An automated apparatus and method for obtaining tissue from a seed

Abstract

The application discloses an automatic device and method for obtaining tissues from seeds, comprising: a seed loading module for arranging seeds from a seed bin into single-seed arrangement and conveying the single-seed to a working area of a seed grabbing tool along a conveying line, the seed grabbing tool transfers the seed to a work station switching module; a sampling module comprising the work station switching module, a plurality of seed fixing devices are arranged on the work station switching module, the work station switching module is used for switching the seed fixing devices on a plurality of work stations, the seed is fixed by the seed fixing device and then switched to a vision and cutting work station, a cutting assembly is installed above the cutting work station, the cutting assembly is used for cutting the seed of the seed fixing device under the guidance of a path planning module; a vision module comprising: a phenotype feature recognition module for recognizing the phenotype feature of the seed, a positioning module for positioning the coordinates of the seed, and a path planning module for planning the seed cutting and sampling path and guiding the cutting assembly to cut the seed.

Description

Technical Field

[0001] This invention relates to the field of seed cutting equipment technology, and in particular to an automated device and method for obtaining tissues from seeds. Background Technology

[0002] In the fields of plant breeding and gene analysis, obtaining tissue samples from seeds non-destructively or minimally invasively is a crucial process. Traditional seed sampling methods rely heavily on manual operation, which is inefficient and prone to contamination, making it difficult to meet the demands of high-throughput and precision breeding in modern breeding.

[0003] Therefore, automated seed sampling technology has been gradually developed. For example, existing automated seed sampling systems extract tissue samples from seeds through mechanical grinding, drilling, or cutting. These systems typically include seed delivery modules, positioning modules, and sampling modules, enabling a certain degree of automation. However, these systems still have the following prominent problems: Existing sampling devices use blind grasping at fixed positions when grabbing seeds, relying on the previous seed transport process. Seeds need to be accurately transported to the fixed grasping position; any deviation will lead to grasping failure. Existing sampling devices use a combination of vision and grasping tools, but this is a static grasping method, requiring a pre-process to disperse the seeds. If the seeds are not dispersed, they cannot be grasped, limiting efficiency. Existing sampling uses physical drills or grinding tools, requiring seeds to be placed in a uniform orientation. If the orientation is incorrect, the seeds will be cut incorrectly, potentially cutting off the embryo and causing the seed to lose its germination activity. This also easily damages the seeds, affecting their germination potential and subsequent seed utilization value. Existing sampling using laser cutting relies on the movement of a laser-controlled machine to complete the cutting. This movement speed is slow, the stroke is limited, the working range is limited, and the number of seeds that can be cut at one time is limited, resulting in low efficiency. Existing sampling methods often follow a four-step process of grasping, orientation, fixing, and cutting, which is reasonable but inefficient. Summary of the Invention

[0004] In order to overcome the above-mentioned shortcomings of the prior art, the purpose of the present invention is to provide an automated device and method for obtaining tissues from seeds, which integrates vision and algorithms to realize seed perception and planning, and is equipped with flexible grasping to adapt to complex seed morphology, so as to solve the problems mentioned in the background art.

[0005] The technical solution adopted by this invention to solve its technical problem is: an automated device for obtaining tissue from seeds, comprising: The seed feeding module is used to arrange the seeds from the hopper into single-seed arrangements and transport the single seeds along the conveyor line to the working area of ​​the seed gripping tool. The seed gripping tool then picks up the seeds and transfers them to the workstation switching module. The sampling module includes a station switching module, which is equipped with multiple seed fixing devices. The station switching module is used to switch the seed fixing devices between multiple stations. After the seed is fixed by the seed fixing device, the module switches to the vision and cutting station. A cutting component is installed above the cutting station. The cutting component is used to cut the seed of the seed fixing device under the guidance of the path planning module. The vision module includes a path planning module, a positioning module, and a phenotypic feature recognition module. The phenotypic feature recognition module is used to identify the phenotypic features of the seed, the positioning module is used to locate the coordinates of the seed, and the path planning module is used to plan the seed cutting sampling path and guide the cutting component to cut the seed.

[0006] As a further improvement of the present invention, the seed feeding module includes: The vibratory plate has seed movement path guide grooves on its edge, and the seeds are arranged into a single row and move along the guide grooves by vibration. A conveyor belt, located at the outlet of the vibratory feeder, is used to transport single seeds to the working area of ​​the seed-grabbing tool. A seed-grabbing tool is used to grab seeds based on their coordinate information and movement speed during the dynamic transport of seeds on the conveyor line, and place the grabbed seeds into a seed fixing device.

[0007] As a further improvement of the present invention, the seed-grabbing tool includes: The motion module drives the positioning of the flexible gripper assembly through movement in the X and Y directions; A flexible gripper assembly is mounted on the drive end of a motion module. The gripping part of the flexible gripper assembly has a deformable flexible gripper that can adapt to seeds of different shapes.

[0008] As a further improvement of the present invention, the workstation switching module has a vision workstation, a seed grasping workstation, a sampling workstation and a collection workstation; The seed fixing device includes: Servo-controlled electric grippers are installed at the vision station, seed grasping station, sampling station, and collection station, respectively. An adaptive gripping fixture, mounted on the servo-driven electric gripper, is used to grip seeds of different shapes that are transferred by a flexible gripper assembly.

[0009] As a further improvement of the present invention, the cutting assembly includes a laser machine consisting of a laser machine, a reflector, and a galvanometer. The laser machine is positioned above the sampling station and is used to perform surface treatment on the seed coat of the seeds in the adaptive clamping fixture at the sampling station, and to cut the seeds in the adaptive clamping fixture under the guidance of the path planning module.

[0010] As a further improvement of the present invention, the positioning and recognition module is disposed at the connection end of the conveyor belt and the vibrating plate. The positioning and recognition module has multiple first cameras. The first cameras acquire motion images of the seeds on the conveyor belt, control the seed grasping tool to perform pose calculation and target recognition based on the target seeds extracted from the motion images, and drive the motion module and flexible gripper assembly to grasp the movement trajectory of the target seeds.

[0011] As a further improvement of the present invention, the phenotypic feature recognition module includes a second camera, which acquires a characterization image of a seed in an adaptive clamping fixture at a vision station, locates the seed position based on the characterization image, and performs target feature recognition on the seed to obtain the phenotypic features of the seed.

