Intelligent picking robot based on deep learning algorithm machine vision

Abstract

The invention belongs to the technical field of intelligent picking robots, and discloses an intelligent picking robot based on deep learning algorithm machine vision, which comprises a main body mechanism, a rotating mechanism is arranged in the main body mechanism, a picking mechanism is arranged in the rotating mechanism, and a fixing mechanism is arranged on the rotating mechanism. The picking mechanism comprises a hydraulic cylinder, a push rod is slidably connected to the hydraulic cylinder, guide rods are fixedly connected to the two sides of the push rod respectively, and the end, away from the hydraulic cylinder, of the push rod is in threaded connection with a bolt; through cooperation of structures such as the movable plate and the clamping jaws, fruits are picked conveniently, the push rod moves in the direction away from the hydraulic cylinder through the hydraulic cylinder, the movable plate is driven to move in the direction close to the fruits, and the clamping jaws are arc-shaped, so that the fruits are picked conveniently. Therefore, the movable plate moves to enable the three clamping jaws to move towards each other through the clamping grooves, and the fruits are clamped in the middle.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent picking robots, and specifically is an intelligent picking robot based on deep learning algorithm machine vision. Background Art

[0002] Smart harvesting robots are a key development in modern agricultural technology, combining the latest research results from multiple fields such as mechanical engineering, electronic engineering, computer science and artificial intelligence. The core of these robots is to use a machine vision system driven by deep learning algorithms to identify and locate ripe fruits, and then the robotic arms complete the picking work. Deep learning is an artificial intelligence technology that simulates the way the human brain processes information through a neural network model and can learn complex patterns and features from large amounts of data. In smart harvesting robots, deep learning algorithms train machine vision systems to identify fruits of different types, maturity and health conditions. This process usually requires a large amount of image data to train the algorithm model so that it has highly accurate recognition capabilities. The application of smart harvesting robots has greatly improved agricultural production efficiency, reduced the intensity of manual labor, and helped solve the problem of agricultural labor shortage. In addition, these robots can also work in adverse weather conditions, reducing crop losses caused by weather factors.

[0003] Most of the existing intelligent picking robots based on deep learning algorithm and machine vision pick fruits by pulling, which not only easily causes damage to the fruits during pulling, but may also cause bending of the branches of the plants, thereby affecting the production of the fruits. At the same time, most of the picking structures are fixed and can only pick fruits of a single specification, and the scope of application is small. Therefore, an intelligent picking robot based on deep learning algorithm and machine vision is proposed, which has the effect of avoiding damage to the fruits caused by pulling and picking and increasing the scope of application of the device. Summary of the invention

[0004] In order to solve the problem that the above-mentioned background technology proposes to pick fruits by pulling, which not only easily causes damage to the fruits during pulling, but may also cause bending of the branches of the plants, thereby affecting the production of the fruits, and most of the picking structures are fixed and can only pick fruits of a single specification, and have a small scope of application, the present invention provides an intelligent picking robot based on deep learning algorithm machine vision.

[0005] To achieve the above object, the present invention provides the following technical solution: an intelligent picking robot based on deep learning algorithm machine vision, comprising a main body mechanism, a rotating mechanism is arranged inside the main body mechanism, a picking mechanism is arranged inside the rotating mechanism, and a fixing mechanism is arranged on the rotating mechanism;

[0006] The picking mechanism includes a hydraulic cylinder, a push rod is slidably connected to the hydraulic cylinder, guide rods are fixedly connected to both sides of the push rod, a bolt is threadedly connected to one end of the push rod away from the hydraulic cylinder, a movable plate is rotatably connected to one end of the push rod away from the hydraulic cylinder, three slots are evenly arranged on the movable plate, clamps are slidably connected inside the three slots, the ends of the three clamps close to the push rod are rotatably connected to the mounting seat, a positioning plate is fixedly connected to one end of the mounting seat close to the push rod, and fixing holes are respectively arranged on both sides of the positioning plate.

