the robot's current posture, ideal picking posture, cage door position and ideal cage door position; controlling the cage door opening and closing mechanism to open the cage door; identifying the dead chicken position according to the image information and controlling the dead chicken picking mechanism to grab the dead chicken; controlling the cage door opening and closing mechanism to close the cage door; updating the point cloud data and image information and returning "acquire point cloud data and image information"; if the detection result is no, returning "acquire point cloud data and image information".
[0011] In one embodiment, if the detection result is yes, the cage door position is identified according to the image information and the posture of the fully automatic dead chicken picking robot is adjusted according to the robot's current posture, ideal picking posture, cage door position and ideal cage door position, specifically including: if the detection result is yes, the cage door is detected according to the image information using an improved YOLO target segmentation algorithm to obtain the pixel coordinates of the cage door area mask; the cage door position is obtained according to the pixel coordinates of the cage door area mask and the calibration parameters of the RGBD depth camera; the deviation between the current posture of the robot and the ideal picking posture is compared, and based on the current posture of the robot, the ideal picking posture is used as the target posture and the deviation between the ideal cage door position and the current cage door position is used as feedback, and the posture of the fully automatic dead chicken picking robot is adjusted using the TEB motion planning algorithm.
[0012] In one embodiment, if the detection result is yes, the cage door position is identified based on the image information and the posture of the fully automatic dead chicken picking robot is adjusted according to the robot's current posture, ideal picking posture, cage door position and ideal cage door position. It also includes: judging whether the posture of the fully automatic dead chicken picking robot is the ideal posture to obtain a first judgment result; if the first judgment result is yes, controlling the cage door opening and closing mechanism to open the cage door; if the first judgment result is no, returning to "comparing the deviation between the current posture of the robot and the ideal picking posture, based on the current posture of the robot, using the ideal picking posture as the target posture and the deviation between the ideal cage door position and the current cage door position as feedback, and using the TEB motion planning algorithm to adjust the posture of the fully automatic dead chicken picking robot."
[0013] In one embodiment, the position of the dead chicken is identified according to the image information and the dead chicken picking mechanism is controlled to grab the dead chicken, specifically including: roughly identifying the position of the dead chicken according to the image information obtained by the RGBD depth camera at the slider flange to obtain a rough identification position; the image information includes image information obtained by the RGBD depth camera at the slider flange and image information obtained by the RGBD depth camera at the end flange of the robotic arm; controlling the robotic arm to perform linear motion according to the rough identification position and updating the image information obtained by the RGBD depth camera at the end flange of the robotic arm; finely identifying the position and posture of the dead chicken according to the updated image information obtained by the RGBD depth camera at the end flange of the robotic arm to obtain a fine position; performing path planning and posture planning on the robotic arm according to the fine position to control the robotic arm to move to the fine position and control the robotic gripper to grab the dead chicken.
[0014] In one embodiment, after identifying the position of the dead chicken according to the image information and controlling the dead chicken picking mechanism to grab the dead chicken, it also includes: judging whether the dead chicken is grabbed to obtain a second judgment result; if the second judgment result is yes, returning to "controlling the cage door opening and closing mechanism to close the cage door"; if the second judgment result is no, returning to "performing path planning and posture planning for the robotic arm according to the fine position to control the robotic arm to move to the fine position and controlling the mechanical gripper to grab the dead chicken."
[0015] According to the specific embodiments provided in this application, this application discloses the following technical effects: This application provides a fully automatic dead chicken picking robot and control method. An RGBD depth camera group and a thermal infrared camera are used to collect image information and transmit it to a visual industrial computer. The visual industrial computer detects dead chickens and cage doors based on the image information and sends control instructions to a motion industrial computer based on the detection results. The motion industrial computer controls a dead chicken picking mechanism to capture dead chickens based on the control instructions. The motion industrial computer also navigates based on point cloud data from a laser radar and controls the cage door opening and closing mechanism to open and close the cage door. A data acquisition module acquires image information and point cloud data, and the visual industrial computer identifies dead chickens based on the image information. The motion industrial computer then uses the control to perform the actual picking, thereby achieving automatic dead chicken picking. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0017] Figure 1Schematic diagram of a fully automatic dead chicken collection robot.
[0018] Figure 2 This is a schematic diagram of the connection relationship of the fully automatic dead chicken picking robot.
[0019] Figure 3 This is a schematic diagram of the control method of the fully automatic dead chicken picking robot.
[0020] Figure 4 This is a schematic diagram for cage door identification.
