Tea leaf tender bud picking device and tea leaf dynamic picking method

Abstract

The application discloses a tea tender bud picking device and a tea dynamic picking method, wherein the tea tender bud picking device comprises a portal-type crawler chassis, a visual detection device, a tea picking device, a tea collecting device and a control device; the visual detection device is used for acquiring a tea garden point cloud image and transmitting the same to the control device; the tea picking device comprises a tea picking mechanism and a tea picking driving mechanism used for driving the tea picking mechanism to perform a tea picking action; the control device is used for processing the received tea garden point cloud image to obtain three-dimensional coordinates of an actual picking point of a tea tender bud, so as to control the tea picking driving mechanism to drive the tea picking mechanism to move to the position of the actual picking point to pick the tea tender bud, and transfer the picked tea tender bud to the tea collecting device. The tea tender bud picking device can realize accurate and non-destructive picking of the tea tender bud, and has higher picking efficiency.

Description

Technical Field

[0001] This invention belongs to the field of tea picking technology, specifically relating to a tea bud picking device and a dynamic tea picking method. Background Technology

[0002] Tea has a vast market in my country, and its production is increasing year by year. The quality of tea production is related to the quality of the tea leaves obtained; high-quality tea is generally produced primarily by harvesting tender tea buds. Currently, tea production faces problems such as labor shortages, and there is an urgent need to mechanize tea picking to improve efficiency.

[0003] Currently available tea-picking robots can be categorized into two types based on their picking methods: clamping-cutting end-gripping mechanisms and finger-like lifting end-gripping mechanisms. The clamping-cutting end-gripping mechanism uses knife-shaped fingers to clamp and pick tea leaves by cutting them during the picking process. However, this method can easily damage the tender tea buds. The finger-like lifting end-gripping mechanism uses soft, smaller fingers to simulate the force exerted by human fingers, picking tea leaves by clamping and lifting. However, this method is prone to missing the target in dynamic environments.

[0004] However, in terms of harvest integrity, the simulated finger-lifting picking mechanism is more suitable for picking tender tea buds; in order to improve the success rate of the simulated finger-lifting picking mechanism, it is necessary to optimize the picking structure and picking method of the simulated finger-lifting picking mechanism. Summary of the Invention

[0005] In order to overcome the shortcomings of the existing technology, the present invention provides a tea bud picking device, which can accurately and without damage pick tea buds and has higher picking efficiency.

[0006] The second objective of this invention is to provide a dynamic tea picking method using a tea bud picking device.

[0007] The technical solution of the present invention to solve the above-mentioned technical problems is:

[0008] A tea bud picking device includes a gantry-type tracked chassis, a vision detection device mounted on the gantry-type tracked chassis, a tea picking device, a tea collecting device, and a control device.

[0009] The visual inspection device is used to acquire point cloud images of the tea garden and transmit them to the control device;

[0010] The tea-picking device includes a tea-picking mechanism and a tea-picking drive mechanism for driving the tea-picking mechanism to perform tea-picking actions, wherein...

[0011] The tea-picking mechanism includes a distal main gripper finger, a distal secondary gripper finger, and a clamping drive mechanism for driving the distal main gripper finger and the distal secondary gripper finger to open or close. The distal main gripper finger is provided with a flexible gripping block; the distal secondary gripper finger is provided with an elastic band. The clamping drive mechanism includes a clamping base and a proximal main gripper finger, a proximal secondary gripper finger, and a clamping motor disposed on the clamping base. One end of the proximal main gripper finger is hinged to the distal main gripper finger, and the other end is rotatably connected to the clamping base via a first rotating shaft. One end of the proximal secondary gripper finger is hinged to the distal secondary gripper finger, and the other end is rotatably connected to the clamping base via a second rotating shaft. Gear sets are disposed on the first and second rotating shafts. The clamping motor is mounted on the clamping base, and the main shaft of the clamping motor is connected to the first / second rotating shaft.

[0012] The control device is used to process the received tea garden point cloud image to obtain the three-dimensional coordinates of the actual picking point of the tea buds. Based on this, the control device controls the tea picking drive mechanism to move to the actual picking point to pick the tea buds and transfer the picked tea buds to the tea collection device.

[0013] Preferably, the gear set includes a first gear disposed on the first rotating shaft and a second gear disposed on the second rotating shaft, wherein the first gear and the second gear mesh, and the outer diameter of the first gear is larger than the outer diameter of the second gear.

[0014] Preferably, a finger clamping link is provided between the distal main clamping finger and the distal secondary clamping finger and the clamping seat respectively; one end of the finger clamping link is hinged to the clamping seat, and the other end is hinged to the distal main clamping finger and the distal secondary clamping finger.

[0015] Preferably, the tea picking drive mechanism includes a connecting seat, an X-axis drive mechanism mounted on the connecting seat for driving the tea picking mechanism to rotate around the X-axis, a Z-axis rotation drive mechanism for driving the X-axis drive mechanism and the tea picking mechanism connected to the X-axis drive mechanism to rotate around the Z-axis, wherein the X-axis drive mechanism includes a U-shaped seat, a connecting frame mounted in the U-shaped seat, and an X-axis motor for driving the connecting frame to rotate; one end of the connecting frame is rotatably connected to the U-shaped seat, and the other end is connected to the clamping seat; the X-axis motor is mounted on the U-shaped seat, and the main shaft of the X-axis motor is connected to the connecting frame; the Z-axis rotation drive mechanism includes a connecting plate mounted on the connecting seat and a Z-axis motor for driving the connecting plate to rotate; the connecting plate is mounted on the U-shaped seat, and the main shaft of the Z-axis motor is connected to the connecting plate.

