A seed potato directional sorting and shunting conveying method, device and medium

Abstract

The application discloses a kind of seed potato directional sequencing and shunt conveying method, device and medium, comprising: in response to loading start signal, control will be processed seed potato interval transport to visual detection station;Obtain the surface image of seed potato at visual detection station, input the surface image into pre-trained target detection model, identify the body area of seed potato, bud eye position and defect area, and output the detection frame of body area, the detection frame of bud eye position and the detection frame of defect area;According to the detection frame of body area and the detection frame of bud eye position, determine the azimuth angle of bud eye relative to the body of seed potato, estimate the physical size grade of seed potato according to the detection frame of body area, and according to azimuth angle, defect area and size grade, generate control instruction set, control instruction set is used to control the bud eye of seed potato is adjusted to preset target posture, and is used to control seed potato is transported to the discharge channel corresponding to size grade and / or defect area.

Description

Technical Field

[0001] This invention relates to the field of seed potato orientation sorting and diversion conveying technology, and more specifically, to a seed potato orientation sorting and diversion conveying method, apparatus and medium. Background Technology

[0002] Currently, most seed potato sorting and distribution operations before processing and planting rely on purely mechanical limit blocks or manual visual sorting, lacking intelligent visual perception and adaptive control functions. Traditional mechanical orientation mechanisms rely solely on mechanical limit blocks for posture correction, exhibiting poor adaptability to differences in seed potato shape, surface damage, and bud orientation, easily leading to problems such as seed potato jamming, posture confusion, and missorting or omission. While existing visual inspection solutions can identify seed potato defects and bud characteristics and perform quality grading, they only reach the information output level of quality grading. They cannot drive the actuator to complete posture correction and distribution based on the identification results, nor do they involve the calculation and feedback control of bud orientation angles. This prevents the realization of a fully automated closed-loop operation from "identification and judgment" to "posture correction" to "grading and distribution," resulting in significant manual intervention still required during subsequent planting and processing. Low automation, low operational efficiency, and bud orientation confusion negatively impact the quality of subsequent cutting and planting. Summary of the Invention

[0003] This invention provides a method, apparatus, and medium for seed potato orientation sorting and diversion conveying, which at least solves the technical problem in the prior art that it is impossible to simultaneously achieve accurate perception of seed potato bud location, automatic posture correction, and integrated closed-loop control of defect size grading and diversion.

[0004] According to one aspect of the present invention, in order to achieve the above-mentioned objective, a method for seed potato directional sorting and diversion transportation is provided, comprising:

[0005] In response to the feeding start signal, the control system will periodically transport the seed potatoes to be processed to the vision inspection station;

[0006] The surface image of the seed potato at the visual inspection station is acquired, and the surface image is input into the pre-trained target detection model to identify the body region, eye position and defect region of the seed potato, and output the detection box of the body region, the detection box of the eye position and the detection box of the defect region.

[0007] Based on the detection frames of the body area and the bud position, the orientation angle of the bud relative to the seed potato body is determined. The physical size grade of the seed potato is estimated based on the detection frames of the body area. Based on the orientation angle, defect area, and size grade, a control instruction set is generated. The control instruction set is used to control the adjustment of the bud of the seed potato to the preset target posture and to control the conveying of the seed potato to the discharge channel corresponding to the size grade and / or defect area.

[0008] Further, acquire surface images of the seed potatoes at the visual inspection station, including:

[0009] When the seed potato is detected to have arrived at the visual inspection station, the surface image of the seed potato is captured.

[0010] The process involves acquiring surface images using a ring-shaped shadowless light source, capturing multiple surface images of the seed potato from different angles, inputting these images into the target detection model, and then fusing the detection results.

[0011] Furthermore, the target detection model is a detection model trained based on the improved YOLOv11 neural network algorithm, and the training dataset used for training is a set of seed potato images pre-labeled with seed potato body, eyes, damage defects, rotten areas and size reference marks;

[0012] The improved YOLOv11 neural network algorithm introduces a weighted bidirectional feature pyramid module into the feature fusion network. The feature pyramid of the weighted bidirectional feature pyramid module includes feature maps of different scales. The weighted bidirectional feature pyramid module is configured to adaptively weight and fuse input features of different scales through learnable weight parameters.

[0013] Furthermore, a small target detection layer is added to the improved YOLOv11 neural network algorithm. This small target detection layer is configured to adjust the size of the anchor frame to fit the size of the bud and / or defect area.