[0021] Figure 5 An analysis agency for dead chicken processing.
[0022] Figure numerals: carbon fiber support rod-1, synchronous belt linear slide-2, three-layer photoelectric switch-3, first RGBD depth camera-4, thermal infrared camera-5, visual industrial computer-6, robotic arm-7, second-layer photoelectric switch-8, first-layer photoelectric switch-9, stepper motor-10, 32-line laser radar-11, hub motor-12, motion control card-13, motion industrial computer-14, colorimetric detection system-15, dead chicken tissue sampling mechanism-16, dead chicken ejection mechanism-17, dead chicken collection device-18, slider flange-19, mechanical gripper-20, live chicken escape prevention device-21, second RGBD depth camera-22, turbine push rod motor-23, cage door pulling mechanism-24, push rod motor-25, four-layer photoelectric switch-26. DETAILED DESCRIPTION
[0023] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.
[0024] This application provides a fully automatic dead chicken collection robot for H-type cage chicken houses. The robot can identify dead chickens within the H-type cage, open the corresponding cage door, activate the device that prevents live chickens from escaping, and then use a robotic arm to grab the dead chicken and place it in a collection device. The collection device uses biosensors to analyze the cause of death and upload the analysis results to the cloud, achieving unmanned management of the H-type cage chicken house.
[0025] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.
[0026] like Figure 1As shown, the present application provides a fully automatic dead chicken picking robot, which includes: a data acquisition mechanism, a dead chicken picking mechanism, a cage door opening and closing mechanism, a slider flange 19 and a control mechanism.
[0027] The control mechanism includes a vision control computer 6 and a motion control computer 14; the data acquisition mechanism includes an RGBD depth camera assembly, a thermal infrared camera 5, and a lidar. The thermal infrared camera 5 of the dead chicken picking mechanism is mounted on the slider flange 19. In actual use, the RGBD depth camera assembly includes a first RGBD depth camera 4 and a second RGBD depth camera 22. The first RGBD depth camera 4 is mounted on the slider flange 19, and the second RGBD depth camera 22 is mounted on the end flange of the robotic arm.
[0028] The RGBD depth camera group and thermal infrared camera 5 are used to collect image information and transmit it to the visual control computer 6. The visual control computer 6 detects dead chickens and cage doors based on the image information and sends control instructions to the motion control computer 14 based on the detection results. The motion control computer 14 controls the dead chicken collection mechanism to complete the capture of dead chickens according to the control instructions. The motion control computer 14 also navigates based on the point cloud data from the laser radar and controls the cage door opening and closing mechanism to complete the opening and closing of the cage door. In actual applications, the laser radar used in this application is a 32-line laser radar 11.
[0029] The image information and point cloud data are acquired through the data acquisition module, the visual industrial computer 6 recognizes the dead chickens according to the image information, and then the motion industrial computer 14 is used to complete the actual picking according to the control, thereby realizing the automatic picking of the dead chickens.
[0030] In an exemplary embodiment, the dead chicken picking mechanism includes: a robotic arm 7, a synchronous belt linear slide 2, a mechanical gripper 20, and a turbine pusher motor 23; the slider flange 19 is mounted on the synchronous belt linear slide 2; the turbine pusher motor 23 is connected to the mechanical gripper 20; both the turbine pusher motor 23 and the mechanical gripper 20 are mounted on the end flange of the robotic arm; the motion control computer 14 controls the turbine pusher motor 23 according to control instructions, driving the mechanical gripper 20 to retrieve the dead chicken. In actual applications, the dead chicken picking mechanism also includes a carbon fiber support rod 1, on which the synchronous belt linear slide 2 is mounted.
[0031] In an exemplary embodiment, the cage door opening and closing mechanism includes: a cage door pulling mechanism 24 and a push rod motor 25 connected to the cage door pulling mechanism 24; the push rod motor 25 is connected to the motion industrial computer 14; the cage door pulling mechanism 24 and the push rod motor 25 are both arranged on the slider flange 19.
[0032] The cage door opening and closing mechanism also includes a live chicken escape prevention device 21, which is connected to the motion control computer 14 and mounted on the joint flange of the robotic arm. The live chicken escape prevention device 21 is an umbrella-shaped structure. When the cage door opening and closing mechanism is pushed open by the push rod motor, the umbrella extends to block the open cage door. When closed, the umbrella retracts.