[0012] As a further improvement of the present invention, the path planning module includes a third camera, which acquires a seed image of a seed in an adaptive clamping fixture at a sampling station, acquires seed positioning and seed characterization information based on the seed image, performs cutting path planning for the target seed based on the seed image, and performs deskinning region planning for the target seed.

[0013] As a further improvement of the present invention, the cutting assembly further includes a collection module, which includes a material tray, comprising a large material tray and a small material tray. The small material tray is used to load tissue samples. The small material tray is transported to the tissue sampling station via a material rack and a conveyor line. After the cutting assembly cuts the seeds into tissue samples, the tissue samples fall into the small material tray for storage via a distribution tray. The large material tray is used to load the cut seeds. The small material tray is transported to the collection station via a material rack and a conveyor line. The cut seeds fall into the large material tray for storage via a distribution tray.

[0014] On one hand, the present invention provides an automated method for obtaining tissue from seeds, comprising the following steps: Seed localization: Acquire motion images of seeds on the conveyor belt, perform pose calculation and target recognition on the target seeds, and obtain the placement direction and embryo position of the target seeds; Seed grasping: Drive the motion module and flexible gripper assembly to grasp the movement trajectory of the target seed; Seed identification: Acquire representational images at the visual workstation, use a pre-trained embryo feature recognition model on the representational images to detect the embryo information of valid target seeds in the image information, and screen seeds; Seed planning: Acquire seed images at the sampling station, plan the cutting path for the target seed based on the seed images, and plan the skin removal region for the target seed; Seed cutting: A laser machine is used to cut and remove the outer skin of the seeds under the guidance of the path planning module.

[0015] As a further improvement of the present invention, the seed positioning step specifically includes: The first camera is used to acquire the original image of the conveyor belt. Preprocess the original image; Extract moving objects from the processed image; The extracted target seeds are used for pose calculation and target recognition. Based on the target recognition results, the target's trajectory is predicted by combining the seed's movement speed to achieve dynamic target capture.

[0016] As a further improvement of the present invention, target extraction includes: performing binarization processing on pixels, distinguishing pixels below a threshold from pixels above a threshold, and forming connected components based on the parts of similar gray values ​​between adjacent pixels; Calculate the minimum bounding rectangle for the connected components, and use the calculated minimum bounding rectangle as the target box; The length and width of the target bounding box are calculated, and the target seed is determined based on the target length and width standards.

[0017] As a further improvement of the present invention, the pose calculation includes: calculating the orientation angle of the target box as the placement direction of the target object based on the extracted target box, and calculating the center of the target box as the target center coordinates; Target identification includes: identifying target seed embryos using a pre-trained embryo feature recognition model.

[0018] As a further improvement of the present invention, the seed cutting path planning step specifically includes: Use a third camera to acquire the raw images; Preprocess the original image; Target extraction is performed on the processed image; Perform path planning on the extracted target; The extracted target area is then used for deskinning and region planning.

[0019] As a further improvement of the present invention, the cutting path planning includes: determining the planning base point for the extracted target, drawing a circular curve with the base point of the path planning as the center and the radius increasing from zero to the target box, and confirming the planned path for the target object.

[0020] As a further improvement of the present invention, the deskinning region planning includes: determining the radius of a circular curve for the extracted target, performing morphological processing on the extracted target, drawing a circular curve for the target based on the morphologically processed target image, and determining the deskinning region.

[0021] As a further improvement of the present invention, the seed identification step specifically includes: The original image acquired by the second camera is labeled with feature regions to generate a labeled reference image. The labeled image is then fed into a deep learning recognition model to train its ability to recognize embryonic features. Image information is acquired using a deep learning recognition model, and valid targets are detected within the image information. Valid targets must meet the following criteria: The target category name is consistent with the feature region category name labeled during the training phase; The target recognition confidence level is greater than or equal to the minimum confidence level; The target recognition overlap rate is less than or equal to the maximum overlap rate; The final result of whether or not a seed embryo exists is obtained based on the valid conditions of the detected target.

[0022] As a further improvement of the present invention, it also includes a material drop recognition module. After the cutting component cuts the seed, when the tissue sample falls into the small material tray, the material drop recognition module identifies whether there is a sample in each compartment of the small material tray. If no sample is identified, it indicates that the cutting component has not cut the seed.

[0023] On the other hand, the present invention also provides an automated method for obtaining tissues from seeds, the method comprising: Picking a single seed from a group of seeds: picking up the single seed using the seed picking tool; Placement and fixing: Move the single seed picked up by the seed grasping tool to the adaptive clamping fixture of the seed fixing device, and place the seed into the adaptive clamping fixture; Cutting and sampling: The cutting component cuts the seed in the adaptive clamping fixture to obtain a tissue sample.

[0024] As a further improvement of the present invention, a vision module is used to plan the cutting path for the seed. After the cutting path planning confirms the planned path of the target object, the beam path of the cutting component is controlled to complete the precise cutting of at least one seed in the seed fixing device.

[0025] As a further improvement of the present invention, the cutting component can perform surface treatment on a surface planned in the epidermal region; the vision module calibrates the seed fixing device, and the cutting component can perform surface degreasing treatment on the calibrated seed fixing device.

[0026] As a further improvement of the present invention, a collection step is also included: collecting the tissue cut from the individually selected seeds into a small material tray, while placing the processed seed with the tissue removed into a large material tray.

[0027] As a further improvement of the present invention, the method also includes genetic and physicochemical analysis of the tissue sample. The genetic information includes genomic DNA sequence, SSR genetic markers, SNP genetic markers, transgenic status, alleles, and methylation patterns. The physicochemical information includes oil content, protein content, starch content, sugar content, heavy metal content, and trace element content.

[0028] As a further improvement of the present invention, it also includes selecting and retaining or removing seeds corresponding to the tissue sample based on the analysis results of the genetic information and physicochemical information. The analysis results are used to determine the agronomic traits of the seeds, including disease resistance, high yield, lodging resistance, insect resistance, plant height, quality, high oil content, high protein content, high sugar content, aroma, taste, and combinations of multiple traits.