[0007] Preferably, the movable plate is located on a side of the bolt close to the push rod, and an end of the push rod away from the hydraulic cylinder penetrates the surface of the mounting seat and extends to a side of the mounting seat close to the clamping claw.

[0008] Preferably, the rotating mechanism includes a fixed seat, the inner wall of which is provided with two linear guide grooves, the two linear guide grooves are provided with an arc guide groove on the side away from the hydraulic cylinder, the surface of the fixed seat is fixedly connected to a limiting block, and a positioning groove is provided on the end of the fixed seat away from the hydraulic cylinder.

[0009] Preferably, the size of the positioning groove is matched with the size of the positioning plate, the positioning groove and the positioning plate are clamped together, and the hydraulic cylinder is slidably connected to the inner wall of the fixing seat.

[0010] Preferably, the two guide rods are slidably connected to the two linear guide grooves and the two arc guide grooves respectively, and the fixing seat is located between the hydraulic cylinder and the mounting seat.

[0011] Preferably, the fixing mechanism comprises two handles, the ends of the two handles close to each other are fixedly connected with a connecting rod, the end of the connecting rod away from the handles is fixedly connected with an inclined block, a sliding groove is provided inside the fixing seat, and the inclined block is elastically connected to the inner wall of the sliding groove through a spring.

[0012] Preferably, the two handles are located on both sides of the fixing seat, and the inclined block penetrates the inner wall of the sliding groove and extends to the inside of the positioning groove.

[0013] Preferably, the size of the inclined block matches the size of the fixing hole, and the inclined block is snap-fitted to the fixing hole.

[0014] Preferably, the main structure includes a vehicle body, the top of which is fixedly connected to a robot body, the top of which is rotatably connected to a camera, a collecting groove is provided on the top of the vehicle body, and a limiting groove is provided inside the robot body.

[0015] Preferably, the hydraulic cylinder is fixedly connected to the robot body, the size of the limit block is matched with the size of the limit slot, the limit block is rotatably connected to the limit slot, and the fixing seat is rotatably connected to the robot body.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] The present invention facilitates the picking of fruits by arranging the cooperation of structures such as the movable plate and the clamping claws. The hydraulic cylinder moves the push rod in the direction away from the hydraulic cylinder, driving the movable plate to move in the direction close to the fruit. Because the shape of the clamping claw is an arc, the movement of the movable plate will cause the three clamping claws to move in the direction close to each other through the clamping groove, clamping the fruit in the middle. At this time, the guide rod moves to the side of the linear guide groove away from the hydraulic cylinder. When the push rod continues to move, the guide rod squeezes the positioning groove, so that the fixed seat rotates inside the robot body under the constraint of the positioning groove, thereby driving the mounting seat fixedly connected thereto to rotate, so that the clamping claw of the fruit holding column drives the fruit to rotate, and the fruit is separated from the branch by the rotational force, avoiding the situation that the fruit is damaged by pulling and picking.

[0018] The present invention facilitates the replacement and loading and unloading of the clamping jaws by arranging the cooperation of structures such as the handle and the inclined block. When fruits of different sizes need to be picked or the clamping jaws are damaged and need to be replaced, it is only necessary to rotate and remove the bolts first to release the clamping connection between the movable plate and the push rod, and then pull the handle away from the fixed seat to drive the inclined block to move to the inside of the slide groove through the connecting rod, so that the clamping connection between the inclined block and the fixed hole is released, and the mounting seat can be removed from the fixed seat to complete the disassembly, thereby improving the application range and working efficiency of the device. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the structure of the present invention;

[0020] Figure 2 It is a schematic diagram of the structural matching relationship between the clamping jaw and the mounting seat of the present invention;

[0021] Figure 3 It is a schematic diagram of the cross-sectional structure of the fixing mechanism of the present invention;

[0022] Figure 4 For the present invention Figure 3 The enlarged structural diagram at A in the middle;

[0023] Figure 5 For the present invention Figure 3 The enlarged structural diagram at B in the middle;

[0024] Figure 6 It is a schematic diagram of the structural matching relationship between the push rod and the mounting seat of the present invention;

[0025] Figure 7 It is a schematic diagram of the structural matching relationship between the positioning plate and the positioning groove of the present invention.