[0033] In an exemplary embodiment, the fully automatic dead chicken picking robot also includes a dead chicken processing and analysis mechanism; the dead chicken processing and analysis mechanism includes a colorimetric detection system 15, a dead chicken tissue sampling mechanism 16, a dead chicken pushing mechanism 17 and a dead chicken collecting device 18; the colorimetric detection system 15, the dead chicken tissue sampling mechanism 16, the dead chicken pushing mechanism 17 and the dead chicken collecting device 18 are all connected to the motion industrial computer 14. Figure 5 This is the dead chicken processing and analysis mechanism structure, wherein 15 is a colorimetric detection system, 16 is a dead chicken tissue sampling mechanism, 17 is a dead chicken pushing mechanism and 18 is a dead chicken collecting device, wherein the dead chicken pushing mechanism 17 is composed of two push rod motors, and the dead chicken tissue sampling mechanism 16 is composed of a collection device and a rotating collection disk. The collected tissue is pretreated by the gold standard processing module, and the pretreated solvent is secondary processed by the microfluidic chip. The treated solution is input onto the colorimetric card paper, and the camera analyzes the color card to obtain the processing result.
[0034] In practical applications, such as Figure 2 As shown, the first RGBD depth camera 4, the second RGBD depth camera 22 and the RGB camera in the colorimetric detection system 15 are connected to the visual industrial computer 6 in a USB3.0 connection mode, and the thermal infrared camera 5 is connected to the visual industrial computer 6 through an Ethernet port. At the same time, the visual industrial computer 6 is connected to a switch, which is located in the motion industrial computer 14. The motion industrial computer 14, the 32-line laser radar 11, and the 4G router are all connected to the switch through an Ethernet port. The 4G router is connected to an external server through a private tunnel. The robotic arm 7 is connected to the motion industrial computer 14 through an Ethernet port. All systems are powered by a lithium battery. It is connected to the power conversion module through a DC plug, the power conversion module is connected to the motion industrial computer 14 through a DC interface, the motion industrial computer 14 is connected to the motion control card 13 through a serial port, the four hub motors 12 are connected to the motion control card 13 through a CAN bus, the push rod motor 25, the cage door mechanism 24, the turbine push rod motor 23, and the stepper motor 10 are connected to the motion control card 13 through a 485 bus, the live chicken prevention device 21, the dead chicken tissue sampling mechanism 16, the dead chicken ejection mechanism 17, the first layer photoelectric switch 9, the second layer photoelectric switch 8, the third layer photoelectric switch 3, and the fourth layer photoelectric switch 26 are all connected to the motion control card 13 in the form of an IO port.
[0035] Based on the same inventive concept, the embodiment of the present application also provides a control method for a fully automatic dead chicken picking robot. The fully automatic dead chicken picking robot control method described in the fully automatic dead chicken picking robot is as follows: Figure 3 As shown, the method includes the following steps.
[0036] Obtain point cloud data and image information.
[0037] Navigation control based on point cloud data.
[0038] Dead chicken detection is performed based on the image information to obtain a detection result.
[0039] If the detection result is yes, the cage door position is identified according to the image information and the posture of the fully automatic dead chicken picking robot is adjusted according to the robot's current posture, the ideal picking posture, the cage door position and the ideal cage door position.
[0040] Control the cage door opening and closing mechanism to open the cage door.
[0041] The position of the dead chicken is identified according to the image information and the dead chicken picking mechanism is controlled to grab the dead chicken.
[0042] Control the cage door opening and closing mechanism to close the cage door.
[0043] Update the point cloud data and image information and return to "Get point cloud data and image information".
[0044] If the detection result is no, then return to "obtaining point cloud data and image information".
[0045] In an exemplary embodiment, if the detection result is yes, the cage door position is identified according to the image information and the posture of the fully automatic dead chicken picking robot is adjusted according to the robot's current posture, ideal picking posture, cage door position and ideal cage door position, specifically including: if the detection result is yes, the cage door is detected according to the image information using the improved YOLO target segmentation algorithm to obtain the pixel coordinates of the cage door area mask; the cage door position is obtained according to the pixel coordinates of the cage door area mask and the calibration parameters of the RGBD depth camera; the deviation between the current posture of the robot and the ideal picking posture is compared, and based on the current posture of the robot, the ideal picking posture is used as the target posture and the deviation between the ideal cage door position and the current cage door position is used as feedback, and the posture of the fully automatic dead chicken picking robot is adjusted using the TEB motion planning algorithm.