[0016] Preferably, the tea picking drive mechanism further includes a posture adjustment mechanism for adjusting the posture of the tea picking mechanism; the posture adjustment mechanism includes a connecting shell, multiple sets of adjusting components disposed between the connecting shell and the connecting seat, and an adjustment motor for driving the adjusting components to move, wherein the connecting shell is mounted on the gantry track chassis; there are multiple sets of adjusting components, and correspondingly, there are also multiple sets of adjustment motors; each set of adjusting components includes a swing arm and a connecting rod; the adjustment motor is mounted inside the connecting shell, the main shaft of the adjustment motor is connected to the swing arm, the other end of the swing arm is hinged to the upper end of the connecting rod, and the lower end of the connecting rod is hinged to the connecting seat.

[0017] Preferably, the tea leaf collecting device includes a collecting frame and a fan disposed within the collecting frame. The collecting frame is mounted on the gantry-type tracked chassis and has a collecting port. The fan is disposed on the side wall of the collecting frame opposite to the collecting port. A filter screen is disposed between the collecting port and the fan. A discharge port is disposed at the bottom of the collecting frame and is connected to a collecting box via a discharge pipe. The collecting box is mounted on the gantry-type tracked chassis.

[0018] Preferably, there are two sets of collection boxes, and correspondingly, there are also two sets of discharge pipes; the inlets of the two sets of discharge pipes are connected by a connecting plate; a servo motor is provided on the side wall of the collection frame, and the output shaft of the servo motor is connected to the connecting plate; when one of the collection boxes is full of tea buds, the servo motor drives the connecting plate to rotate, so as to cause the inlet of the discharge pipe connected to the other collection box to connect with the discharge outlet of the collection frame.

[0019] A dynamic tea-picking method includes the following steps:

[0020] Step S1: The gantry-type tracked chassis moves along the tea bushes in the tea garden. The vision inspection device takes a 60° downward angle shot of the tea bushes in front to obtain RGB color images and depth images of the tea garden. Based on the camera intrinsics in the vision inspection device, the depth image is converted into a three-dimensional point cloud.

[0021] Step S2: After acquiring the RGB color image and 3D point cloud data of the tea garden, the YOLOv11 target detection framework is used to identify and select tea buds in the RGB color image; after obtaining the detection box of the tea bud, based on the principle of projection transformation, the center point of the detection box is mapped from the 2D image coordinate system to the 3D point cloud space using the pre-calibrated camera intrinsic and extrinsic parameter matrices; based on the depth information corresponding to the center point of the detection box, the priority planning of tea bud picking is carried out.

[0022] Step S3: Using a direct-pass filtering algorithm, based on the spatial distribution characteristics of tea gardens and tea bushes, the coordinate ranges of the X, Y, and Z axes are set to filter out the three-dimensional point cloud data within the target area; then, the algorithm is used to downsample the target point cloud; next, a statistical filter is used to remove outliers; finally, the Euclidean clustering extraction algorithm is used to segment the three-dimensional point cloud data of tea gardens and tea bushes to obtain the separated branch point cloud clusters;

[0023] Step S4: For the highest priority tea buds, a traversal operation will be performed in the segmented branch point cloud clusters based on the three-dimensional coordinates of the center point of the detection frame; by calculating the Euclidean distance between the three-dimensional coordinate point and all three-dimensional coordinate points in each branch point cloud cluster, the distance values ​​are compared one by one, and the branch point cloud closest to the three-dimensional coordinates of the center point of the detection frame is selected as the tea bud picking reference point.

[0024] Step S5: By introducing the Kalman filter algorithm, the spatial probability distribution of the future moment of the tea bud picking reference point is predicted and optimized. The centroid of the spatial probability distribution is extracted as the optimal estimation point. The approximation speed of the tea picking device is dynamically adjusted by combining the motion delay of the tea picking device and the covariance of the probability distribution. Finally, the compensated three-dimensional coordinates are used as the actual picking point of the tea bud.

[0025] Step S6: The control device generates control commands based on the actual picking coordinates, thereby controlling the tea picking device to pick the tender tea buds.

[0026] Preferably, in step S5, the step of predicting and optimizing the spatial probability distribution of the tea bud picking point at future times using the Kalman filter algorithm is as follows:

[0027] Step S501: Using the three-dimensional coordinates of the tea bud picking reference point output by the first frame of the visual inspection device as the initial observation value, the initial velocity component is estimated by the position difference between adjacent frames, thereby defining the initial state of the tea buds in the base coordinate system of the tea picking device; the initial state includes position, velocity, and acceleration, and the initial covariance matrix is ​​defined.

[0028] Step S502: Use the three-dimensional coordinates of the tea bud picking reference point output in real time by the visual inspection device as the observation input, initialize the observation noise covariance parameter according to the characteristics of visual measurement error, realize the prediction-update loop optimization through the state transition matrix and the observation matrix, and output the spatial probability distribution of the tea bud picking reference point at future time.