[0026] In the figure: 1. picking mechanism; 101. hydraulic cylinder; 102. push rod; 103. guide rod; 104. bolt; 105. movable plate; 106. slot; 107. clamping claw; 108. mounting seat; 109. positioning plate; 110. fixing hole; 2. rotating mechanism; 201. fixing seat; 202. linear guide groove; 203. arc guide groove; 204. limit block; 205. positioning groove; 3. fixing mechanism; 301. handle; 302. connecting rod; 303. inclined block; 304. spring; 305. slide groove; 4. main mechanism; 401. vehicle body; 402. robot body; 403. camera; 404. collecting trough; 405. limit groove. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] like Figures 1 to 4 As shown, the present invention provides an intelligent picking robot based on deep learning algorithm machine vision, including a main body mechanism 4, a rotating mechanism 2 is arranged inside the main body mechanism 4, a picking mechanism 1 is arranged inside the rotating mechanism 2, and a fixing mechanism 3 is arranged on the rotating mechanism 2;

[0029] The picking mechanism 1 includes a hydraulic cylinder 101, on which a push rod 102 is slidably connected, and guide rods 103 are fixedly connected on both sides of the push rod 102, and a bolt 104 is threadedly connected to one end of the push rod 102 away from the hydraulic cylinder 101, and a movable plate 105 is rotatably connected to one end of the push rod 102 away from the hydraulic cylinder 101, and three slots 106 are evenly arranged on the movable plate 105, and the insides of the three slots 106 are all slidably connected with clamping claws 107, and the ends of the three clamping claws 107 close to the push rod 102 are rotatably connected to a mounting seat 108, and a positioning plate 109 is fixedly connected to one end of the mounting seat 108 close to the push rod 102, and fixing holes 110 are respectively arranged on both sides of the positioning plate 109.

[0030] The movable plate 105 is located on a side of the bolt 104 close to the push rod 102 . An end of the push rod 102 away from the hydraulic cylinder 101 penetrates the surface of the mounting seat 108 and extends to a side of the mounting seat 108 close to the clamping claw 107 .

[0031] The above scheme is adopted: by setting the cooperation of structures such as the movable plate 105 and the clamping claw 107, it is convenient to pick the fruit. The push rod 102 is moved in the direction away from the hydraulic cylinder 101 by the hydraulic cylinder 101, driving the movable plate 105 to move in the direction close to the fruit. Because the shape of the clamping claw 107 is arc-shaped, the movement of the movable plate 105 will cause the three clamping claws 107 to move in the direction close to each other through the card slot 106, clamping the fruit in the middle. At this time, the guide rod 103 moves to the side of the linear guide groove 202 away from the hydraulic cylinder 101. When the push rod 102 continues to move, the guide rod 103 will squeeze the positioning groove 205, so that the fixed seat 201 rotates inside the robot body 402 under the constraint of the positioning groove 205, thereby driving the mounting seat 108 fixedly connected thereto to rotate, so that the clamping claw 107 clamping the fruit drives the fruit to rotate, and the fruit is separated from the branch by the rotational force, avoiding the situation where the fruit is damaged by pulling and picking.

[0032] like Figures 3 to 7 As shown, the rotating mechanism 2 includes a fixed seat 201, the inner wall of which is provided with two linear guide grooves 202, an arc guide groove 203 is provided on the side of the two linear guide grooves 202 away from the hydraulic cylinder 101, a limiting block 204 is fixedly connected to the surface of the fixed seat 201, and a positioning groove 205 is provided on the end of the fixed seat 201 away from the hydraulic cylinder 101.

