CN-122004047-A - Visual-guiding multi-branch vacuum adsorption fruit picking manipulator and picking strategy method

CN122004047ACN 122004047 ACN122004047 ACN 122004047ACN-122004047-A

Abstract

The invention relates to the technical field of agricultural robots, and aims to provide a visual guiding multi-branch vacuum adsorption fruit picking manipulator and a picking strategy method, so as to improve the picking success rate and instantaneity and reduce damage. The technical scheme is that the visual guide multi-branch vacuum fruit picking manipulator comprises a mechanical arm, an actuator arranged at the front end of the mechanical arm, a visual camera and a controller, wherein the actuator comprises a main support, a motor, a clamping jaw mechanism used for grabbing fruits and driven by the motor, and an adsorption mechanism used for adsorbing the fruits. A visual guiding multi-branch vacuum adsorption fruit picking strategy method comprises the following steps of detecting fruit gestures by a visual gesture detection model, generating a picking priority queue, enabling an actuator to be aligned to a first target fruit by a mechanical arm, generating negative pressure by a vacuum generator, grabbing the fruit by the actuator when the adsorption force reaches a stable threshold, and controlling the actuator to execute screwing and separating actions by the mechanical arm, wherein the target fruit is separated from a fruit tree.

Inventors

MENG ZHICHAO
PAN YUCAI
PAN JUAN
YANG QINGHUA
MA HANG

Assignees

浙江农林大学

Dates

Publication Date: 20260512
Application Date: 20260326

Claims (9)

1. A visual guide multi-branch vacuum fruit picking manipulator is characterized by comprising a mechanical arm (1), an actuator arranged at the front end of the mechanical arm, a visual camera and a controller, wherein the actuator comprises a main bracket (4), a motor (11), a clamping jaw mechanism used for grabbing fruits and driven by the motor, and an adsorption mechanism used for adsorbing the fruits.
2. The vision-guided multi-arm vacuum suction fruit picking manipulator of claim 1, wherein the jaw mechanism comprises a fixed seat (14) provided at a front end of the main frame, a movable seat (13) axially slidably positioned on the main frame, a screw (15) rotatably positioned on the main frame and coaxially connected to a motor shaft, a nut engaged with the screw and fixed to the movable seat, and three jaws (7) disposed around the main frame and connecting the movable seat and the fixed seat.
3. The vision-guided multi-branch vacuum suction fruit picking manipulator of claim 1, wherein the motor is fixed at the rear end of the main support, the movable seat is arranged between the motor and the fixed seat, the middle part of the clamping jaw is swingably positioned on the fixed seat, and the rear end of the clamping jaw is slidably positioned on the movable seat.
4. The vision-guided multi-branch vacuum fruit picking manipulator according to claim 1, wherein the adsorption mechanism comprises a vacuum generator (3), a plurality of suckers (8) arranged on the clamping surface of the clamping jaw and the front surface of the fixing plate, a first shunt (10) and a second shunt (12), the vacuum generator is communicated with the first shunt through a main path (5-1), the first shunt is respectively communicated with the four second shunts through four branches (5-2), each second shunt is respectively communicated with a group of suckers through a plurality of branches (5-3), and a logic valve (9) is further arranged on each branch pipe.
5. The vision-guided multi-branch vacuum-suction fruit picking manipulator of claim 1, wherein the controller is electrically connected to a mechanical arm, a vision camera, a motor, a vacuum generator, a first splitter, a second splitter, and a logic valve.
6. A vision-guided multi-arm vacuum-adsorbing fruit picking strategy method employing the vision-guided multi-arm vacuum-adsorbing fruit picking manipulator of any one of claims 1-5, comprising the steps of: S1, acquiring RGB-D images of fruits by using a visual camera, inputting the RGB images into a visual gesture detection model to detect fruit gestures, acquiring three-dimensional positions of the fruits by combining depth image indexes, and judging fruit picking feasibility; S2, picking and sorting the pickable fruits according to a rule of 'near distance-small shielding and stable posture', and generating a picking priority queue; Step S3, acquiring a three-dimensional coordinate of a target fruit at the first position of a picking priority queue, enabling the mechanical arm to axially align an actuator with the target fruit, enabling the actuator to control a clamping jaw mechanism to open, enabling the mechanical arm to move the actuator to a preset position for alignment, and enabling three clamping jaws to surround the target fruit; S4, generating negative pressure by a vacuum generator, and distributing the negative pressure to all branches through the pressure equalization of the first current divider and the second current divider, wherein logic valves of all branches are self-adaptively switched on and off according to the attaching state of the sucker and the target fruit, so that air leakage prevention control is realized; S5, when the adsorption force reaches a stable threshold, the actuator controls the clamping jaw mechanism to fold and execute clamping action, so that fruits are accurately grabbed; And S6, the mechanical arm control actuator performs screwing and separating actions according to the direction of the main shaft of the fruit, so that the target fruit is separated from the fruit tree, and the target fruit is conveyed to an external collecting device.
7. The vision-guided, multi-arm vacuum-adsorbing fruit picking robot of claim 6 wherein the vision pose detection model comprises an MSCA attention module embedded between the Neck neck network and the Head network of the YOLOv-Pose model.
8. The vision-guided multi-arm vacuum suction fruit picking manipulator of claim 7, wherein the feasibility determination comprises entering the next step when the sequence of keypoints comprises end-to-end keypoints and the confidence level is above a threshold, otherwise, leaving to be manually processed or secondarily observed.
9. The vision-guided multi-branch vacuum adsorption fruit picking manipulator of claim 8, wherein the air leakage prevention control comprises the steps that when a sucker is attached to a fruit surface and forms a seal, negative pressure of a branch is quickly increased, a logic valve conducts the branch to establish stable adsorption force, and when the sucker is not attached to a target fruit or has an air leakage trend, the negative pressure of the branch is unchanged or slowly increased, the logic valve automatically throttles/closes the branch, and only trace ventilation of the branch is reserved.