[0046] In an exemplary embodiment, if the detection result is yes, the cage door position is identified according to the image information and the posture of the fully automatic dead chicken picking robot is adjusted according to the robot's current posture, ideal picking posture, cage door position and ideal cage door position. It also includes: judging whether the posture of the fully automatic dead chicken picking robot is the ideal posture to obtain a first judgment result; if the first judgment result is yes, controlling the cage door opening and closing mechanism to open the cage door; if the first judgment result is no, returning "comparing the deviation between the current posture of the robot and the ideal picking posture, based on the current posture of the robot, using the ideal picking posture as the target posture and the deviation between the ideal cage door position and the current cage door position as feedback, and using the TEB motion planning algorithm to adjust the posture of the fully automatic dead chicken picking robot."
[0047] In an exemplary embodiment, the position of the dead chicken is identified according to the image information and the dead chicken picking mechanism is controlled to grab the dead chicken, specifically including: roughly identifying the position of the dead chicken according to the image information obtained by the RGBD depth camera at the slider flange to obtain a rough identification position; the image information includes image information obtained by the RGBD depth camera at the slider flange and image information obtained by the RGBD depth camera at the end flange of the robotic arm; controlling the robotic arm to move in a straight line according to the rough identification position and updating the image information obtained by the RGBD depth camera at the end flange of the robotic arm; finely identifying the position and posture of the dead chicken according to the updated image information obtained by the RGBD depth camera at the end flange of the robotic arm to obtain a fine position; performing path planning and posture planning on the robotic arm 7 according to the fine position to control the robotic arm 7 to move to the fine position and control the robotic gripper 20 to grab the dead chicken.
[0048] In an exemplary embodiment, after identifying the position of the dead chicken according to the image information and controlling the dead chicken picking mechanism to grab the dead chicken, it also includes: judging whether the dead chicken is caught to obtain a second judgment result; if the second judgment result is yes, returning to "controlling the cage door opening and closing mechanism to close the cage door"; if the second judgment result is no, returning to "performing path planning and posture planning for the robotic arm 7 according to the fine position to control the robotic arm 7 to move to the fine position and control the mechanical gripper 20 to grab the dead chicken."
[0049] This application also describes the workflow of a fully automatic dead chicken picking robot in practical applications.
[0050] Step 1: Use 32-line laser radar 11, Faster-LIO algorithm and point cloud to raster map method to create an accurate two-dimensional raster map.
[0051] The step 1 further includes the following steps.
[0052] S10. Sensor calibration.
[0053] In a chicken coop environment, two frames of original point cloud data and IMU data obtained by the motion control card 13 are first collected. The rotation matrix and translation matrix between the two frames of point cloud are calculated through the NDT algorithm, and the rotation between the two frames of IMU is calculated at the same time; the relative rotation matrix and translation matrix of several groups of IMU and LiDAR point clouds are collected, which is equivalent to knowing a 4n×4 matrix, solving them jointly, and using the SVD decomposition method to obtain the optimal solution of a 4×1 vector.
[0054] S11. Data collection.
[0055] The point cloud data of the laser radar in the standardized H-type cage chicken house environment is collected through the 32-line laser radar 11, and the IMU data is obtained through the motion control card 13.
[0056] S12. Point cloud map creation.
[0057] The original lidar data is stored through incremental voxels (iVox), and this type of data is used as the input of the Faster-LIO algorithm to output a point cloud map.
[0058] S13. Establishment of two-dimensional grid map.
[0059] The acquired point cloud map is fed into a pass-through filter tailored to the robot's dimensions to obtain the processed point cloud map. This point cloud map is then projected onto the XOY plane. The minimum x, maximum x, minimum y, and maximum y values of all two-dimensional points are then found to determine the coordinates of the map's lower-left and upper-right corners. The grid resolution is then set, and the projected two-dimensional points are traversed. The grid cell to which each point belongs is determined based on its coordinates. The number of points within each grid cell is then counted. Based on the number of points within the grid cell, each grid cell is assigned a status value. Common status values include occupied, idle, and unknown. For example, if the number of points within a grid cell exceeds a set threshold, the grid cell is considered occupied; if the number of points is less than the set threshold, the grid cell is considered idle; and if there is insufficient information to determine the status of the grid cell, the grid cell is considered unknown.
[0060] Step 2: Based on the existing two-dimensional grid map of the chicken house, set the charging point, starting point, inspection point and inspection route. Each inspection route needs to repeat the feeding layer several times. At the starting point, the motion industrial computer 14 controls the stepper motor 10 to raise the synchronous belt linear slide 2 fixed on the carbon fiber support rod 1 and the dead chicken picking mechanism assembled with the slider flange 19 of the dead chicken picking device until all layers are traversed and pass through the first-layer photoelectric switch 9, the second-layer photoelectric switch 8, the third-layer photoelectric switch 3 and the fourth-layer photoelectric switch 26.