[0029] Step S503: Extract the centroid of the spatial probability distribution as the optimal estimation point, and perform coordinate system transformation in combination with the motion delay of the tea picking device to correct the three-dimensional coordinates of the optimal estimation point; predict the dynamic position of the tea picking location within the time series based on the probability distribution covariance, and adjust the approximation speed of the tea picking device accordingly. After the above-mentioned picking prediction coordinate correction and approximation speed adjustment, calculate the contact coordinates between the picking device and the predicted picking point as the actual picking point.

[0030] Preferably, in step S6, the step of the tea picking device picking tea buds is as follows:

[0031] The tea-picking drive mechanism moves the tea-picking machine to the actual picking point;

[0032] Based on the predicted direction of movement of the tea bud, the clamping motor controls the distal main clamping finger and the distal secondary clamping finger to open, so that the elastic band on the distal secondary clamping finger faces the approaching tea bud; when the elastic band contacts the tea bud, the disturbance of the tea bud is suppressed by the elastic band.

[0033] The clamping motor controls the closure of the distal main clamping finger and the distal secondary clamping finger. The flexible clamping block on the distal main clamping finger fits against the elastic band on the distal secondary clamping finger to clamp the tea buds. The tea clamping drive mechanism pulls the clamped tea buds off the branches and then throws them backward into the tea collection device.

[0034] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0035] 1. In the tea bud picking device of the present invention, during the picking process, a point cloud image of the tea garden is first acquired by a visual detection device and transmitted to a control device; the control device processes the received point cloud image of the tea garden to obtain the three-dimensional coordinates of the actual picking point of the tea bud, thereby controlling the tea picking drive mechanism to move to the position of the actual picking point to pick the tea bud, and transferring the picked tea bud to the tea collection device; thus achieving precise picking of tea bud.

[0036] 2. The tea bud picking device of the present invention, based on the predicted movement direction of the tea bud, controls the distal main gripping finger and the distal secondary gripping finger to open, so that the elastic band on the distal secondary gripping finger faces the approaching tea bud; when the elastic band contacts the tea bud, the disturbance of the tea bud is suppressed by the elastic band; the gripping motor controls the distal main gripping finger and the distal secondary gripping finger to close, wherein the flexible gripping block on the distal main gripping finger fits against the elastic band on the distal secondary gripping finger to clamp the tea bud; the tea clamping drive mechanism pulls the clamped tea bud from the branch and then throws it backward into the tea collection device, thereby achieving non-damaging picking of tea leaves.

[0037] 3. The dynamic tea picking method of the present invention uses Kalman prediction and trajectory coordination of the tea picking device to consider the movement time of the tea picking drive mechanism and the tea picking mechanism. It uses Kalman filtering algorithm to predict the tea picking point to compensate for the picking delay, thereby minimizing the phenomenon of empty picking by the tea holding mechanism. Attached Figure Description

[0038] Figure 1 and Figure 2 These are two perspective views of the tea bud picking device of the present invention from different angles.

[0039] Figure 3 and Figure 4 These are two different perspective views of a tea-picking device.

[0040] Figure 5 and Figure 6 These are two perspective views of the tea-picking device (excluding some connecting parts).

[0041] Figure 7 A 3D diagram of a tea-picking facility.

[0042] Figure 8 A 3D view of the tea-picking mechanism (excluding the clamps).

[0043] Figure 9 and Figure 10 Two perspective views of the tea-collecting device.

[0044] Figure 11 This is a schematic diagram illustrating the principle of converting a depth image into a 3D point cloud.

[0045] Figure 12 This is a schematic diagram illustrating the process of converting a depth map into a 3D point cloud.

[0046] Figure 13 A schematic diagram illustrating the principle of selecting reference points for picking tea buds.

[0047] Figure 14This is a flowchart for statistical filtering and obtaining branch point cloud clusters.

[0048] Figure 15 This is a flowchart of dynamic prediction of tea buds based on the Kalman filter algorithm.

[0049] Figure 16 This is a graph showing the prediction results. Detailed Implementation

[0050] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0051] See Figures 1-16 The tea bud picking device of the present invention includes a gantry tracked chassis 1, a vision detection device 2 mounted on the gantry tracked chassis 1, a tea picking device 3, a tea collecting device 4, and a control device.

[0052] See Figures 1-16 The visual inspection device 2 is used to acquire point cloud images of the tea garden and transmit them to the control device. The visual inspection device 2 is a visual sensor, such as a binocular camera.

[0053] See Figures 1-16 The tea picking device 3 includes a tea picking mechanism 9 and a tea picking drive mechanism for driving the tea picking mechanism 9 to perform tea picking actions.