[0029] Compared with the prior art, the beneficial effects of the present invention are: This invention achieves real-time pose calculation and trajectory prediction of moving seeds through multiple cameras and dynamic algorithms. A flexible gripper assembly grasps seeds on the conveyor line and transfers them to a motion-adaptive clamping fixture, ensuring the success rate and efficiency of dynamic grasping. A workstation switching module integrates visual recognition, seed grasping, tissue sampling, and sample collection processes, enabling parallel operation and assembly line production, eliminating waiting time between processes, and thus significantly improving overall operational efficiency. A deep learning model accurately identifies key features such as seed embryos, achieving precise identification of seed embryos and endosperm, ensuring that specified tissues are cut according to requirements. This invention performs real-time cutting path planning and epidermal removal area planning on seed images, accurately processing target tissues. By using laser cutting components for non-contact cutting, it replaces traditional mechanical drills or grinding wheels, avoiding damage to the seed embryo caused by physical compression and friction. The laser beam can precisely control the cutting range, removing only the target tissue or performing seed coat surface treatment, thereby greatly improving the germination rate and viability of the sampled seeds. There is no need to move the laser machine; cutting of specific tissues is achieved by controlling the movement of the beam. This invention only requires three steps—grabbing, fixing, and cutting—to complete the cutting and sampling process, simplifying the cutting process. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the overall structure of the present invention; Figure 2 This is a schematic diagram of the structure of the conveyor belt and the positioning and identification module of the present invention. Figure 3 This is a schematic diagram of the structure of the seed-grabbing tool of the present invention; Figure 4 This is a schematic diagram of the workstation switching module of the present invention; Figure 5 This is a schematic diagram showing the position of the visual module of the present invention; Figure 6 This is a schematic diagram of the material tray module of the present invention; Reference numerals: 1. Seed feeding module; 11. Vibratory feeder; 12. Conveyor belt; 13. Seed gripping tool; 131. Motion module; 132. Flexible gripper assembly; 133. Flexible gripper; 2. Station switching module; 21. Vision station; 22. Seed gripping station; 23. Sampling station; 24. Collection station; 31. Positioning recognition module; 32. Phenotypic feature recognition module; 4. Cutting assembly; 5. Seed fixing device; 51. Servo electric gripper; 52. Adaptive clamping fixture; 71. Large material tray; 72. Small material tray; 8. Conveyor line; 9. Distributor tray. Detailed Implementation

[0031] In order to clearly and completely understand the technical solution, the present invention will be further described in conjunction with the embodiments and accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0033] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0034] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0035] Embodiments of the present invention provide an automated apparatus for obtaining tissue from seeds, comprising: The seed feeding module 1 is used to arrange the seeds from the hopper into single-seed arrangements and transport the single seeds along the conveyor line to the working area of ​​the seed gripping tool 13, which then grips and transfers the seeds into the seed fixing device 5. Specifically, the seed feeding module 1 includes: a vibratory feeder 11, a conveyor belt 12, and a seed gripping tool 13; the vibratory feeder 11 has seed movement path guide grooves on its edge, and the seeds are arranged into a single column along the guide grooves by vibration; using vibration and the guide grooves on its edge, the randomly piled seeds in the hopper are automatically arranged into a single column with uniform spacing, realizing the initial conversion from batch processing to single-seed processing; the conveyor belt 12 is set at the outlet of the vibratory feeder 11, receiving the output of the vibratory feeder 11, and serving as a buffer and transmission channel to orderly transport the single-seeds to the designated working area; then, in conjunction with the positioning and recognition module 31, the seed position and posture are identified, and the seeds are processed according to the real-time coordinates provided by the positioning and recognition module 31. Based on information and movement speed, the seed grasping tool 13 dynamically transports seeds along the conveyor line. Instead of waiting at a fixed point, the tool grasps seeds based on their coordinates and movement speed, dynamically adjusting its trajectory to achieve synchronous tracking and grasping of the moving seeds. The grasped seeds are then placed within the seed fixing device 5. The coordinated operation of the vibrating plate 11 and the conveyor belt 12 effectively constrains the position and orientation range of the seeds upon entering the grasping area. The seed grasping tool 13 dynamically grasps the seeds based on their position and orientation, transferring them to the seed fixing device 5. This eliminates the need to search for targets over a large area, improving recognition speed and success rate. After a seed is grasped and placed into a seed fixing device 5, the seed fixing device 5 becomes the seed's dedicated carrier. Through the workstation switching module 2, the entire seed fixing device 5 and the seed are moved together to various workstations. This means that the seed's position and orientation relative to its seed fixing device 5 remain unchanged after being fixed once. Subsequent visual recognition and laser cutting are based on the fixed relationship between the seed and the seed fixing device 5, avoiding cumulative errors caused by multiple grasping and placement, and preventing multiple grasping and releasing of the same seed, thus reducing the risk of seed damage.

[0036] The workstation switching module 2 is equipped with multiple seed fixing devices 5. The workstation switching module 2 is used to switch the seed fixing devices 5 between multiple workstations. The workstation switching module 2 adjusts the traditional single-workstation sequential operation mode to a multi-workstation parallel operation mode, eliminating the waiting time in the production process, so that various processes such as feeding, identification, cutting and collection can be carried out simultaneously, thereby improving the overall output speed of the equipment.

[0037] The vision module includes a path planning module, a positioning and recognition module 31, and a phenotypic feature recognition module 32; The positioning and recognition module 31 is used to locate the coordinates of the seed. The positioning and recognition module 31 not only provides the spatial coordinates of the seed, but also calculates the speed of the seed through continuous frame image analysis or in combination with the encoder. The control center predicts the future trajectory of the seed based on the coordinate and speed information, and instructs the seed grabbing tool 13 to track it synchronously, so as to complete the accurate grabbing during the continuous movement of the seed. The phenotypic feature recognition module 32 is used to identify the phenotypic features of seeds, such as defects like seed integrity, abnormal color, mold, shriveling, and poor development. This avoids wasting control center resources and time on inferior seeds that cannot provide effective samples, thus improving the overall quality of the final tissue samples. The phenotypic feature recognition module 32 enables the control center to perform adaptive processing based on individual phenotypic differences of seeds, identifying the size, shape, and relative position of seeds within the seed. It can also dynamically adjust the optimal laser cutting path for each seed, ensuring that the cutting operation can completely obtain the target tissue sample while guaranteeing the accuracy of sampling.