[0061] Step 3: Obtain image information through the thermal infrared camera 5 and the RGBD depth camera, and use the visual industrial computer 6 to detect dead chickens in real time during the inspection process. If a dead chicken is detected, the visual industrial computer 6 will use Ethernet to send information to the motion industrial computer 14. After receiving the information, the motion industrial computer 14 will send a motion stop signal to the motion control card 13 through the serial port to control the hub motor 12.
[0062] Step 4: Adjust the robot's picking posture.
[0063] Said step 4 further comprises the following steps.
[0064] S41. Sensor calibration.
[0065] According to formula (1), the robot arm 7 and the first RGBD depth camera 4 are calibrated. A checkerboard calibration plate is placed at the end of the robot arm 7 to perform external parameter calibration on the coordinates of the first RGBD depth camera 4 and the robot arm 7.
[0066] (1) In the formula is the pose of the calibration plate relative to the camera coordinate system, is the position of the base of the robot arm 7 relative to the camera coordinate system, is the inverse of the pose matrix of the end of the manipulator 7 relative to the base of the manipulator 7, is the inverse matrix of the pose matrix of the calibration plate relative to the end coordinate system of the manipulator 7. The value of can be obtained through the feedback of the robot arm 7, The value of is a fixed constant. It is a homogeneous matrix composed of a rotation matrix and a translation matrix, where the rotation matrix is obtained through the pixel coordinates of the checkerboard corner points, and the translation matrix is obtained through the depth information.
[0067] S42. Cage door identification.
[0068] like Figure 4 As shown, the first RGBD depth camera 4 data and the improved YOLO v8 target segmentation algorithm are used to detect the chicken house cage door, identify the cage door area and obtain the corresponding depth point cloud data of the cage door area. The improved YOLO v8 target segmentation algorithm is an improvement of YOLO v8. The improvement of YOLO v8 is to replace the convolution layer of the neck network with a multi-scale dynamic snake convolution to improve the segmentation area extraction of straight line features. The RTX4060Ti graphics card is used in the training process. The X-AnyLabeling software is used to perform OBB rotation segmentation and annotation on the collected 5000 RGB image data, and the training set, validation set and test set are divided into 7:2:1 ratios. After extracting the segmented area and obtaining the corresponding rotation angle After that, the depth value of the area is converted into point cloud data. Let the obtained point cloud be , divide the point cloud into four point sets according to the distribution of the segmentation area , process the four point sets, and use the point set Hedianji For example, the point set The points in the figure are fitted with a straight line using the improved RANSAC algorithm. The improvement lies in adding voxel downsampling before the RANSAC algorithm processing. The sampling size of the voxel downsampling is 1 / 10 of the width of the recognized box. Then the sampling method is changed to sequential sampling, and the threshold is set to 1 / 2 of the width of the recognized box. The fitting straight line equation can be obtained by formula (2). Similarly, the point set The fitting straight line equation can be obtained according to formula 3, the point set and The corresponding fitting straight line equation can also be obtained by formula (4) and formula (5), and then the coordinates of the four corner points of the cage door relative to the robot body are obtained using formula (6) and formula (7): .
[0069] (2) (3) (4) (5) (6) (7) in, is the slope of the fitted line on the upper side of the identification box, is the slope of the fitted line on the right side of the identification box, is the slope of the fitted line below the identification box, The slope of the fitted line to the left of the identification box. is the intercept of the fitted straight line on the upper side of the identification frame, is the intercept of the fitted straight line on the right side of the identification box, is the intercept of the fitted straight line below the identification box, is the intercept of the fitted straight line on the left side of the identification box, To fit the minimum depth within the identified cage door area, Fits all depth values within the identified cage door area.
[0070] S43. Relative error calculation.
[0071] The RANSAC algorithm is used to perform finite plane fitting on the cage door area, and the plane equation is obtained by formula (8).
[0072] (8) Where { } are the parameters of the plane equation, is the normal vector of the plane in the camera coordinate system, The direction vector of the axis is , calculate the rotation matrix according to formula (9) .
[0073] (9) The translation vector difference of the camera relative to the plane can be calculated by formula (10).
[0074] (10) In the formula For the standard grasping relative cage door position, This is the standard grasping posture.