[0014] Furthermore, based on the detection frames of the main body region and the detection frames of the bud positions, the orientation angle of the bud relative to the seed potato body is determined, including:

[0015] Extract the center point coordinates of the detection box in the body region and the center point coordinates of the detection box at the bud position;

[0016] Calculate the relative position vector of the bud center point with respect to the body center point;

[0017] Based on the relative position vector, the angular deviation value of the bud relative to the preset reference direction is determined as the azimuth angle.

[0018] Furthermore, based on the azimuth angle, defect area, and size level, a control instruction set is generated, including:

[0019] When a defect is detected that the defect area contains a rotten area or the damaged area exceeds a preset threshold, a defect rejection instruction is generated. The defect rejection instruction is used to control the pushing of seed potatoes to the waste channel.

[0020] When no defects are detected that contain rotten areas or whose damaged areas exceed a preset threshold, an attitude correction command is generated based on the orientation angle, and a graded diversion command is generated based on the size level.

[0021] Furthermore, the attitude correction command is used to control the rotation of the servo drive mechanism to adjust the seed potato's buds to a preset target attitude.

[0022] Furthermore, after generating the control instruction set, it also includes:

[0023] Acquire real-time speed data during the conveying process;

[0024] Based on real-time speed data, dynamic compensation and correction are performed on the timing of attitude correction actions.

[0025] Update the control instruction set based on the corrected action execution time.

[0026] According to one embodiment of the present invention, a seed potato orientation sorting and diversion conveying device is also provided, comprising:

[0027] The response module is used to respond to the feeding start signal and control the interval delivery of the seed potatoes to be processed to the vision inspection station;

[0028] The acquisition module is used to acquire surface images of seed potatoes at the visual inspection station, input the surface images into a pre-trained target detection model, identify the body region, eye location and defect region of the seed potato, and output the detection boxes of the body region, eye location and defect region.

[0029] The calculation module is used to determine the orientation angle of the bud relative to the seed potato body based on the detection frame of the body area and the detection frame of the bud position, estimate the physical size grade of the seed potato based on the detection frame of the body area, and generate a control instruction set based on the orientation angle, defect area and size grade. The control instruction set is used to control the adjustment of the bud of the seed potato to the preset target posture and to control the conveying of the seed potato to the discharge channel corresponding to the size grade and / or defect area.

[0030] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is executed, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of the present invention.

[0031] In this embodiment of the invention, seed potatoes are periodically transported to the visual inspection station in response to a feeding start signal. Surface images are acquired and input into the target detection model, simultaneously identifying the seed potato body, bud positions, and defect areas. The orientation angle of the bud relative to the seed potato body is calculated based on the coordinates of the center points of the body detection frame and the bud detection frame. At the same time, the physical size level is estimated based on the body detection frame. Finally, the orientation angle, defect areas, and size level are fused to generate a control instruction set to drive subsequent execution actions, realizing integrated intelligent closed-loop control of the entire process of seed potato "visual perception - posture calculation - comprehensive judgment - execution control". This method completely overcomes the shortcomings of traditional solutions, such as a single perception dimension and a disconnect between detection and execution. It avoids the problem of damaging the buds during subsequent cutting due to neglecting the orientation of the buds, and prevents defective seed potatoes from mixing into qualified materials and affecting planting quality. It can perfectly adapt to seed potato materials with different varieties, sizes, and appearances, greatly improving the accuracy of seed potato orientation and sorting and the level of automation in diversion and transportation. It effectively reduces the cost of manual intervention and comprehensively improves the efficiency and reliability of seed potato pretreatment operations. In turn, it solves the technical problem that existing technologies cannot simultaneously achieve accurate perception of seed potato bud orientation, automatic posture correction, and integrated closed-loop control of defect size grading and diversion. Attached Figure Description

[0032] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0033] Figure 1 This is a flowchart of a seed potato directional sorting and diversion transportation method according to one embodiment of the present invention;

[0034] Figure 2 This is a structural block diagram of a seed potato directional sorting and diversion conveying device according to one embodiment of the present invention. Detailed Implementation

[0035] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0036] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0037] According to an embodiment of the present invention, a method for seed potato orientation sorting and diversion transportation is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0038] This method embodiment can be executed in an electronic device or similar computing device that includes memory and a processor. Taking operation on a terminal as an example, the terminal may include one or more processors (processors may include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), digital signal processing (DSP) chips, microcontroller units (MCUs), field-programmable gate arrays (FPGAs), neural network processors (NPUs), tensor processors (TPUs), artificial intelligence (AI) type processors, etc.) and memory for storing data. Optionally, the terminal may also include transmission devices, input / output devices, and display devices for communication functions. Those skilled in the art will understand that the above structural description is merely illustrative and does not limit the structure of the terminal. For example, the terminal may include more or fewer components than described above, or have a different configuration than described above.