[0054] See Figures 1-16 The tea picking mechanism 9 includes a distal main gripper finger 902, a distal secondary gripper finger 903, and a clamping drive mechanism for driving the distal main gripper finger 902 and the distal secondary gripper finger 903 to open or close. The distal main gripper finger 902 is provided with a flexible clamping block 904; the distal secondary gripper finger 903 is provided with an elastic band 905. The clamping drive mechanism includes a clamping base 901 and a proximal main gripper finger 908, a proximal secondary gripper finger 909, and a clamping motor disposed on the clamping base 901. One end of the proximal main gripper finger 908 is hinged to the distal main gripper finger 902, and the other end is rotatably connected to the clamping base 902 via a first rotating shaft. In 01; one end of the proximal auxiliary gripper finger 909 is hinged to the distal auxiliary gripper finger 903, and the other end is rotatably connected to the clamping seat 901 via a second rotating shaft; gear sets are provided on the first rotating shaft and the second rotating shaft; a gripper connecting rod 907 is provided between the distal main gripper finger 902 and the distal auxiliary gripper finger 903 and the clamping seat 901 respectively; one end of the gripper connecting rod 907 is hinged to the clamping seat 901, and the other end is hinged to the distal main gripper finger 902 and the distal auxiliary gripper finger 903; the clamping motor is mounted on the clamping seat 901, and the main shaft of the clamping motor is connected to the first rotating shaft / second rotating shaft;

[0055] In this embodiment, the gear set includes a first gear 910 disposed on the first rotating shaft and a second gear 911 disposed on the second rotating shaft. The first gear 910 and the second gear 911 mesh, and the outer diameter of the first gear 910 is larger than the outer diameter of the second gear 911. This allows the first gear 910 to rotate, driving the second gear 911 to rotate, causing the distal main gripper finger 902 and the distal secondary gripper finger 903 to move simultaneously. Because the outer diameter of the first gear 910 is larger than that of the second gear 911, and because the linear velocity of the first gear 910 and the linear velocity of the second gear 911 are the same, the angular velocity of the second gear 911 will be greater than that of the first gear 910. The angular velocity ultimately causes the movement speed of the distal secondary gripper 903 to be greater than that of the distal primary gripper 902, thereby allowing the elastic band 905 on the distal secondary gripper 903 to make contact with the tea bud first. The elastic band 905 suppresses the disturbance of the tea bud, thus positioning the tea bud and keeping it stationary. Subsequently, the clamping motor controls the distal primary gripper 902 and the distal secondary gripper 903 to close, wherein the flexible clamping block 904 on the distal primary gripper 902 fits against the elastic band 905 on the distal secondary gripper 903 to clamp the tea bud. The tea clamping drive mechanism drives the tea clamping mechanism 9 to move, thereby pulling the clamped tea bud off the branch and then throwing it backward into the tea collecting device 4.

[0056] See Figures 1-16 The tea picking drive mechanism includes a connecting seat 701, an X-axis drive mechanism 8 mounted on the connecting seat 701 for driving the tea picking mechanism 9 to rotate around the X-axis, a Z-axis rotation drive mechanism 7 for driving the X-axis drive mechanism 8 and the tea picking mechanism 9 connected to the X-axis drive mechanism 8 to rotate around the Z-axis, wherein...

[0057] The X-axis drive mechanism 8 includes a U-shaped base 801, a connecting frame 803 disposed in the U-shaped base 801, and an X-axis motor 802 for driving the connecting frame 803 to rotate; one end of the connecting frame 803 is rotatably connected to the U-shaped base 801, and the other end is connected to the clamping base 901; the X-axis motor 802 is mounted on the U-shaped base 801, and the main shaft of the X-axis motor 802 is connected to the connecting frame 803;

[0058] The Z-axis rotation drive mechanism 7 includes a connecting disk 703 disposed on the connecting seat 701 and a Z-axis motor 702 for driving the connecting disk 703 to rotate; the connecting disk 703 is mounted on the U-shaped seat 801, and the spindle of the Z-axis motor 702 is connected to the connecting disk 703.

[0059] In this embodiment, the tea picking drive mechanism further includes a posture adjustment mechanism for adjusting the posture of the tea picking mechanism 9; the posture adjustment mechanism includes a connecting shell 601, multiple sets of adjusting components disposed between the connecting shell 601 and the connecting seat 701, and an adjustment motor 602 for driving the adjusting components to move, wherein there are multiple sets of adjusting components, and correspondingly, there are also multiple sets of adjustment motors 602; each set of adjusting components includes a swing arm 603 and a connecting rod 604; the adjustment motor 602 is installed inside the connecting shell 601, the main shaft of the adjustment motor 602 is connected to the swing arm 603, the other end of the swing arm 603 is hinged to the upper end of the connecting rod 604, and the lower end of the connecting rod 604 is hinged to the connecting seat 701.

[0060] Through the coordinated movement of the X-axis drive mechanism 8, the Z-axis rotation drive mechanism 7, and the posture adjustment mechanism, the tea picking mechanism 9 is driven to move in all directions, so that the tea picking mechanism 9 can move to the actual picking point of the tea buds, pick the tea buds, and transfer the picked tea buds to the tea collecting device 4.

[0061] In this embodiment, there are two sets of connecting rods 604 in each set of adjusting components, and the two sets of connecting rods 604 are arranged side by side.

[0062] See Figures 1-16 The control device is used to process the received tea garden point cloud image to obtain the three-dimensional coordinates of the actual picking point of the tea buds, and thereby control the tea picking drive mechanism to move the tea picking mechanism 9 to the position of the actual picking point to pick the tea buds, and transfer the picked tea buds to the tea collecting device 4.