[0075] In practical applications, plane fitting methods in relative error calculation include but are not limited to RANSAC.
[0076] S44. Motion control.
[0077] Obtain the global coordinate point of the robot's current position, and calculate the coordinates of the robot's grasping target position relative to the global coordinate system according to formula (11).
[0078] (11) In the formula is the coordinate point of the robot's current position in the global coordinate system, obtained through the robot's positioning information. is the coordinate of the cage in the robot coordinate system, through the corner coordinates Get, The relative error calculated by and After obtaining the coordinates of the grasping target position relative to the global coordinate system, adjust the robot's position and posture by combining the grid map described in step 2 and the TEB motion planning algorithm.
[0079] In practical applications, motion control includes but is not limited to the TEB algorithm mentioned above.
[0080] S45. Obtaining depth value.
[0081] In this posture, the dead chicken is identified for the second time using the first RGBD depth camera 4 and the thermal infrared camera 5 , and the depth value of the location of the dead chicken is obtained by the first RGBD depth camera 4 .
[0082] Step 5: Robotic dead chicken collection.
[0083] S50. Sensor calibration.
[0084] The second RGBD depth camera 22 and the robotic arm 7 are calibrated using a checkerboard calibration plate and Zhang Zhengyou calibration method, and the pixel coordinates obtained by the depth camera are synchronized to the base coordinates of the robotic arm 7.
[0085] S51. Control the push rod motor 25 to push out the cage door mechanism 24 until there is push-in feedback.
[0086] S52. The robot arm 7 searches for the optimal gait for the robot arm 7 and the robot's posture at this time in the motion gait library of the specified layer and moves to the initial state of entering the cage door.
[0087] S53 . Generate an obstacle of the cage size based on the cage door identified by the first RGBD depth camera 4 .
[0088] The coordinates of the four corner points of the cage door relative to the robot body obtained by S42 Obstacle models are constructed based on standard cage door dimensions, feeding trough dimensions, egg tray dimensions, and the robot body.
[0089] S54. Use the cage door pulling mechanism 24 to open the cage door and turn on the device 21 to prevent live chickens from running out.
[0090] S55. The robotic arm 7 performs linear motion. After reaching the depth obtained in S45, the motion control card 13 sends an opening instruction to the turbine pusher motor 23, thereby opening the mechanical gripper 20.
[0091] S56. After the second RGBD depth camera 22 acquires the image data, the visual industrial computer 6 will perform a third recognition to accurately identify the position and posture of the dead chicken and calculate the optimal grasping posture of the end of the robotic arm 7.
[0092] S57. The robotic arm 7 moves to the specified grasping posture, and the motion control card 13 sends a closing instruction to the turbine pusher motor 23, thereby closing the mechanical gripper 20 to clamp the dead chicken.
[0093] S58. The robot arm 7 returns to the third recognition position, and then returns to the second recognition position in a linear motion.
[0094] S59. Use the cage door pulling mechanism 24 to close the cage door and the device 21 to prevent live chickens from running out.
[0095] S510. Control the push rod motor 25 to retract the cage door mechanism 24 until there is feedback that the cage door is completely retracted.
[0096] S511. The robotic arm 7 returns to the initial position, and the stepper motor 10 is controlled by the motion industrial computer 14 to move the synchronous belt linear slide 2 fixed on the carbon fiber support rod 1 and the dead chicken picking mechanism assembled by the slider flange 19 of the dead chicken picking device to the designated recycling layer.
[0097] S512. Place the dead chicken to the dead chicken tissue sampling mechanism 16 according to a fixed trajectory.
[0098] Step 6: Use the colorimetric detection system 15 to quickly detect pathogenic bacteria in the tissues of dead chickens.
[0099] S61. Tissue sampling of dead chickens.
[0100] The dead chicken tissue sampling mechanism 16 samples the dead chicken tissue through a needle tube or a rotating cutter located below.
[0101] S61. Collection of dead chickens.
[0102] The dead chicken pushing mechanism 17 is used to push the dead chickens after sampling into the dead chicken collecting device 18. The dead chicken collecting device 18 adopts a single-sided flip cover design. When pushed in, the cover opens due to the thrust, and when retracted, the cover closes, ensuring the sealing of the dead chicken collecting device 18.
[0103] S62. Samples are standardized using a microfluidic chip.
[0104] The sampled tissue is placed in a colorimetric detection system 15 , which includes a dead chicken tissue gold standard processing module and a microfluidic chip. The tissue fluid flowing out of the microfluidic chip reacts with the colorimetric card.