[0039] The memory can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the seed potato directional sorting and diversion transportation method in this embodiment of the invention. The processor executes various functional applications and data processing by running the computer program stored in the memory, thereby realizing the aforementioned seed potato directional sorting and diversion transportation method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to a mobile terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0040] The transmission device is used to receive or send data via a network. Specific examples of the network mentioned above may include a wireless network provided by the mobile terminal's communication provider. In one example, the transmission device includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0041] Display devices can be, for example, touchscreen liquid crystal displays (LCDs) and touch displays (also referred to as "touchscreens" or "touch displays"). The LCD allows users to interact with the user interface of the mobile terminal. In some embodiments, the mobile terminal has a graphical user interface (GUI), which allows users to interact with the GUI through finger contact and / or gestures on a touch-sensitive surface. Optional human-computer interaction functions include: creating web pages, drawing, word processing, creating electronic documents, playing games, video conferencing, instant messaging, sending and receiving emails, call interfaces, playing digital video, playing digital music, and / or web browsing, etc. Executable instructions for performing the above human-computer interaction functions are configured / stored in one or more processor-executable computer program products or readable storage media.

[0042] Figure 1 This is a flowchart of a seed potato directional sorting and diversion transportation method according to one embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0043] Step S110: In response to the feeding start signal, the seed potatoes to be processed are intermittently conveyed to the vision inspection station. The specific details are as follows:

[0044] In step S110, before the seed potato orientation sorting and diversion conveying system is officially started, the entire system is in standby mode. Each functional module, including the feeding mechanism, conveying mechanism, vision inspection unit, and control unit, completes its pre-operational self-checks and parameter initialization to ensure the stable operation of subsequent actions. When on-site personnel put the mixed seed potatoes of varying shapes and sizes into the hopper of the feeding mechanism, a feeding start signal is issued through the system control panel or automatic triggering. This start signal is transmitted in real time to the system's main controller via signal cable or wireless communication module, serving as the trigger condition for the official start of the entire orientation sorting and diversion conveying process.

[0045] Upon receiving the feeding start signal, the main controller immediately initiates the control process for the conveying mechanism. The core task is to transport the seed potatoes to be processed to the visual inspection station in an orderly and spaced manner. Specifically, the feeding mechanism uses a vibrating feeder to disperse and loosen the mixed stacked seed potatoes. By adjusting the vibration frequency and amplitude, the vibrating feeder can evenly spread the piled seed potatoes, preventing them from overlapping and obstructing each other. The discharge end of the vibrating feeder connects to a front-mounted flat belt conveyor. After falling from the vibrating feeder, the seed potatoes are arranged sequentially on the flat belt conveyor. The surface of the flat belt conveyor is equipped with anti-slip rubber treads, which increase the friction between the seed potatoes and the belt, effectively reducing the rolling and slipping phenomenon during transportation and ensuring that the seed potatoes are transported forward with a stable posture and speed. The belt speed of the flat belt conveyor line supports frequency conversion speed regulation. The main controller can adjust the belt running speed in real time according to the detection frame rate of the subsequent vision inspection unit and the overall production line cycle requirements. This ensures that each seed potato passes through the vision photography station individually and at intervals, avoiding seed potato stacking and obstruction due to excessive conveying speed, or affecting the overall production efficiency due to excessively slow speed.

[0046] Flexible limiting barriers are installed on both sides of the flat belt conveyor. These barriers extend along the conveying direction and effectively prevent seed potatoes from falling laterally from the sides of the belt due to vibration or tilting during transport, ensuring stable transport within the belt's load-bearing capacity. A visual inspection station is located in the middle of the flat belt conveyor. This station is a pre-defined, fixed area equipped with image acquisition equipment and photoelectric trigger sensors. The photoelectric trigger sensor is installed at the entrance of the visual inspection station. When a seed potato moves along the conveyor to the sensor's position, the sensor sends a positioning detection signal, which is transmitted in real-time to the main controller, triggering the image acquisition equipment to perform image acquisition. Simultaneously, the image acquisition equipment is equipped with a ring-shaped shadowless light source, which illuminates synchronously when the equipment starts, providing uniform, shadow-free illumination for the visual inspection station. This ensures clear and detailed images of the seed potato surface, providing a high-quality image data foundation for accurate identification of subsequent target detection models.