[0038] The path planning module is used to plan the seed cutting and sampling path and guide the cutting component 4 to cut the seeds. The path planning module dynamically adjusts the optimal laser cutting path for each seed according to the seed characteristics to ensure that the cutting operation can obtain tissue samples completely and ensure the accuracy of sampling.

[0039] The sampling module includes a cutting component 4, which is used to cut the seeds of the seed fixing device 5 under the guidance of the path planning module. The sampling module receives instructions from the path planning module and drives the cutting component 4 (such as a laser machine) to perform precise cutting actions. A non-contact laser cutting component 4 is used, and the operation is performed under visual guidance. The laser beam itself has no mass and will not compress the seeds, greatly reducing the risk of damage to the embryo and maximizing the protection of the seed's germination potential and viability. Traditional drills can only perform simple circular or linear movements, while the cutting component 4 of this application executes any complex curved path generated by the path planning module, adapting to the unique shape and target position of each seed.

[0040] In one embodiment of the present invention, the seed-grabbing tool 13 includes: The motion module 131 drives the positioning of the flexible gripper assembly 132 through movement in the X and Y directions; The flexible gripper assembly 132 is mounted on the drive end of the motion module 131. The gripping part of the flexible gripper assembly 132 has a deformable flexible gripper 133, which can adapt to seeds of different shapes.

[0041] In this embodiment, firstly, the irregular shape of the seeds makes it difficult for traditional gripping mechanisms to adapt to seeds of different shapes and sizes, easily causing seed damage during gripping or fixing. Traditional rigid grippers make point or line contact, resulting in enormous pressure that can easily crush or scratch the fragile seed coat. The flexible gripper 133 utilizes its deformable properties to adaptively deform upon contact with the seed, thereby increasing the contact area and evenly distributing the gripping force. Seeds of various shapes, sizes, and hardness can be safely and stably gripped, reducing the risk of damage during operation. The motion module 131 quickly moves the gripper above the target seed, and the flexible gripper 133, using its flexibility and adaptive capabilities, successfully grips the seed. The flexible gripper assembly 132 is pneumatically controlled, and its gripping part can deform to adapt to seeds of different shapes, ensuring that the seeds are not damaged by excessive force during the gripping process; or the flexible gripper assembly 132 adopts an adaptive electric gripper, which can adapt to seeds of different sizes by setting a constant gripping force; the seed gripping tool 13 is communicatively connected to the vision module, and the vision module sends the seed's movement speed, seed posture and coordinate information to the control center, and the control center sends a gripping command to the seed gripping tool 13 according to the seed's movement speed, seed posture and coordinate information.

[0042] In one embodiment of the present invention, the workstation switching module 2 has a vision workstation 22, a seed grasping workstation 21, a sampling workstation 23 and a collection workstation 24; The vision station 22, seed grasping station 21, sampling station 23 and collection station 24 are switched by rotating a turntable; The seed fixing device 5 includes: Servo electric grippers 51 are assembled on vision station 22, seed grasping station 21, sampling station 23 and collection station 24 respectively; An adaptive clamping fixture 52 is mounted on the drive end of the servo electric gripper 51. The adaptive clamping fixture 52 is used to clamp seeds of different shapes transferred by the flexible gripper assembly 132.

[0043] Vision station 22 is dedicated to performing high-definition imaging and feature recognition on fixed seeds; seed grasping station 21 is dedicated to receiving seeds from flexible grippers 133; sampling station 23 is dedicated to performing time-consuming laser cutting; and collection station 24 is dedicated to unloading the sampled seeds.

[0044] In this embodiment, after the seed is grasped by the flexible gripper assembly 132, it is placed into an adaptive clamping fixture. The adaptive clamping fixture is controlled by a servo electric gripper 51 and is moved to various workstations via the workstation switching module 2. The position and orientation of the seed relative to its seat do not change after the initial placement. The coordinate reference of the cutting area calibrated at the vision workstation 22 can be transmitted to the cutting workstation without deviation, eliminating the cumulative error caused by multiple grasping and placing, and providing a spatial reference for laser cutting. The servo electric gripper 51 is a force-controlled electric gripper. The force-controlled electric gripper can rotate at any angle. The force-controlled electric gripper means that the force used to fix the seed can be set to stabilize the sampling position of the seed during the cutting process, and the force is controllable.

[0045] In one embodiment of the present invention, the cutting component 4 includes a laser machine, which is disposed above the sampling station 23. The laser machine is used to perform surface treatment on the seed coat of the seeds in the adaptive clamping fixture 52 on the sampling station 23, and to identify and cut the seeds in the adaptive clamping fixture 52 under the guidance of the path planning module.

[0046] In this embodiment, the cutting component 4 includes a laser machine, a reflector, and a galvanometer. The galvanometer can be a 2D galvanometer, a 2.5D galvanometer, or a 3D galvanometer. The laser cutting speed is fast and controlled by the galvanometer, resulting in extremely fast beam movement. The laser beam has no mass and acts on the seed in the form of light energy, without direct contact with the seed, thus maximizing the protection of the seed embryo and greatly improving the survival rate and germination potential of the seeds after sampling. Furthermore, the laser system includes a laser machine control cabinet, the laser machine, and a laser machine adjustment mechanism. The same laser system can perform two different precision operations by adjusting the power, speed, and scanning mode, using a lower-power laser to quickly scan the seed coat. The laser can remove wax, lint, or minor contamination from the surface, achieving cleaning, disinfection, or alteration of wettability to prepare for subsequent high-quality cutting. Using a high-power laser, it precisely cuts along the planned path to obtain samples. When cutting seeds, the laser machine is stationary, requiring no movement or adjustment of the seed's orientation. It simply cuts and removes the outer skin of the seed using the path planned by the path planning module. The laser machine can quickly respond to and execute the trajectory generated by the path planning module, performing a customized cutting plan for each seed with a different shape, and can extract tissue with regular shapes and neat edges.

