Method for detecting and removing vegetables

Abstract

The application provides a vegetable detection and removal method, which extracts spatial structure features of a to-be-detected vegetable through a lens, establishes a corresponding spatial model, matches feature information of various kinds of vegetables with the spatial structure features, identifies the kind of the to-be-detected vegetable, further identifies whether there are problem features such as yellowing, rotting and worm holes on the spatial model and accurately locates them, intelligently judges whether the to-be-detected vegetable is qualified, and accurately removes the area of unqualified vegetables through a sorting mechanism and a removal mechanism. The application extracts spatial features and matches other features through vision to identify vegetables, accurately locates and removes the area of problem features through a spatial model, has a low false detection rate, a high successful removal rate, and achieves a good detection and removal effect.

Description

Technical Field

[0001] This invention relates to the field of vegetable detection and removal, and in particular to a method for vegetable detection and removal. Background Technology

[0002] In recent years, my country's agricultural development model has been shifting from traditional to modern intelligent agriculture. However, the inspection and harvesting of most vegetables still rely on manual labor, hindering the development of intelligent agriculture. The production and demand of agricultural products are also constantly increasing, leading to a large labor force consumption when vegetable production is high. Research on vegetable inspection and harvesting, which uses visual recognition of local features to determine vegetable quality, provides important technical support for solving the aforementioned problems. Summary of the Invention

[0003] The purpose of this invention is to provide a method for detecting and removing vegetables. This machine can quickly and accurately detect and remove vegetable targets, realizing intelligent integration of detection and removal, thereby solving the problem of excessive labor consumption in existing machines.

[0004] A further objective of this invention is to determine the type of vegetables by capturing the spatial structural features of the vegetables through the lens of the original image acquisition component, and to accurately locate the problem area by identifying features such as yellowing and rot by the rating module, thereby selecting unqualified vegetables and moving them to the removal mechanism for precise removal of the problem area.

[0005] According to one aspect of the present invention, a method for detecting vegetables is provided, comprising:

[0006] S1. The vegetables to be tested are transported through a conveyor mechanism, and the original image set of the vegetables to be tested from various angles is obtained through the testing mechanism.

[0007] S2. The original image set is sent to the image feature recognition module to obtain the features of the vegetables to be detected, and the type of vegetables to be detected is determined by feature matching;

[0008] S3. A spatial model is established through the built-in neural network module of the image feature recognition component. The model is used to identify whether there are problem features in the spatial features of the original image set and to determine whether the vegetables are qualified. If they are qualified, they are directly output to the discharge port. If they are not qualified, step S4 is executed. S4. The detection mechanism sends a digital signal to the subsequent sorting mechanism. The mechanical arm is controlled to clamp the unqualified vegetables onto the second conveyor belt and then to the removal mechanism to accurately remove the problem features of the unqualified vegetables.

[0009] S5. Finally, output qualified vegetables to the discharge port.

[0010] In step S1, the detection mechanism and the sorting mechanism are placed sequentially along the conveying direction of the conveying mechanism, with the sorting mechanism placed alongside the removal mechanism.

[0011] In step S1, the vegetable detection mechanism includes an image raw acquisition component and an image feature recognition component that are interconnected.

[0012] The image acquisition component includes a first lens group and a second lens group, which are respectively installed on both sides of the conveying mechanism, with the lenses facing the conveying mechanism, for recording the image data of the vegetable to be detected and establishing an original image set.

[0013] The first lens group and the second lens group each include four lenses arranged vertically and horizontally at an angle of 15 to 20 degrees, used to capture images of the vegetables to be tested from multiple angles.

[0014] The image feature recognition component has a built-in neural network that includes a recognition module and a rating module. The recognition module is used to build a spatial model based on the original image set to match the features of various vegetables in the database. The rating module is used to determine whether there are problematic features in the spatial model to determine whether the vegetables are qualified.

[0015] The rating module is used to verify whether there are problematic features among the structural features matched by the identification module to determine whether the vegetables are qualified. The problematic features include rot, wormholes, and color that is different from the database records.

[0016] In step S4, the sorting mechanism includes a two-claw robotic arm and a five-claw robotic arm.

[0017] The two-claw manipulator includes a power rod, a spring, a left connecting rod, and a right connecting rod. The power rod and the spring are used to control the opening and closing of the two-claw manipulator. The inner side of the top of the left connecting rod and the right connecting rod is equipped with a clamping plate, which is used to firmly clamp the vegetables.

[0018] The five-claw manipulator includes a cylinder, a connecting rod, and an elastic silicone gripper. The cylinder and the connecting rod are used to control the opening and closing of the five-claw manipulator. The inner side of the top of the elastic silicone gripper has serrated patterns to prevent vegetables from slipping.