[0063] See Figures 1-16 The tea collection device 4 includes a collection frame 401 and a fan 404 disposed in the collection frame 401. The collection frame 401 is mounted on the gantry tracked chassis 1 and has a collection port. The fan 404 is disposed on the side wall of the collection frame 401 opposite to the collection port. A filter screen 402 is disposed between the collection port and the fan 404. A discharge port 403 is disposed at the bottom of the collection frame 401 and is connected to the collection box 5 through a discharge pipe 407. The collection box 5 is mounted on the gantry tracked chassis 1.

[0064] In this embodiment, there are two sets of collection boxes 5, and correspondingly, there are also two sets of discharge pipes 407; the inlets of the two sets of discharge pipes 407 are connected by a connecting plate 406; a servo motor 405 is provided on the side wall of the collection frame 401, and the output shaft of the servo motor 405 is connected to the connecting plate 406; when one of the collection boxes 5 is full of tea buds, the servo motor 405 drives the connecting plate 406 to rotate, so as to cause the inlet of the discharge pipe 407 connected to the other collection box 5 to connect with the discharge port 403 of the collection frame 401.

[0065] With the above setup, the picked tea buds are drawn by the suction of the fan 404 and blocked by the filter screen 402, sliding towards the only outlet 403 of the collection frame 401, and then entering the collection box 5 through the outlet pipe 407, thus completing the picking process. When one of the collection boxes 5 is full of tea buds, the servo motor 405 will drive the connecting plate 406 to rotate, so that the inlet of the outlet pipe 407 connected to the other collection box 5 will connect with the outlet 403 of the collection frame 401, so that the picked tea buds will enter the corresponding collection box 5 through the outlet pipe 407.

[0066] See Figures 1-16 The dynamic tea picking method of the present invention includes the following steps:

[0067] Step S1: The gantry tracked chassis 1 moves along the tea bushes in the tea garden. The vision inspection device 2 takes a 60° downward angle shot of the tea bushes in front to obtain RGB color images and depth images of the tea garden. Based on the camera intrinsics in the vision inspection device 2, the depth image is converted into a three-dimensional point cloud.

[0068] In this embodiment, the principle of converting a depth image into a 3D point cloud is as follows: Figure 11 As shown, O c Centered at the camera, (x) ω y ω , z ω (x) represents the real-world coordinates, z = 0. c The depth from the target to the camera. For image coordinates, The coordinates of the image center;

[0069] Therefore, the formula for converting a depth image from two-dimensional coordinates to a three-dimensional point cloud is:

[0070]

[0071] Step S2: After acquiring the RGB color image and 3D point cloud data of the tea garden, the YOLOv11 target detection framework is used to identify and select tea buds in the RGB color image; after obtaining the detection box of the tea bud, based on the principle of projection transformation, the center point of the detection box is mapped from the 2D image coordinate system to the 3D point cloud space using the pre-calibrated camera intrinsic and extrinsic parameter matrices; based on the depth information corresponding to the center point of the detection box, the priority planning of tea bud picking is carried out.

[0072] Step S3: Using a direct-pass filtering algorithm, based on the spatial distribution characteristics of tea gardens and tea bushes, the coordinate ranges of the X, Y, and Z axes are set to filter out the 3D point cloud data within the target area; then, the algorithm is used to downsample the target point cloud to reduce the point cloud density; next, a statistical filter is used to remove outliers, achieving point cloud downsampling and noise reduction; finally, the Euclidean clustering extraction algorithm is used to segment the 3D point cloud data of tea gardens and tea bushes to obtain the separated branch point cloud clusters;

[0073] Step S4: For the highest priority tea buds, a traversal operation will be performed in the segmented branch point cloud clusters based on the three-dimensional coordinates of the center point of the detection frame; by calculating the Euclidean distance between the three-dimensional coordinate point and all three-dimensional coordinate points in each branch point cloud cluster, the distance values ​​are compared one by one, and the branch point cloud closest to the three-dimensional coordinates of the center point of the detection frame is selected as the tea bud picking reference point.

[0074] Step S5: By introducing the Kalman filter algorithm, the spatial probability distribution of the future moment of the tea bud picking reference point is predicted and optimized. The centroid of the spatial probability distribution is extracted as the optimal estimation point. The approximation speed of the tea picking device 3 is dynamically adjusted by combining the motion delay of the tea picking device 3 and the probability distribution covariance. Finally, the compensated three-dimensional coordinates are used as the actual picking point of the tea bud.

[0075] Step S6: The control device generates control commands based on the actual picking coordinates, thereby controlling the tea picking device 3 to pick the tender tea buds.

[0076] In step S5, the process of predicting and optimizing the spatial probability distribution of future tea bud picking points using the Kalman filter algorithm is as follows: Figure 15 As shown, specifically:

[0077] Step S501: Using the three-dimensional coordinates of the tea bud picking reference point output by the first frame of the visual inspection device 2 as the initial observation value, the initial velocity component v0 is estimated by the position difference between adjacent frames, thereby defining the initial state of the tea buds in the base coordinate system of the tea picking device 3; the initial state includes position, velocity, and acceleration, and the initial covariance matrix is ​​defined.