[0047] In one embodiment of the present invention, the positioning and recognition module 31 is disposed at the connection end between the conveyor belt 12 and the vibrating plate 11. The positioning and recognition module 31 has a plurality of first cameras. The first cameras acquire motion images of the seeds on the conveyor belt 12, control the seed grasping tool 13 to perform pose calculation and target recognition based on the target seeds extracted from the motion images, and drive the motion module 131 and the flexible gripper assembly 132 to adjust the grasping and placement pose to grasp the movement trajectory of the target seeds. The positioning and recognition module 31 is set at the connection end between the conveyor belt 12 and the vibrating plate 11, so that the seed is immediately locked as soon as it enters the conveyor belt 12. The control center has enough time to process the image, calculate the pose, predict the trajectory, and plan the motion path of the gripper arm. Multiple first cameras (such as the top, left rear 45°, and right side) are used to form a three-dimensional visual sensor network, which can observe the seed from different angles. A single top-view camera can only obtain two-dimensional planar coordinates and cannot know the three-dimensional posture of the seed such as tilting or rolling, nor can it judge the protrusion or depression. Through the fusion analysis of multi-view images, the three-dimensional coordinates and three-dimensional posture of the seed can be calculated, providing complete spatial information for the seed gripping tool 13, ensuring that the flexible gripper assembly 132 approaches the seed at the best angle. By analyzing the continuous frame images, the control center can calculate the movement speed of the seed. Combined with the existing motion model or algorithm, the future movement trajectory of the seed can be predicted. The control center controls the motion module 131 and the flexible gripper 133 to move to an interception point in advance instead of going to the current position of the seed, so as to grasp the moving seed.

[0048] The phenotypic feature recognition module 32 includes a second camera, which acquires a characterization image of the seed in the adaptive clamping fixture on the vision station 22, locates the seed position based on the characterization image, and performs target object feature recognition on the characterization image to obtain the phenotypic features of the seed. After the seeds are fixed to the seed fixing device 5, before laser cutting, the seeds are screened to ensure that only qualified seeds are cut. At the vision station 22, the seeds are stably fixed by the adaptive clamping fixture. The second camera acquires the characteristic image of the seeds. The control center can identify the seed's color uniformity, whether there are lesions, mold, wrinkles, insect infestation or mechanical damage on the surface, and other phenotypic features. In addition, by analyzing the characteristic image, the phenotypic feature recognition module 32 can evaluate the size, shape and position of the target area. This information can be transmitted to the path planning system to dynamically adjust the laser cutting range.

[0049] The path planning module includes a third camera, which may be mounted within the laser system. The third camera acquires seed images of seeds within the adaptive clamping fixture at sampling station 23, obtains seed positioning and seed characterization information based on the seed images, performs cutting path planning for the target seed based on the seed images, and performs deskin region planning for the target seed. Alternatively, the third camera may function in conjunction with the second camera, integrating positioning recognition, path planning, and phenotypic feature recognition into the second camera. After the seed undergoes transfer, fixation, and station switching, its final posture in the adaptive clamping fixture may deviate slightly from that at vision station 22. The third camera performs a final image at the cutting station, and path planning based on this image is the most direct, ensuring that the cutting path matches the seed's current real-time state, minimizing positioning and processing errors. An area is planned, and a shallow, rapid scan is performed using a laser to remove the seed coat without damaging the interior. Within the exposed area, a cutting path is planned to obtain tissue samples.

[0050] In one embodiment of the present invention, the cutting assembly further includes a collection module, which includes a material tray, comprising a large material tray 71 and a small material tray 72. The small material tray 72 is used to load tissue samples. The small material tray 72 is transported to the sampling station 23 via a material rack and a conveyor line 8. After the cutting assembly 4 cuts the seeds to the target tissue, the tissue samples fall into the small material tray 72 via a distribution tray 9 for storage. The large material tray 71 is used to load the cut seeds. The small material tray 72 is transported to the collection station 24 via a material rack and a conveyor line 8. The cut seeds fall into the large material tray 71 via the distribution tray 9 for storage.

[0051] In one embodiment of the present invention, a material drop recognition module is also included. After the cutting component 4 cuts the seed, when the tissue sample falls into the small material tray 72, the material drop recognition module identifies whether there is a tissue sample in each compartment of the small material tray 72. If no sample is identified, it indicates that the cutting component 4 has not cut the seed into tissue.

[0052] On the other hand, the present invention also provides an automated method for obtaining tissues from seeds, comprising the following steps: S1. Seed positioning: Acquire motion images of seeds on conveyor belt 12, perform pose calculation and target recognition on the target seeds, and obtain the placement direction and embryo position of the target seeds; S2, Seed Grasping: Drive the motion module 131 and the flexible gripper assembly 132 to grasp the movement trajectory of the target seed; S3, Seed Identification: Acquire the representational image on the visual station 22, use a pre-trained embryo feature recognition model on the representational image to detect the embryo information of valid target seeds in the image information, and screen the seeds. S4. Seed planning: Obtain the seed image on sampling station 23, plan the cutting path for the target seed based on the seed image, and plan the skin removal region for the target seed. S5. Seed cutting: The seeds are cut and peeled using a laser machine guided by the path planning module.

[0053] In one embodiment of the present invention, the seed positioning step specifically includes: S11. Use the first camera to acquire the original images directly above, 45 degrees to the left rear, and directly to the right of the input end of the conveyor belt 12; S12. Preprocess the original image; Algorithms such as Gaussian filtering and median filtering are used to eliminate random noise caused by camera sensor, lighting fluctuations, or electronic interference, preventing these noise points from being misjudged as seed features in subsequent steps. Histogram equalization and other methods are used to stretch the dynamic range of image grayscale, making the edges and textures of the seeds more distinct from the background. This is particularly beneficial for capturing details in low-light conditions, suppressing irrelevant interference information in the image, and making the outline and features of the seeds clearer. Image correction is performed to compensate for overly bright or dark areas in the image, making the lighting of the entire image uniform. In fixed scenes, by subtracting the background image, the moving seeds can be greatly highlighted, making their outlines clear, which is convenient for the positioning and recognition module 31 to extract targets. It also eliminates or reduces the adverse effects of uneven ambient lighting, shadows, reflections, etc., ensuring that the vision system can obtain consistent recognition results at different times and in different environments. S13. Extract moving objects from the processed image; S131. Extract regions from the processed image: perform binarization on the pixels, distinguish pixels below the threshold from pixels above the threshold, and form connected components based on the similar gray values ​​between adjacent pixels. By setting a threshold, the image is simplified to a pure foreground (target, white) and background (black), discarding redundant information such as color and texture, and retaining only the most essential shape and position information; S132. Extract the target bounding box of the target region: calculate the minimum bounding rectangle of the connected component and use the calculated minimum bounding rectangle as the target bounding box; By identifying interconnected foreground pixels as the same object, the problem of distinguishing multiple seed or noise points from each other is solved, ensuring that the system can identify each white area in the image as an independent candidate target. S133. Determine the target: Calculate the length and width of the target box, and determine the target seed based on the target length and width standards; This rectangle provides four key parameters: center point coordinates (X, Y), orientation angle (θ), length, and width. The center point coordinates (X, Y) directly tell the seed grabbing tool the precise location of the target in two-dimensional space. It provides a visual size estimate of the seed. By using the known seed size range, it filters connected components, which can effectively eliminate interference boxes caused by image noise, foreign objects, or seed fragments, ensuring that only real, appropriately sized seeds are used as grabbing targets. S14. Perform pose calculation and target recognition on the extracted target seed; S141. Perform preliminary pose calculation on the extracted target: Based on the extracted target bounding box, calculate the direction angle of the target bounding box as the placement direction of the target object, and calculate the center of the target bounding box as the target center coordinates. The target center coordinates (X, Y) provide the position of the seed on the two-dimensional plane, which directly determines the target point that the motion module 131 needs to move the gripper to. The orientation angle (θ) provides the orientation of the seed, which determines the stability of the gripper assembly 132 after reaching the target point. In the dynamic grasping scenario, the control center can fit the seed's movement speed and future trajectory by calculating a series of center coordinates and orientation angles from consecutive frame images.