[0019] Traditional machine vision methods for vegetable inspection require complex processing of raw images, followed by manual feature extraction and image recognition using a feature trainer. Machine vision demands high-quality raw images, necessitates feature analysis for each type of vegetable, and relies heavily on human experience, making the process overly complex. This invention combines visual intelligent feature extraction with spatial image model processing. The raw image acquisition component captures the spatial structural features of vegetables to determine their type. A rating module identifies yellowing, rot, and other characteristics to precisely locate problem areas, selecting substandard vegetables. These substandard vegetables are then moved by a sorting mechanism to a removal mechanism for precise removal, achieving fully automated detection and removal of substandard vegetables. This method offers high accuracy and speed in vegetable detection, demonstrating strong applicability for vegetable inspection and removal tasks and is suitable for widespread adoption. Attached Figure Description

[0020] Figure 1 This is a structural diagram of the vegetable detection and removal method in this invention;

[0021] Figure 2 This is a structural diagram of the two-claw robotic arm of the sorting mechanism in this invention;

[0022] Figure 3 yes Figure 2 Side sectional view;

[0023] Figure 4 This is a structural diagram of the five-claw robotic arm of the sorting mechanism in this invention;

[0024] Figure 5 yes Figure 4 Enlarged view of region A in the middle;

[0025] Figure 6 This is a flowchart of the vegetable detection and removal method in this invention.

[0026] In the diagram: 1-First conveyor belt motor; 2-First conveyor belt; 3-Second conveyor belt; 4-Image raw acquisition component; 5-Image feature recognition component; 6-Two-claw manipulator; 7-Bracket; 8-Second conveyor belt motor; 9-Outer wall; 10-Fixing block; 11-Power rod; 12-Left connecting rod; 13-Right connecting rod; 14-Left clamping plate; 15-Right clamping plate; 16-Spring; 17-Five-claw manipulator; 18-Connecting rod; 19-Air nozzle; 20-Cylinder; 21-Connecting plug; 22-Elastic silicone claw. Detailed Implementation

[0027] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0028] This invention discloses a method for detecting and removing vegetables, comprising the following steps:

[0029] S1. The vegetables to be tested are transported through a conveyor mechanism, and the original image set of the vegetables to be tested from multiple angles is obtained by the testing mechanism;

[0030] S2. The original image set is sent to the image feature recognition module to obtain the features of the vegetables to be detected, and the type of vegetables to be detected is determined by feature matching;

[0031] S3. Establish a spatial model through the built-in neural network module of the image feature recognition component 5, identify whether there are problem features in the spatial features of the original image set, and determine whether the vegetables are qualified. If they are qualified, they are directly output to the discharge port. If they are not qualified, proceed to step S4.

[0032] S4. The testing agency sends a digital signal to the subsequent sorting agency, which controls the robotic arm to clamp the unqualified vegetables onto the second conveyor belt and then to the removal agency to precisely remove the unqualified vegetables based on their problematic characteristics.

[0033] S5. Finally, output qualified vegetables to the discharge port.

[0034] The following is a detailed implementation process of the present invention.

[0035] S1. The vegetable testing mechanism is equipped with a first conveyor belt 2 and a second conveyor belt 3 in parallel. The first conveyor belt 2 and the second conveyor belt 3 are respectively equipped with a first conveyor belt motor 1 and a second conveyor belt motor 8. The vegetables to be tested are conveyed through the first conveyor belt 2 and passed through the image raw acquisition component 4. The first lens group and the second lens group record the image data of the vegetables from various angles on the conveying mechanism and establish the raw image set.

[0036] S2. The feature recognition module built into the image feature recognition component 5 reads the spatial feature data of the vegetable, such as color and length, and establishes a spatial model of the vegetable to be detected. Based on all the read spatial features, it matches the features of various types of vegetables in the database. When the number of matched features reaches a certain value, it confirms that the type is the target to be identified in the image to be identified, thereby confirming the type of the vegetable.

[0037] S3. The rating module of the built-in neural network of the image feature recognition component 5 identifies and locates the problem features in the spatial features of different types of vegetables. For leafy vegetables such as cabbage and scallions, it identifies whether there are rotten or yellowing problems on their main spatial features. For tuberous vegetables such as potatoes, it identifies whether there are rotten or wormhole problems on their main spatial features. The identification results are fed back to the rating module and the problematic vegetables are moved to the appropriate robotic arm. Leafy vegetables are moved to the underside of the five-claw robotic arm, and tuberous vegetables are moved to the underside of the two-claw robotic arm.

[0038] S4. If the problem feature is not identified, the vegetables are directly output to the discharge port. If the problem feature is identified, the rating module will classify the vegetables as unqualified vegetables and convert the spatial location information of the problem feature into a digital signal and send it to the subsequent sorting mechanism. Then, according to the type of vegetable, it will be transported to the designated type of robotic arm area through the first conveyor belt 2.