[0105] S63.RGB camera for colorimetric acquisition.
[0106] The color comparison results are analyzed by the RGB camera of the colorimetric detection system 15 .
[0107] S64. Automatically change the sample colorimetric card.
[0108] After the results are obtained, the sample colorimetric card is replaced by the automatic card changing device of the colorimetric detection system 15 .
[0109] S65. Upload the results to the cloud server.
[0110] Use easy-rsa to create SSL / TLS certificates and keys, and then use OpenVPN to communicate with the server through a private link to ensure the secure upload of data. The uploaded data includes the location of the dead chickens, the time of dead chicken detection, the dead chicken number, the cause of death of the dead chickens, photos of the cages, and photos of color cards.
[0111] Step 7: The motion industrial computer 14 controls the stepper motor 10 to move the synchronous belt linear slide 2 fixed on the carbon fiber support rod 1 and the dead chicken picking mechanism assembled with the slider flange 19 to rise to the original position, and determine whether there are other dead chickens. If there are, repeat steps 5-6. If not, continue the inspection until all inspection tasks are completed and return to the charging point.
[0112] The robot provided by this application fills the gap in the field of fully automatic dead chicken picking robots in China. Compared with other chicken house inspection robots, it has added a dead chicken picking mechanism, which enables it to automatically identify dead chickens, open cage doors, and pick up dead chickens, and takes into account factors such as live chickens running out and in-situ measurement of the cause of death of dead chickens; the method adopts three-stage recognition to improve the accuracy of dead chicken picking, and considers the influence of the posture of the robot body in the first stage of recognition on the grasping accuracy, and uses deviation recognition combined with motion control methods to improve the grasping accuracy.
[0113] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0114] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.
Claims
1. A fully automatic dead chicken picking robot, characterized in that: The fully automatic dead chicken picking robot comprises: a data acquisition mechanism, a dead chicken picking mechanism, a cage door opening and closing mechanism, a slider flange and a control mechanism; The control mechanism includes a visual industrial computer and a motion industrial computer; the data acquisition mechanism includes an RGBD depth camera group, a thermal infrared camera and a laser radar; the dead chicken picking mechanism and the thermal infrared camera are arranged on the slider flange; The RGBD depth camera group and the thermal infrared camera are used to collect image information and transmit it to the visual industrial computer; the visual industrial computer detects dead chickens and cage doors based on the image information and sends control instructions to the motion industrial computer based on the detection results; the motion industrial computer controls the dead chicken picking mechanism to complete the capture of dead chickens according to the control instructions; the motion industrial computer also performs navigation and posture correction based on the point cloud data of the laser radar and controls the cage door opening and closing mechanism to complete the opening and closing of the cage door.
2. The fully automatic dead chicken picking robot according to claim 1 is characterized in that: The dead chicken picking mechanism includes: a robotic arm, a synchronous belt linear slide, a mechanical gripper and a turbine push rod motor; The slider flange is arranged on the synchronous belt linear slide; the turbine push rod motor is connected to the mechanical gripper; the turbine push rod motor and the mechanical gripper are both arranged on the end flange of the robotic arm; the motion industrial computer controls the turbine push rod motor to drive the mechanical gripper to complete the grabbing of the dead chicken according to the control instructions.
3. The fully automatic dead chicken picking robot according to claim 1, characterized in that: The cage door opening and closing mechanism includes: a cage door pulling mechanism and a push rod motor connected to the cage door pulling mechanism; The push rod motor is connected to the motion industrial computer; the cage door pulling mechanism and the push rod motor are both arranged on the slider flange.
4. The fully automatic dead chicken picking robot according to claim 3 is characterized in that: The cage door opening and closing mechanism also includes: a device to prevent live chickens from running out; The device for preventing live chickens from running out is connected to a motion industrial computer; and the device for preventing live chickens from running out is arranged on a joint flange of a robotic arm.
5. The fully automatic dead chicken picking robot according to claim 1, characterized in that: It also includes dead chicken disposal analysis agencies; The dead chicken processing and analysis mechanism includes a colorimetric detection system, a dead chicken tissue sampling mechanism, a dead chicken pushing mechanism and a dead chicken collecting device; the colorimetric detection system, the dead chicken tissue sampling mechanism, the dead chicken pushing mechanism and the dead chicken collecting device are all connected to the motion industrial computer.