[0054] S142. Feature recognition of the target object: Use a pre-trained embryo feature recognition model to identify the target seed embryo; By combining PLC prediction of the target's movement trajectory, dynamic target capture is achieved; specifically: During the training phase, the image processed in step S2 is labeled with feature regions (embryos), and the model is trained to recognize the features of the target object through a convolutional neural network. In the application phase, the image processed in step S2 is fed into the trained recognition model for target feature recognition. By training a convolutional neural network (CNN) to learn a large number of manually labeled embryo features, the CNN model learns the visual features of the embryo (such as shape, texture, and relative position to surrounding tissues) using the labeled data, enabling it to acquire knowledge of embryo recognition. In the application stage, the system achieves accurate recognition capability. The preprocessed image is input into the trained model, and the CNN model identifies the position of the embryo and the seed pose data obtained through traditional algorithms, combining the seed pose and the embryo position.

[0055] In one embodiment of the present invention, the seed planning step specifically includes: S41. Use a third camera to acquire the raw image; S42. Preprocess the original image; S43. Extract the target from the processed image; S44. Perform cutting path planning on the extracted target; S441. Determine the planning base point for the extracted target. Based on the extracted target box, take the upper right corner of the target box as the base point for path planning. S442. Draw a circular curve for the target object, using the baseline of the path planning as the center and the radius increasing with the length from zero to the target box. S443. Confirm the planned path for the target object. After traversing and drawing circular curves, perform pixel statistics on the pixel values ​​of the intersection area between the circular area and the target area. Calculate the area of ​​the intersection area based on the principle that the gray value of the target pixel is 255 and the background pixel is 0 in the binary image. Use the planned cutting area as a threshold to confirm the circular curve with the closest area of ​​the intersection area as the lower curve of the cutting area. Perform edge extraction processing on the intersection area to confirm the extracted edge curve as the upper curve of the cutting area. The lower curve and the upper curve of the cutting area are combined to form a complete cutting planning path. S45. Perform skin removal region planning on the extracted target; S451. Determine the planning base point for the extracted target. Based on the extracted target box, take the upper right corner of the target box as the base point for path planning. S452. Determine the radius of the circular curve for the extracted target, and obtain the radius of the circular curve in S443 that is closest to the area of ​​the intersecting region with the planned cutting area as the threshold. Set the radius of the circular curve of the peeling region to be slightly smaller than this radius. S453. Perform morphological processing on the extracted target. To meet the requirement that the contour of the skin removal area should be slightly smaller than the contour of the cutting path, the image erosion effect needs to be achieved. Traverse the image pixels of the target binary image, and set the shape of the structural element (usually a rectangle) with the pixel as the origin and the origin as the center. If all the pixels in the structural element shape are white, the origin pixel is kept white; otherwise, the origin pixel is set to black. S454. Draw a circular curve on the target object based on the morphologically processed target image. Draw a circular curve on the morphologically processed target image based on the confirmed base point center and radius. S455. Determine the deskinning region. Set the intersection region between the morphologically processed target binary image and the drawn circle as the deskinning region.

[0056] In one embodiment of the present invention, the seed identification step specifically includes: The original image acquired by the second camera is labeled with feature regions to generate a labeled reference image. The labeled image is then fed into a deep learning recognition model to train its ability to recognize features. During the training phase, the original image is labeled with feature regions to generate a labeled reference image. The labeled image is then fed into a deep learning recognition model to train its ability to recognize features in the reference image. In the application phase, image information is acquired based on a deep learning recognition model, and valid targets are detected within the image information. Valid targets must meet the following criteria: 1. The target category name is consistent with the feature region category name labeled during the training phase; 2. The target recognition confidence level is greater than or equal to the minimum confidence level; 3. The target recognition overlap rate is less than or equal to the maximum overlap rate; The final result of whether or not a seed embryo exists is obtained based on the valid conditions of the detected target.

[0057] Through deep learning models, the system gains the ability to identify the key feature of "embryo," thereby achieving pre-processing quality control. During training, manual annotation teaches the model what an "embryo" is, enabling it to learn the visual representation of an embryo on various seeds. In application, the system can automatically and quickly determine whether a seed contains a complete and healthy embryo, ensuring that only seeds with life potential and research value can proceed to the subsequent expensive and time-consuming laser cutting process, thus improving the overall value of the final sample from the source. Target category name consistency: objects detected by the model must be classified as "embryos," ensuring the system focuses only on embryos and does not mistake other parts of the seed (such as seed coat damage, shadows) or image noise for embryos, guaranteeing the specificity of the screening. Confidence level ≥ minimum confidence level: fuzzy results that the model "looks like but is not quite sure" are filtered out, improving the certainty of judging qualified seeds. Only seeds that the model is certain have an embryo will pass, ensuring the accuracy of the screening. Overlap rate ≤ maximum overlap rate: for the same target, only the detection box with the highest confidence is retained, suppressing duplicate detection and ensuring that a seed is judged only once, providing the most accurate embryo location. This provides a clear and unique coordinate basis for subsequent cutting path planning that may require avoidance based on embryo position, ensuring data cleanliness.