[0078] In this embodiment, the motion and observation equations and Gaussian distribution of the Kalman filter algorithm are as follows:

[0079]

[0080] In the formula: x k+1 For state variables, u k For input variables, w k For process noise, A k Let B be the state transition matrix. k To control the input matrix; z k C is the output vector (the observations obtained by the visual sensor). k The observation matrix (describing the state vector x) k To the observed value z k (linear relationship between them), v k To measure noise (error in the measurement process); Q k and R k These are the process noise covariance matrices and the measurement noise covariance matrices, respectively; where the process noise w is mentioned in the formula... k and v k All variables have a mean of 0 and are Gaussian distributed random variables.

[0081] Velocity difference estimation:

[0082]

[0083] Set state variables:

[0084]

[0085] Initial covariance estimation:

[0086]

[0087] In the formula: This is the expected estimate of the state; Let x be the state variable predicted at time k-1. k-1 The actual state variables at time k-1; Let be the covariance of the state error.

[0088] Step S502: Take the three-dimensional coordinates of the tea bud picking reference point output in real time by the visual inspection device 2 as the observation input, initialize the observation noise covariance parameter according to the characteristics of visual measurement error, realize the cyclic optimization of prediction-update iteration through the state transition matrix and the observation matrix, and output the spatial probability distribution of the tea bud picking reference point at future time.

[0089] The state transition matrix is ​​as follows:

[0090]

[0091] In the formula: the twig and bud are equivalent to a simple pendulum motion k=θ / L, where k is the restoring force coefficient of the twig, g is the weight acceleration, L is the equivalent pendulum length, and c is the damping coefficient of the twig.

[0092] The prediction-update iteration process is as follows:

[0093] Next-moment state prediction:

[0094]

[0095] Predicted state covariance matrix:

[0096]

[0097] Kalman gain calculation:

[0098]

[0099] State prediction update:

[0100]

[0101] Update the covariance matrix:

[0102]

[0103] In the formula: Let A be the state variable predicted at time k. k-1 Let k be the state transition matrix from time k-1 to time k. For the predicted state at time k-1, B k-1 To control the input matrix, u k-1 To control the input, w k-1 This is process noise; For the predicted covariance matrix, P k-1 Let Q be the covariance matrix of the previous time step. k-1 K is the process noise covariance matrix; k To balance the weighted Kalman gain between predictions and actual observations, C k To describe the observation matrix that maps the state space to the observation space, R... k The observation noise covariance matrix characterizing the observation error; z k These are actual observed values. For the predicted observations, The observation residual, P, is used to characterize the difference between predicted and observed values. k For the updated covariance matrix, K k C k This is the product of the Kalman gain and the observation matrix.

[0104] Therefore, the entire implementation process of the Kalman filter algorithm is as follows:

[0105] Based on the state vector changes of tea buds in adjacent frames, a state transition matrix A containing position, velocity, and acceleration is constructed. k-1 Based on the state vector and state transition matrix of the previous frame, predict the state of the tea buds in the next frame. And based on the covariance P of the previous frame k-1 and process noise Q k-1 Obtain the predicted state covariance matrix at the current moment. (i.e., the uncertainty or error range of the current state); with the observation matrix C k Current observation noise R k Covariance matrix and predicted covariance matrix Calculate the Kalman gain K k and the current predicted state and the current observation value z k By combining these steps, we obtain the corrected state update. Ultimately, through Kalman gain K k With predicted covariance The updated covariance matrix P is calculated. k .

[0106] Step S503: Extract the centroid of the spatial probability distribution as the optimal estimation point, and perform coordinate system transformation in combination with the motion delay of the tea picking device to correct the three-dimensional coordinates of the optimal estimation point; predict the dynamic position of the tea picking location within the time series based on the probability distribution covariance, and adjust the approximation speed of the tea picking device accordingly. After the above-mentioned picking prediction coordinate correction and approximation speed adjustment, calculate the contact coordinates between the picking device and the predicted picking point as the actual picking point.

[0107] In this embodiment, the centroid of the spatial probability distribution is extracted as the optimal estimated point for the location of the tea bud. The centroid comprehensively considers the probability of the tea bud appearing at each possible location, and can most representatively reflect the actual location of the tea bud, laying the foundation for subsequent precise operations. Given the unavoidable motion delay between receiving the instruction and executing the picking action, to ensure that the tea picking device 33 can accurately pick the bud when it reaches the target location, a coordinate system transformation is performed based on this delay characteristic. The three-dimensional coordinates of the optimal estimated point are corrected, accurately transforming it to the actual location after considering the delay factor. The approximation speed of the tea picking device 3 is dynamically adjusted according to the probability distribution covariance. Among them, covariance directly reflects the uncertainty of the position of tea buds: when the covariance is large, it means that the position of tea buds is highly uncertain. In order to avoid mispicking, the tea picking device 3 will automatically slow down to allow sufficient time for accurate calibration. When the covariance is small, it indicates that the position of tea buds is relatively certain. The tea picking device 3 can appropriately increase its speed to effectively improve picking efficiency. After the dual optimization of the above coordinate correction and speed adjustment, the final three-dimensional coordinates are the actual picking points. Sending these coordinates to the picking device can achieve accurate picking of tea buds.