[0033] The size of the positioning groove 205 is compatible with the size of the positioning plate 109, the positioning groove 205 and the positioning plate 109 are clamped together, the hydraulic cylinder 101 is slidingly connected to the inner wall of the fixed seat 201, the two guide rods 103 are slidingly connected to the two linear guide grooves 202 and the two arc guide grooves 203 respectively, and the fixed seat 201 is located between the hydraulic cylinder 101 and the mounting seat 108.

[0034] The fixing mechanism 3 includes two handles 301, and the ends of the two handles 301 that are close to each other are fixedly connected to a connecting rod 302, and the end of the connecting rod 302 that is away from the handles 301 is fixedly connected to an inclined block 303. A sliding groove 305 is opened inside the fixing seat 201, and the inclined block 303 is elastically connected to the inner wall of the sliding groove 305 through a spring 304.

[0035] The two handles 301 are located on both sides of the fixing seat 201 , and the inclined block 303 passes through the inner wall of the slide groove 305 and extends to the inside of the positioning groove 205 . The size of the inclined block 303 matches the size of the fixing hole 110 , and the inclined block 303 is clamped with the fixing hole 110 .

[0036] The above scheme is adopted: by setting the cooperation of structures such as the handle 301 and the inclined block 303, it is convenient to replace and load and unload the clamp 107. When fruits of different sizes need to be picked or the clamp 107 is damaged and needs to be replaced, it is only necessary to rotate and remove the bolt 104 first to release the clamping connection between the movable plate 105 and the push rod 102, and then pull the handle 301 away from the fixed seat 201, and drive the inclined block 303 to move to the inside of the slide groove 305 through the connecting rod 302, so that the clamping connection between the inclined block 303 and the fixing hole 110 is released, and the mounting seat 108 is removed from the fixed seat 201 to complete the disassembly, thereby improving the application range and work efficiency of the device.

[0037] like Figure 1 As shown, the main structure 4 includes a vehicle body 401, a robot body 402 is fixedly connected to the top of the vehicle body 401, a camera 403 is rotatably connected to the top of the robot body 402, a collecting groove 404 is arranged on the top of the vehicle body 401, and a limiting groove 405 is opened inside the robot body 402.

[0038] The hydraulic cylinder 101 is fixedly connected to the robot body 402 , the size of the limit block 204 matches the size of the limit slot 405 , the limit block 204 is rotationally connected to the limit slot 405 , and the fixed seat 201 is rotationally connected to the robot body 402 .

[0039] It should be noted that a processing module is provided in the body 401 of the robot, the processing module is electrically connected to the camera 403, and the processing module stores a pre-trained deep learning model, target detection algorithm, maturity assessment algorithm, and path planning algorithm;

[0040] The camera 403 captures images of the target crop area in real time and preprocesses the captured images to make the features of the target crop more prominent; the preprocessing includes operations such as denoising, enhancement, and color space conversion (such as from RGB to HSV).

[0041] The deep learning model performs target detection on the preprocessed images and identifies the target crops in the images (such as citrus, apples, peaches, pears, etc.).

[0042] The deep learning model is a pre-trained CNN model (such as ResNet, VGG, EffectiveNet, etc.). The embodiments of the present application do not limit the specific types of deep learning models. These models can automatically extract useful features from images in image recognition tasks (such as real recognition).

[0043] The object detection algorithm extracts the bounding box of the target crop and determines its position and size in the image.

[0044] The target detection algorithm may be Faster R-CNN, YOLO, etc. The embodiment of the present application does not limit the specific type of the target detection algorithm.

[0045] The maturity assessment algorithm classifies the maturity of the target crop (such as immature, semi-mature, fully mature) based on the crop's color, texture, shape and other characteristics.

[0046] The maturity assessment algorithm includes pre-trained classifiers, such as support vector machines (SVMs), random forests, etc. The embodiments of the present application do not limit the specific types of classifiers.

[0047] The path planning algorithm plans the robot's movement path based on the location and maturity information of the target crop and finds the optimal path from the current position to the target position.