6. A fully automatic dead chicken picking robot control method, characterized in that: The control method of the fully automatic dead chicken picking robot is applied to the fully automatic dead chicken picking robot according to any one of claims 1 to 5, and the control method of the fully automatic dead chicken picking robot includes: Obtain point cloud data and image information; Navigation control based on point cloud data; Performing dead chicken detection based on the image information to obtain a detection result; If the detection result is yes, the cage door position is identified according to the image information and the posture of the fully automatic dead chicken picking robot is adjusted according to the robot's current posture, the ideal picking posture, the cage door position and the ideal cage door position; Control the cage door opening and closing mechanism to open the cage door; Identify the location of dead chickens based on image information and control the dead chicken picking mechanism to grab the dead chickens; Control the cage door opening and closing mechanism to close the cage door; Update the point cloud data and image information and return "Get point cloud data and image information"; If the detection result is no, return to "obtaining point cloud data and image information".
7. The control method of the fully automatic dead chicken picking robot according to claim 6, characterized in that: If the detection result is yes, the cage door position is identified according to the image information and the posture of the fully automatic dead chicken picking robot is adjusted according to the robot's current posture, the ideal picking posture, the current cage door position and the ideal cage door position, specifically including: If the detection result is yes, the cage door is detected using the improved YOLO target segmentation algorithm according to the image information to obtain the pixel coordinates of the cage door area mask; The current cage door position is obtained based on the pixel coordinates of the cage door area mask and the calibration parameters of the RGBD depth camera; The deviation between the robot's current posture and the ideal picking posture is compared. Based on the robot's current posture, the ideal picking posture is used as the target posture, and the deviation between the ideal cage door position and the current cage door position is used as feedback. The TEB motion planning algorithm is used to adjust the posture of the fully automatic dead chicken picking robot.
8. The control method of the fully automatic dead chicken picking robot according to claim 7, characterized in that: If the detection result is yes, the cage door position is identified according to the image information and the posture of the fully automatic dead chicken picking robot is adjusted according to the robot's current posture, the ideal picking posture, the cage door position and the ideal cage door position, further comprising: Determine whether the posture of the fully automatic dead chicken picking robot is an ideal posture, and obtain a first judgment result; If the first judgment result is yes, controlling the cage door opening and closing mechanism to open the cage door; If the first judgment result is no, then return to "compare the deviation between the current posture of the robot and the ideal picking posture, based on the current posture of the robot, use the ideal picking posture as the target posture and the deviation between the ideal cage door position and the current cage door position as feedback, and use the TEB motion planning algorithm to adjust the posture of the fully automatic dead chicken picking robot." 9. The control method of the fully automatic dead chicken picking robot according to claim 6, characterized in that: Identify the location of dead chickens based on image information and control the dead chicken picking mechanism to grab the dead chickens, specifically including: Roughly identify the location of the dead chicken based on image information acquired by the RGBD depth camera at the slider flange to obtain a rough identification position; the image information includes image information acquired by the RGBD depth camera at the slider flange and image information acquired by the RGBD depth camera at the end flange of the robotic arm; Controlling the robotic arm to perform linear motion according to the coarse recognition position and updating image information acquired by the RGBD depth camera of the end flange of the robotic arm; The position and posture of the dead chicken are precisely identified based on the image information obtained by the RGBD depth camera at the end flange of the updated robotic arm to obtain the precise position; Path planning and posture planning are performed on the robotic arm according to the precise position to control the robotic arm to move to the precise position and control the robotic gripper to grab the dead chicken.
10. The control method of the fully automatic dead chicken picking robot according to claim 9, characterized in that: The method further includes identifying the location of the dead chicken according to the image information and controlling the dead chicken picking mechanism to grab the dead chicken: Determine whether the dead chicken is caught and obtain a second determination result; If the second judgment result is yes, then return to "control the cage door opening and closing mechanism to close the cage door"; If the second judgment result is no, return to "perform path planning and posture planning for the robotic arm according to the fine position to control the robotic arm to move to the fine position and control the mechanical gripper to grab the dead chicken."
Citation Information
Patent Citations
Chicken house safety monitoring and dead chicken picking integrated system
CN109343441A
Dead chicken picking device and control system and method of dead chicken picking device
CN112847384A
Dead chicken picking robot for standardized caged chicken house and method of dead chicken picking robot
CN115890677A
Cage rearing dead broiler chicken grabbing system and method
CN116369247A
Dead cage chicken detection method based on deep learning
CN117831074A
Cited By
Device for picking up chickens in bad state
CN121014549A