[0039] S5. Large vegetables such as pumpkins and eggplants are conveyed to the area of ​​the two-claw robotic arm 6. The two-claw robotic arm 6 on the support 7 will control the internal power rod 11 to retract along the direction of the fixed block 10, pulling the upper ends of the left connecting rod 12 and the right connecting rod 13 into the outer wall 9, so that the left clamping plate 14 and the right clamping plate 15 at the end close, firmly clamping the vegetables. Then, the robotic arm extends above the second conveyor belt 3 and controls the power rod to extend, so that the upper ends of the left connecting rod 12 and the right connecting rod 13 are squeezed out of the outer wall 9. The spring 16 will pop open the left connecting rod 12 and the right connecting rod 13, so that the left clamping plate 14 and the right clamping plate 15 open and place the vegetables onto the second conveyor belt. Small vegetables such as potatoes and leafy greens are conveyed into the picking mechanism and transported to the area of ​​the five-claw robotic arm 17. The vacuum valve is opened, causing the vacuum pump to draw a vacuum through the air nozzle 19 on the cylinder 20 connected to the air pipe, controlling the rise of the connecting plug 21. At the same time, the five connecting rods 18 rise and pull the inner side of the five elastic silicone grippers 22, so that the grippers close and grab the vegetables. After the vacuum valve is closed, the robotic arm extends above the second conveyor belt 3. The cylinder 20 is depressurized, causing the connecting plug 21 to fall, which drives the five connecting rods 18 to fall and push the five elastic silicone grippers 22 to open, so that the vegetables fall onto the second conveyor belt 3 and are sent into the picking mechanism.

[0040] S6. The removal agency uses the spatial location information sent by the rating module to accurately remove the problematic characteristics of the substandard vegetables.

[0041] S7. Finally, the qualified vegetables are output to the discharge port.

[0042] In summary, the detection and removal method provided by this invention employs a visual extraction method for local spatial features and performs matching verification using global spatial structure feature information to determine the type of vegetable to be detected. This invention intelligently detects vegetables and performs subsequent removal operations based on global spatial feature information and local problem features, resulting in a low false detection rate, a high successful removal rate, and the ability to intelligently implement the entire process, achieving good detection and removal effects.

[0043] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method of inspecting and removing a vegetable, comprising a conveying mechanism for placing the vegetable, characterized in that, The method comprises the following steps: S1, transporting the vegetables to be detected by a conveying mechanism, and obtaining a set of original images of the vegetables to be detected at each angle by a detection mechanism; S2, delivering the set of original images to an image feature recognition module to obtain the spatial features of the vegetables to be detected, and determining the type of the vegetables to be detected through spatial feature matching; S3, establishing a spatial model through a neural network module built in the image feature recognition component to identify whether there is a problem feature in the spatial features of the set of original images, and determining whether the vegetables are qualified, and if qualified, directly outputting to a discharge port, and if not qualified, executing step S4; leafy vegetables are moved under a five-claw mechanical hand, and blocky vegetables are moved under a two-claw mechanical hand; S4, the detection mechanism sends a digital signal to a subsequent sorting mechanism, and controls the mechanical arm to clamp the unqualified vegetables onto a second conveying belt and then to a removing mechanism to accurately remove the problem features of the unqualified vegetables; According to the type of the vegetables, the first conveying belt delivers the vegetables to a designated type of mechanical arm area; S5, finally outputting the qualified vegetables to the discharge port.

2. The vegetable detection and removal method according to claim 1, wherein In step S1, the detection mechanism, the sorting mechanism, and the removing mechanism are placed side by side along the conveying direction of the conveying mechanism.

3. The method of claim 1, wherein In step S1, the vegetable detection mechanism comprises an image original acquisition component and an image feature recognition component connected to each other.

4. The vegetable detection and removal method according to claim 3, wherein The image original acquisition component comprises a first lens group and a second lens group, and the first lens group and the second lens group are respectively installed on both sides of the conveying mechanism, and all lenses face the conveying mechanism to record image data of the vegetables to be detected and establish a set of original images.

5. The method of claim 4, wherein The first lens group and the second lens group each comprise four lenses placed up and down and left and right to be opened at an angle of 15-20 degrees, and are used to shoot images of the vegetables to be detected at multiple angles.

6. The method of claim 3, wherein The neural network built in the image feature recognition component comprises an identification module and a rating module, the identification module is used to establish a spatial model according to the set of original images to match the features of various types of vegetables in a database, and the rating module is used to determine whether there is a problem feature in the spatial model to determine whether the vegetables are qualified.

7. The method of claim 6, wherein The rating module is used to verify whether there is a problem feature in the spatial features matched by the identification module to determine whether the vegetables are qualified, and the problem features include rot, worm hole, and color different from the database record.

8. The method of claim 1, wherein In step S4, the sorting mechanism comprises a two-claw mechanical hand and a five-claw mechanical hand.

9. The method of claim 8, wherein The two-claw mechanical hand comprises a power rod, a spring, a left connecting rod, and a right connecting rod, the power rod and the spring are used to control the opening and closing of the two-claw mechanical hand, the left connecting rod and the right connecting rod are provided with clamping plates on the inner side of the top end, and the clamping plates are used to stably clamp the vegetables.

10. The method of claim 8, wherein The five-claw mechanical hand comprises a pneumatic cylinder, a connecting rod, and a flexible silicone claw hand, the pneumatic cylinder and the connecting rod are used to control the opening and closing of the five-claw mechanical hand, and the flexible silicone claw hand is provided with sawtooth patterns on the inner side of the top end to prevent the vegetables from falling off.

Images