Description

Visual-guiding multi-branch vacuum adsorption fruit picking manipulator and picking strategy method Technical Field The invention relates to the technical field of agricultural robots, in particular to a visual guiding multi-branch vacuum adsorption fruit picking manipulator and a picking strategy method. Background Under the background of large-scale production and labor cost rise of the orchard, the fruit picking robot gradually becomes important equipment of the intelligent orchard. Because the picking operation is in natural illumination and unstructured space environment, the fruits often have the conditions of backlight, branch and leaf shielding, dense overlapping, scale difference, frequent change of postures and the like, the system needs to realize stable and low-damage picking, an end effector needs to be provided with flexible adaptation and reliable grabbing capability, accurate position and posture information is also needed to be provided by visual perception, and a picking strategy coordinated with the executing action is formed to ensure the stability of the whole process of approaching, aligning, adsorbing, grabbing and separating. Existing end effectors mainly include two types, finger grip and vacuum suction. The finger-clamping type actuator has higher grabbing stability, but has limited self-adaptive capability on targets with different sizes and postures, and is easy to cause skin damage due to over-tight clamping or drop due to insufficient clamping. The vacuum adsorption type actuator can reduce extrusion damage, but is difficult to stably establish adsorption sealing under conditions of fruit surface curvature change, local concave areas, branch and leaf interference and the like, and particularly in a multi-sucker structure, branch air leakage is easy to occur when part of suckers are not attached, total negative pressure attenuation is caused, so that adsorption is unstable and conveying is fallen off, and further picking success rate and efficiency are affected. In the aspect of visual perception, the existing system mainly positions a two-dimensional boundary box or a central point, has insufficient expression of constraint information on the main shaft direction of fruits and fragile parts such as fruit stalks, calyx and the like, is easy to have partial characteristic deficiency under the conditions of backlight, shielding and overlapping, so that the attitude estimation is unstable, and the precise labeling of multiple key points can enhance the attitude expression, but obviously increases the labeling cost, easily introduces consistency noise and restricts engineering application. Therefore, a technical scheme capable of providing stable gesture sensing information in a complex orchard environment and cooperating with an end effector and a picking strategy depth is needed, so as to solve the problems of insufficient visual guidance, easy air leakage due to multi-branch adsorption, unreliable alignment and the like, thereby improving the picking success rate, instantaneity and low damage level. Disclosure of Invention The invention aims to overcome the defects in the background technology and provide a visual guiding multi-branch vacuum adsorption fruit picking manipulator and a picking strategy method so as to improve the picking success rate, real-time performance and reduce damage. The technical scheme of the invention is as follows: A visual guide multi-branch vacuum fruit picking manipulator comprises a mechanical arm, an actuator, a visual camera and a controller, wherein the actuator is arranged at the front end of the mechanical arm, and the actuator comprises a main support, a motor, a clamping jaw mechanism used for grabbing fruits and driven by the motor, and an adsorption mechanism used for adsorbing the fruits. The clamping jaw mechanism comprises a fixed seat arranged at the front end of the main support, a movable seat axially and slidably positioned on the main support, a screw rod rotatably positioned on the main support and coaxially connected with a motor rotating shaft, a nut meshed with the screw rod and fixed with the movable seat, and three clamping jaws which are arranged around the main support and connected with the movable seat and the fixed seat. The motor is fixed at the rear end of the main support, the movable seat is arranged between the motor and the fixed seat, the middle part of the clamping jaw can be positioned on the fixed seat in a swinging manner, and the rear end of the clamping jaw can be positioned on the movable seat in a sliding manner. The adsorption mechanism comprises a vacuum generator, a plurality of suckers, a first shunt and a second shunt, wherein the suckers are arranged on the clamping face of the clamping jaw and the front face of the fixing plate, the vacuum generator is communicated with the first shunt through a main path, the first shunt is respectively communicated with the four second shunts through four branches,