[0108] In step S6, the tea picking device 3 picks the tender tea buds as follows:

[0109] The tea picking drive mechanism moves the tea picking mechanism 9 to the actual picking point;

[0110] Based on the predicted direction of movement of the tea buds, the clamping motor controls the distal main clamping finger 902 and the distal secondary clamping finger 903 to open, so that the elastic band 905 on the distal secondary clamping finger 903 faces the approaching tea bud; when the elastic band 905 contacts the tea bud, the disturbance of the tea bud is suppressed by the elastic band 905.

[0111] The clamping motor controls the closure of the distal main clamping finger 902 and the distal secondary clamping finger 903. The flexible clamping block 904 on the distal main clamping finger 902 fits against the elastic band 905 on the distal secondary clamping finger 903 to clamp the tea buds. The tea clamping drive mechanism pulls the clamped tea buds off the branches and then throws them backward into the tea collecting device 4.

[0112] The above are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above content. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for dynamic tea picking, characterized in that, Tea leaves are harvested using a tea bud picking device. The tea bud picking device includes a gantry-type tracked chassis, a vision detection device mounted on the gantry-type tracked chassis, a tea picking device, a tea collecting device, and a control device. The vision detection device acquires point cloud images of the tea garden and transmits them to the control device. The tea picking device includes a tea picking mechanism and a tea picking drive mechanism for driving the tea picking mechanism to perform tea picking actions. The tea picking mechanism includes a distal main gripper finger, a distal secondary gripper finger, and a clamping drive mechanism for driving the distal main gripper finger and the distal secondary gripper finger to open or close. A flexible clamping block is provided on the distal main gripper finger; an elastic band is provided on the distal secondary gripper finger. The clamping drive mechanism includes a clamping seat and a proximal main gripper finger mounted on the clamping seat. The system includes a proximal auxiliary gripper finger and a gripping motor; one end of the proximal main gripper finger is hinged to the distal main gripper finger, and the other end is rotatably connected to the gripping seat via a first rotating shaft; one end of the proximal auxiliary gripper finger is hinged to the distal auxiliary gripper finger, and the other end is rotatably connected to the gripping seat via a second rotating shaft; gear sets are provided on the first and second rotating shafts; the gripping motor is mounted on the gripping seat, and the main shaft of the gripping motor is connected to the first / second rotating shaft; the control device is used to process the received tea garden point cloud image to obtain the three-dimensional coordinates of the actual picking point of the tea buds, thereby controlling the tea picking drive mechanism to move to the position of the actual picking point to pick the tea buds, and transfer the picked tea buds to the tea collecting device; The specific harvesting method includes the following steps: Step S1: The gantry-type tracked chassis moves along the tea bushes in the tea garden. The vision inspection device takes a 60° downward angle shot of the tea bushes in front to obtain RGB color images and depth images of the tea garden. Based on the camera intrinsics in the vision inspection device, the depth image is converted into a three-dimensional point cloud. Step S2: After acquiring the RGB color image and 3D point cloud data of the tea garden, the YOLOv11 object detection framework is used to identify and select tea buds in the RGB color image. After obtaining the detection box of the tea bud, based on the principle of projection transformation, the center point of the detection box is mapped from the 2D image coordinate system to the 3D point cloud space using the pre-calibrated camera intrinsic and extrinsic parameter matrices. Based on the depth information corresponding to the center point of the detection box, the priority of tea bud picking is planned. Step S3: Using a direct-pass filtering algorithm, based on the spatial distribution characteristics of tea gardens and tea bushes, the coordinate ranges in the X, Y, and Z axes are set to filter out the three-dimensional point cloud data within the target area; then, the algorithm is used to downsample the target point cloud; next, a statistical filter is used to remove outliers; finally, the Euclidean clustering extraction algorithm is used to segment the three-dimensional point cloud data of tea gardens and tea bushes to obtain the separated branch point cloud clusters; Step S4: For the highest priority tea buds, a traversal operation will be performed in the segmented branch point cloud clusters based on the three-dimensional coordinates of the center point of the detection frame; by calculating the Euclidean distance between the three-dimensional coordinate point and all three-dimensional coordinate points in each branch point cloud cluster, the distance values ​​are compared one by one, and the branch point cloud closest to the three-dimensional coordinates of the center point of the detection frame is selected as the tea bud picking reference point. Step S5: By introducing the Kalman filter algorithm, the spatial probability distribution of the tea bud picking reference point at future moments is predicted and optimized. The centroid of the spatial probability distribution is extracted as the optimal estimation point. The approximation speed of the tea picking device is dynamically adjusted by combining the motion delay of the tea picking device and the covariance of the probability distribution. Finally, the compensated three-dimensional coordinates are taken as the actual picking point of the tea bud. The steps for predicting and optimizing the spatial probability distribution of the tea bud picking point at future moments using the Kalman filter algorithm are as follows: Step S501: Using the three-dimensional coordinates of the tea bud picking reference point output by the first frame of the visual inspection device as the initial observation value, the initial velocity component is estimated by the position difference between adjacent frames, thereby defining the initial state of the tea bud in the base coordinate system of the tea picking device; the initial state includes position, velocity, and acceleration, and the initial covariance matrix is ​​defined. Step S6: The control device generates control commands based on the actual picking coordinates, thereby controlling the tea picking device to pick the tender tea buds.