[0048] The path planning algorithm may be an A* algorithm, a Dijkstra algorithm, etc., and the specific type of the path planning algorithm is not limited in the present application embodiment. It may be combined with a navigation system (such as GPS, laser radar, etc.) to realize autonomous navigation and movement of the robot.

[0049] The above solution is adopted: by setting a 360° rotating camera 403, it is convenient to perform an all-round scan of the surrounding environment, so that the image of the crop and related data such as distance, size and maturity can be collected more comprehensively, which is conducive to the deep learning model to recognize the pre-processed image.

[0050] The working principle and use process of the present invention are as follows: first, after the device is started, the camera 403 and the sensor carried thereon start to scan the surrounding environment, collect images of crops and relevant data such as distance, size and maturity, and then pre-process the collected raw data through an image processing algorithm, and analyze the pre-processed images through a deep learning model to extract features related to crop identification and positioning, and then based on the extracted features, use a classifier to identify mature crops, and determine the position of the crops through a target detection algorithm, and calculate the optimal path from the current position of the robot to the target crop position through a path planning algorithm, and in combination with a navigation system, control the vehicle body 401 to reach the target crop position with the minimum moving distance, and according to the path planning result, the robot body 402 drives the gripper 107 to move to the side of the target crop;

[0051] Then, the hydraulic cylinder 101 causes the push rod 102 to move away from the hydraulic cylinder 101, driving the movable plate 105 to move toward the direction close to the crop. Because the shape of the clamping claw 107 is an arc, the movement of the movable plate 105 causes the three clamping claws 107 to move toward each other through the clamping groove 106, clamping the crop in the middle. At this time, the guide rod 103 moves to the side of the linear guide groove 202 away from the hydraulic cylinder 101. When the push rod 102 continues to move, the guide rod 103 squeezes the positioning groove 20 5. The fixing seat 201 is rotated inside the robot body 402 under the constraint of the positioning groove 205, thereby driving the mounting seat 108 fixedly connected thereto to rotate, so that the clamping claws 107 of the crop holding column drive the crop to rotate, so that the fruit is separated from the branches, and then the robot body 402 drives the clamping claws 107 to move to the top of the collecting groove 404, and the push rod 102 slides toward the direction of the hydraulic cylinder 101, so that the three clamping claws 107 open, so that the crop falls into the collecting groove 404 to complete the picking work;

[0052] When it is necessary to pick crops of different sizes or the clamping jaw 107 is damaged and needs to be replaced, it is only necessary to rotate and remove the bolt 104 first to release the clamping connection between the movable plate 105 and the push rod 102, and then pull the handle 301 away from the fixing seat 201, and drive the inclined block 303 to move to the inside of the slide groove 305 through the connecting rod 302, so that the clamping connection between the inclined block 303 and the fixing hole 110 is released, and the mounting seat 108 is removed from the fixing seat 201 to complete the disassembly. The plate 109 is aligned with the positioning groove 205 on the fixing seat 201 and is moved toward the positioning groove 205, so that the positioning plate 109 squeezes the inclined surface of the inclined block 303, and squeezes the inclined block 303 into the inside of the slide groove 305 through the inclined surface, and squeezes the spring 304. When the positioning plate 109 drives the fixing hole 110 to move to the side of the inclined block 303, the inclined block 303 is clamped with the fixing hole 110 under the action of the elastic force of the spring 304. Finally, the movable plate 105 is fixed to the push rod 102 by the bolt 104 to complete the installation and replacement.