[0058] In one embodiment of the present invention, this embodiment also provides a target recognition and localization algorithm based on machine vision. Taking the linkage processing of the vision processor at 45 degrees directly above, directly to the right, and to the left rear as an example, the method includes the following steps: 101. The vision processors at 45-degree angles from directly above, directly to the right, and to the left rear acquire the original images of the target material; 102. Convert the original target image captured by the visual processor directly above into a grayscale image; 103. Perform BLOB analysis on the processed grayscale image directly above to obtain preliminary target pose data; Specifically, the BLOB analysis in step 103 includes the following steps: image binarization, connected component analysis, feature calculation, and target selection.

[0059] 1031. Image binarization: By setting a grayscale value threshold, pixels below the threshold in a grayscale image are distinguished from pixels above the threshold. The target area is determined by polarity selection (brighter than the background or darker than the background) and converted into a binary image with only black (0) and white (255) pixel values, thereby separating the target from the background.

[0060] 1032. Connected Component Analysis: Identify and mark the regions formed by all interconnected white (or black) dots in a binary image. Connected components are formed based on the similar gray values ​​between adjacent pixels. Each independent connected component is a BLOB. This example uses the 8-connectivity rule for connected component analysis (a pixel and its eight adjacent pixels above, below, left, right, upper left, upper right, lower left, and lower right are considered connected. 8-connectivity usually yields a more "complete" region).

[0061] 1033. Feature Calculation: Calculate the coordinates of the top left, top right, bottom left, and bottom right vertices of the rectangle based on the minimum bounding rectangle of the BLOB obtained from the connected domain analysis, and output the rectangle features (center point, width, height, angle, and area).

[0062] 1034. Target Filtering: This example filters out target BLOBs based on area, width, and height to eliminate actual interference items.

[0063] 104. Convert the original target images captured by the vision processor at a 45-degree angle from the right and left rear to grayscale images; 105. Continuously capture images of different targets in various poses as the original dataset for training deep learning object detection models; 106. Label the target feature parts and class names in the original dataset to form a complete training dataset; 107. Input the labeled training dataset to train the deep learning object detection model for the task; 108. Call the trained deep learning object detection model to detect feature parts in the target images captured by the visual processors on the right and left rear sides; 109. Based on different detection results, the targets on conveyor belt 12 are divided into graspable and non-graspable items (targets whose long axis is parallel to the direction of movement of conveyor belt 12 are graspable). In this example, the structural design of conveyor belt 12 satisfies that the long axis of the target is parallel to the direction of movement of conveyor belt 12 under normal circumstances. To eliminate interference from targets with high roundness, the graspable items are screened by combining images captured by the vision processor at 45 degrees to the right and left rear. The specific situation is as follows: 1091. If a feature area is detected directly to the right, the target is a grabbable item; 1092. If no feature is detected to the right but a feature is detected at 45 degrees to the left rear, obtain the coordinates (x, y) of the center point of the model detection result box and the coordinates (x0, y0) of the image center. Determine the relationship between x and x0. When x > x0, the target is an ungraspable item; otherwise, it is a grabable item. 1093. If no feature is detected at 45 degrees to the right and left rear, it means the feature is facing directly forward, and the target is not a grabbable item.

[0064] On the other hand, the present invention also provides an automated method for obtaining tissues from seeds, the method comprising: Picking a single seed from a group of seeds: picking up the single seed using the seed picking tool; Placement and fixing: Move the single seed picked up by the seed grasping tool to the adaptive clamping fixture of the seed fixing device, and place the seed into the adaptive clamping fixture; Cutting and sampling: The cutting component cuts the seed in the adaptive clamping fixture to obtain a tissue sample.

[0065] In one embodiment of the present invention, a vision module is used to plan the cutting path for the seed. After the cutting path planning confirms the planned path for the target object, the beam path of the cutting component is controlled to complete the precise cutting of at least one seed in the seed fixing device.

[0066] In one embodiment of the present invention, the cutting component can perform surface treatment on a surface planned in the epidermal region; the vision module calibrates the seed fixing device, and the cutting component can perform surface degreasing treatment on the calibrated seed fixing device.

[0067] In one embodiment of the invention, a collection step is also included: collecting tissue samples cut from the individually selected seeds into a small tray, while placing the processed seed with the tissue removed into a large tray.

[0068] In one embodiment of the present invention, the method further includes performing genetic and physicochemical information analysis on the tissue sample. The genetic information includes genomic DNA sequence, SSR genetic markers, SNP genetic markers, transgenic status, alleles, and methylation patterns. The physicochemical information includes oil content, protein content, starch content, sugar content, heavy metal content, and trace element content.

[0069] In one embodiment of the present invention, the method further includes selecting and retaining or removing seeds corresponding to the tissue sample based on the analysis results of the genetic and physicochemical information. The analysis results are used to determine the agronomic traits of the seeds, including disease resistance, high yield, lodging resistance, insect resistance, plant height, quality, high oil content, high protein content, high sugar content, aroma, taste, and combinations of multiple traits.

[0070] This invention achieves real-time pose calculation and trajectory prediction of moving seeds through multiple cameras and dynamic algorithms. A flexible gripper assembly grasps the seeds on the conveyor line and transfers them to a motion-adaptive clamping fixture, ensuring the success rate and efficiency of dynamic grasping. A workstation switching module integrates visual recognition, seed grasping, tissue sampling, and sample collection processes, enabling parallel operation and assembly line production, eliminating waiting time between processes, and thus significantly improving overall work efficiency. A deep learning model accurately identifies key features such as seed embryos, achieving precise identification of seed embryos and endosperm, ensuring that specified tissues are cut according to requirements. Based on seed images, real-time cutting path planning and epidermal removal area planning are performed to precisely process target tissues. Non-contact cutting using laser cutting components replaces traditional mechanical drills or grinding wheels, avoiding damage to the seed embryo from physical compression and friction. The laser beam can precisely control the cutting range, removing only target tissues or treating the seed coat surface, thereby greatly improving the germination rate and viability of seeds after sampling. No need to move the laser machine; cutting of specific tissues is achieved by controlling the movement of the beam. This invention simplifies the cutting process by requiring only three steps: grasping, fixing, and cutting. In summary, after reading this invention document, those skilled in the art can make various other corresponding modifications to the technical solutions and concepts based on this invention without creative mental effort, and all of these modifications fall within the scope of protection of this invention.