2. The dynamic tea picking method according to claim 1, characterized in that, The gear set includes a first gear disposed on the first rotating shaft and a second gear disposed on the second rotating shaft, wherein the first gear and the second gear mesh, and the outer diameter of the first gear is larger than the outer diameter of the second gear.

3. The dynamic tea picking method according to claim 1, characterized in that, The distal main clamping finger and the distal secondary clamping finger are respectively provided with a clamping link between them and the clamping seat; one end of the clamping link is hinged to the clamping seat, and the other end is hinged to the distal main clamping finger and the distal secondary clamping finger.

4. The dynamic tea picking method according to claim 1, characterized in that, The tea picking drive mechanism includes a connecting seat, an X-axis drive mechanism mounted on the connecting seat for driving the tea picking mechanism to rotate around the X-axis, a Z-axis rotation drive mechanism for driving the X-axis drive mechanism and the tea picking mechanism connected to the X-axis drive mechanism to rotate around the Z-axis, wherein the X-axis drive mechanism includes a U-shaped seat, a connecting frame mounted in the U-shaped seat, and an X-axis motor for driving the connecting frame to rotate; one end of the connecting frame is rotatably connected to the U-shaped seat, and the other end is connected to the clamping seat; the X-axis motor is mounted on the U-shaped seat, and the main shaft of the X-axis motor is connected to the connecting frame; the Z-axis rotation drive mechanism includes a connecting plate mounted on the connecting seat and a Z-axis motor for driving the connecting plate to rotate; the connecting plate is mounted on the U-shaped seat, and the main shaft of the Z-axis motor is connected to the connecting plate.

5. The dynamic tea picking method according to claim 4, characterized in that, The tea picking drive mechanism further includes a posture adjustment mechanism for adjusting the posture of the tea picking mechanism; the posture adjustment mechanism includes a connecting shell, multiple sets of adjusting components disposed between the connecting shell and the connecting seat, and an adjustment motor for driving the adjusting components to move, wherein the connecting shell is mounted on the gantry track chassis; there are multiple sets of adjusting components, and correspondingly, there are also multiple sets of adjustment motors; each set of adjusting components includes a swing arm and a connecting rod; the adjustment motor is mounted inside the connecting shell, the main shaft of the adjustment motor is connected to the swing arm, the other end of the swing arm is hinged to the upper end of the connecting rod, and the lower end of the connecting rod is hinged to the connecting seat.

6. The dynamic tea picking method according to claim 5, characterized in that, The tea leaf collecting device includes a collecting frame and a fan installed in the collecting frame. The collecting frame is mounted on the gantry-type tracked chassis and has a collecting port. The fan is installed on the side wall of the collecting frame opposite to the collecting port. A filter screen is installed between the collecting port and the fan. The bottom of the collecting frame has a discharge port, which is connected to a collecting box via a discharge pipe. The collecting box is mounted on the gantry-type tracked chassis.

7. The dynamic tea picking method according to claim 6, characterized in that, There are two sets of collection boxes, and correspondingly, there are also two sets of discharge pipes; the inlets of the two sets of discharge pipes are connected by a connecting plate; a servo motor is provided on the side wall of the collection frame, and the output shaft of the servo motor is connected to the connecting plate; when one of the collection boxes is full of tea buds, the servo motor drives the connecting plate to rotate, so as to cause the inlet of the discharge pipe connected to the other collection box to connect with the outlet of the collection frame.

8. The dynamic tea picking method according to claim 7, characterized in that, In step S5, the step of predicting and optimizing the spatial probability distribution of future tea bud picking points using the Kalman filter algorithm also includes: Step S502: Use the three-dimensional coordinates of the tea bud picking reference point output in real time by the visual inspection device as the observation input, initialize the observation noise covariance parameter according to the characteristics of visual measurement error, realize the prediction-update loop optimization through the state transition matrix and the observation matrix, and output the spatial probability distribution of the tea bud picking reference point at future time. Step S503: Extract the centroid of the spatial probability distribution as the optimal estimation point, and perform coordinate system transformation in combination with the motion delay of the tea picking device to correct the three-dimensional coordinates of the optimal estimation point; dynamically adjust the approximation speed of the tea picking device according to the covariance of the probability distribution, and after the above coordinate correction and approximation speed adjustment, the obtained three-dimensional coordinates are used as the actual picking point.

9. The dynamic tea picking method according to claim 1, characterized in that, In step S6, the step of the tea picking device picking tea buds is as follows: The tea-picking drive mechanism moves the tea-picking machine to the actual picking point; Based on the predicted direction of movement of the tea bud, the clamping motor controls the distal main clamping finger and the distal secondary clamping finger to open, so that the elastic band on the distal secondary clamping finger faces the approaching tea bud; when the elastic band contacts the tea bud, the disturbance of the tea bud is suppressed by the elastic band. The clamping motor controls the closure of the distal main clamping finger and the distal secondary clamping finger. The flexible clamping block on the distal main clamping finger fits against the elastic band on the distal secondary clamping finger to clamp the tea buds. The tea clamping drive mechanism pulls the clamped tea buds off the branches and then throws them backward into the tea collection device.

Images