[0053] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0054] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent picking robot based on deep learning algorithm and machine vision, comprising a main body (4), characterized in that: A rotating mechanism (2) is arranged inside the main body mechanism (4), a picking mechanism (1) is arranged inside the rotating mechanism (2), and a fixing mechanism (3) is arranged on the rotating mechanism (2); The picking mechanism (1) comprises a hydraulic cylinder (101), a push rod (102) is slidably connected to the hydraulic cylinder (101), guide rods (103) are fixedly connected to both sides of the push rod (102), a bolt (104) is threadedly connected to one end of the push rod (102) away from the hydraulic cylinder (101), a movable plate (105) is rotatably connected to one end of the push rod (102) away from the hydraulic cylinder (101), three slots (106) are evenly arranged on the movable plate (105), the insides of the three slots (106) are slidably connected to clamping claws (107), the ends of the three clamping claws (107) close to the push rod (102) are rotatably connected to a mounting seat (108), a positioning plate (109) is fixedly connected to one end of the mounting seat (108) close to the push rod (102), and fixing holes (110) are respectively arranged on both sides of the positioning plate (109).

2. The intelligent picking robot based on deep learning algorithm machine vision according to claim 1 is characterized in that: The movable plate (105) is located on a side of the bolt (104) close to the push rod (102), and one end of the push rod (102) away from the hydraulic cylinder (101) passes through the surface of the mounting seat (108) and extends to a side of the mounting seat (108) close to the clamp (107).

3. The intelligent picking robot based on deep learning algorithm machine vision according to claim 1 is characterized in that: The rotating mechanism (2) comprises a fixed seat (201), the inner wall of the fixed seat (201) is provided with two linear guide grooves (202), the two linear guide grooves (202) are provided with an arc guide groove (203) on the side away from the hydraulic cylinder (101), a limiting block (204) is fixedly connected to the surface of the fixed seat (201), and a positioning groove (205) is provided on the end of the fixed seat (201) away from the hydraulic cylinder (101).

4. The intelligent picking robot based on deep learning algorithm machine vision according to claim 3 is characterized by: The size of the positioning groove (205) is matched with the size of the positioning plate (109), the positioning groove (205) and the positioning plate (109) are clamped together, and the hydraulic cylinder (101) and the inner wall of the fixing seat (201) are slidably connected.

5. The intelligent picking robot based on deep learning algorithm machine vision according to claim 4 is characterized in that: The two guide rods (103) are respectively slidably connected to the two linear guide grooves (202) and the two arc guide grooves (203), and the fixed seat (201) is located between the hydraulic cylinder (101) and the mounting seat (108).

6. The intelligent picking robot based on deep learning algorithm machine vision according to claim 3 is characterized by: The fixing mechanism (3) comprises two handles (301), one end of the two handles (301) close to each other is fixedly connected to a connecting rod (302), and one end of the connecting rod (302) away from the handles (301) is fixedly connected to an inclined block (303), a sliding groove (305) is provided inside the fixing seat (201), and the inclined block (303) is elastically connected to the inner wall of the sliding groove (305) through a spring (304).

7. The intelligent picking robot based on deep learning algorithm machine vision according to claim 6 is characterized by: The two handles (301) are located on both sides of the fixing seat (201), and the inclined block (303) penetrates the inner wall of the sliding groove (305) and extends to the inside of the positioning groove (205).

8. The intelligent picking robot based on deep learning algorithm machine vision according to claim 7 is characterized by: The size of the inclined block (303) matches the size of the fixing hole (110), and the inclined block (303) is snap-fitted to the fixing hole (110).

9. The intelligent picking robot based on deep learning algorithm machine vision according to claim 6, characterized in that: The main body mechanism (4) comprises a vehicle body (401), the top of the vehicle body (401) is fixedly connected to a robot body (402), the top of the robot body (402) is rotatably connected to a camera (403), the top of the vehicle body (401) is provided with a collecting groove (404), and a limiting groove (405) is provided inside the robot body (402).

10. The intelligent picking robot based on deep learning algorithm and machine vision according to claim 9, characterized in that: The hydraulic cylinder (101) is fixedly connected to the robot body (402), the size of the limit block (204) is matched to the size of the limit slot (405), the limit block (204) is rotationally connected to the limit slot (405), and the fixed seat (201) is rotationally connected to the robot body (402).

Images