Claims

1. An automated device for obtaining tissues from seeds, characterized in that, include: The seed feeding module is used to arrange the seeds from the hopper into single-seed arrangements and transport the single seeds along the conveyor line to the working area of ​​the seed gripping tool. The seed gripping tool then picks up the seeds and transfers them to the workstation switching module. The sampling module includes a station switching module, which is equipped with multiple seed fixing devices. The station switching module is used to switch the seed fixing devices between multiple stations. After the seed is fixed by the seed fixing device, the module switches to the vision and cutting station. A cutting component is installed above the cutting station. The cutting component is used to cut the seed fixed by the seed fixing device under the guidance of the path planning module. The vision module includes a path planning module, a positioning module, and a phenotypic feature recognition module. The phenotypic feature recognition module is used to identify the phenotypic features of the seeds, the positioning module is used to locate the coordinates of the seeds, and the path planning module is used to plan the seed cutting and sampling path and guide the cutting component to cut the seeds.

2. The automated device for obtaining tissue from seeds according to claim 1, characterized in that, The seed feeding module includes: The vibratory plate has seed movement path guide grooves on its edge, and the seeds are arranged into a single row and move along the guide grooves by vibration. A conveyor belt, located at the outlet of the vibratory feeder, is used to transport individual seeds to the working area of ​​the seed-grabbing tool.

3. The automated device for obtaining tissues from seeds according to claim 2, characterized in that, The seed grabbing tool is used to grab seeds according to their coordinate information and movement speed during the dynamic transport of seeds on the conveyor line, and place the grabbed seeds in the workstation switching module. The seed scraping tool includes: The motion module drives the positioning of the flexible gripper assembly through movement in the X and Y directions; A flexible gripper assembly is mounted on the drive end of a motion module. The gripping part of the flexible gripper assembly has a deformable flexible gripper that can adapt to seeds of different shapes.

4. The automated device for obtaining tissue from seeds according to claim 1, characterized in that, The workstation switching module has a seed clamping and fixing workstation; The seed fixing device includes: A servo-driven electric gripper, which can output a set gripping force, is assembled at a seed-grabbing fixed station. The servo-driven electric gripper is equipped with an adaptive clamping fixture at its end, which is used to clamp and fix seeds of different shapes.

5. An automated device for obtaining tissues from seeds according to claim 1, characterized in that, The cutting assembly includes a laser machine consisting of a laser machine, a reflector, and a galvanometer. The laser machine is positioned above the sampling station and is used to perform surface treatment on the seed coat of the seeds in the adaptive clamping fixture at the sampling station, and to cut the seeds in the adaptive clamping fixture under the guidance of the path planning module.

6. An automated device for obtaining tissue from seeds according to claim 1, characterized in that, The positioning and recognition module is located at the connection end between the conveyor belt and the vibratory feeder. The positioning and recognition module has multiple first cameras. The first cameras acquire motion images of the seeds on the conveyor belt, control the seed grasping tool to perform pose calculation and target recognition based on the target seeds extracted from the motion images, and drive the motion module and flexible gripper assembly to grasp the movement trajectory of the target seeds.

7. An automated device for obtaining tissues from seeds according to claim 1, characterized in that, The phenotypic feature recognition module includes a second camera, which acquires a representational image of a seed in an adaptive clamping fixture at a vision station, locates the seed position based on the representational image, and performs target feature recognition on the seed to obtain the phenotypic features of the seed.

8. An automated device for obtaining tissue from seeds according to claim 1, characterized in that, The path planning module includes a third camera, which acquires seed images within the adaptive clamping fixture at the sampling station, obtains seed positioning and seed characterization information based on the seed images, performs cutting path planning for the target seed based on the seed images, and performs deskinning region planning for the target seed.

9. An automated device for obtaining tissues from seeds according to claim 1, characterized in that, The cutting assembly also includes a collection module, which includes a material tray, comprising a large material tray and a small material tray. The small material tray is used to load tissue samples and is transported to the tissue sampling station via a material rack and conveyor line. After the cutting assembly cuts the seeds into tissue samples, the tissue samples fall into the small material tray for storage via a distribution tray. The large material tray is used to load the cut seeds, and the small material tray is transported to the collection station via a material rack and conveyor line. The cut seeds fall into the large material tray for storage via a distribution tray.

10. An automated method for obtaining tissue from seeds, the method comprising: Picking a single seed from a group of seeds: Picking the single seed using a seed picking tool; Placement and Fixing: The single seed grasped by the seed grasping tool is moved to the adaptive clamping fixture of the seed fixing device, and the adaptive clamping fixture fixes the seed; Cutting and sampling: The cutting component cuts the seeds fixed in the adaptive clamping fixture to obtain tissue samples.

11. An automated method for obtaining tissue from seeds according to claim 10, characterized in that, It also includes genetic and physicochemical information analysis of the tissue samples. The genetic information includes genomic DNA sequence, SSR genetic markers, SNP genetic markers, transgenic status, alleles, and methylation patterns. The physicochemical information includes oil content, protein content, starch content, sugar content, heavy metal content, and trace element content.

12. The automated method for obtaining tissue from seeds according to claim 10, characterized in that, It also includes analysis results based on the genetic and physicochemical information, selecting to retain or remove seeds corresponding to the tissue sample. The analysis results are used to determine the agronomic traits of the seeds, including disease resistance, high yield, lodging resistance, insect resistance, plant height, quality, high oil content, high protein content, high sugar content, aroma, taste, and combinations of multiple traits.

13. An automated method for obtaining tissue from seeds according to claim 10, characterized in that, It also includes material drop recognition. After the cutting component cuts the seed, when the sample falls into the small material tray, the material drop recognition module identifies whether there is a sample in each compartment of the small material tray. If no sample is identified, it means that the cutting component has not cut the seed.

14. The automated method for obtaining tissue from seeds according to claim 10, characterized in that, The vision module calibrates the seed fixing device, and the cutting component can perform surface degreasing treatment on the calibrated seed fixing device.

Images