[0047] When seed potatoes pass through the vision inspection station, they are arranged in a single row with intervals. The main controller precisely controls the discharge rhythm of the vibrating feeder and the running speed of the flat belt conveyor to ensure sufficient spacing between adjacent seed potatoes. This allows each seed potato to be completely and independently imaged at the vision inspection station, without multiple seed potatoes occluding or partially overlapping in the image. This ensures that the target detection model can accurately and completely identify each seed potato. After passing through the vision inspection station and completing image acquisition, the seed potatoes continue to be transported along the flat belt conveyor to the rear station for subsequent posture correction and diversion conveying. The entire conveying process is continuous, orderly, and fully automated.

[0048] Step S120: Obtain the surface image of the seed potato at the visual inspection station, input the surface image into the pre-trained target detection model, identify the body region, eye location, and defect region of the seed potato, and output the detection boxes for the body region, eye location, and defect region. The specific content is as follows:

[0049] In step S120, when the seed potato moves along the flat conveyor belt to the vision inspection station, the photoelectric trigger sensor located at the entrance of the station senses the arrival of the seed potato, generates an arrival detection signal, and sends it to the main controller. Upon receiving the signal, the main controller immediately triggers the image acquisition device to perform image acquisition, and simultaneously illuminates the ring-shaped shadowless light source. The ring-shaped shadowless light source remains constantly lit during image acquisition, and its uniform light covers the entire vision inspection station, effectively eliminating local reflections caused by the irregular curved surface of the seed potato and cluttered shadows caused by the surrounding environment. This ensures that the acquired image of the seed potato surface has clear and complete texture details and edge contours, providing a reliable image data foundation for the high-precision recognition of the subsequent target detection model.

[0050] Upon receiving a trigger signal, the image acquisition device captures multiple surface images of the seed potato located at the visual inspection station from different angles. Because seed potatoes have a three-dimensional structure, features such as buds, damage, and rot may be distributed across any facing surface. A single-angle two-dimensional image cannot cover all surface information, easily leading to missed detections or misjudgments. By acquiring multiple surface images of the seed potato from various locations, comprehensive information on the skin condition of the seed potato in all directions can be obtained, enabling the subsequent detection model to make more accurate identifications based on complete data input. The image data acquired from each angle is converted from analog to digital and preprocessed within the image acquisition device before being transmitted in real-time to the industrial control computer via data cable or industrial Ethernet, serving as input data for the target detection model.

[0051] A pre-trained object detection model was deployed on an industrial control computer. This model was built upon an improved YOLOv11 neural network algorithm and trained via transfer learning. The training dataset used during training consisted of a large number of pre-collected field images of seed potatoes. These images covered intact and healthy seed potatoes, seed potatoes with damaged skin, rotten seed potatoes, seed potatoes with insect holes, and seed potato samples with different placement postures, different bud orientations, and different sizes, covering various forms and states that seed potatoes might encounter in actual transportation scenarios. Each training image was meticulously manually annotated using a labeling tool. The labels were divided into five categories: seed potato body, buds, damage defects, rotten areas, and size reference markers. The size reference markers were reference objects with known actual sizes placed next to the seed potatoes to help the model establish the mapping relationship between pixel size and physical size. After annotation, data augmentation processing was performed on the dataset, including random rotation, brightness perturbation, blurring, and noise simulation, to simulate the complex and variable lighting conditions and noise environment of an industrial site, enhancing the model's generalization ability and robustness. The augmented dataset is divided into training, validation, and test sets according to a preset ratio for transfer training, parameter tuning, and performance evaluation of the YOLOv11 model. Finally, a dedicated weight file for seed potato detection is obtained and deployed in the inference engine.

[0052] In terms of model structure, the improved YOLOv11 neural network algorithm introduces a weighted bidirectional feature pyramid module in the feature fusion network. This module's feature pyramid includes feature maps at different scales and can adaptively weight and fuse input features at different scales through learnable weight parameters. Specifically, for each scale in the feature map, the module first performs an upsampling operation to match the higher-scale feature map. The upsampled feature map is then added element-wise to the original feature map to obtain a bidirectionally connected feature map. Subsequently, the feature maps at each scale are fused. During the fusion process, the contribution of each scale feature is automatically adjusted by the learnable weight parameters to highlight important features and suppress irrelevant features, thereby achieving efficient fusion of multi-scale features. In addition, the improved YOLOv11 neural network algorithm adds a small target detection layer. This layer is configured to adjust the size of the anchor boxes to accommodate the small-sized features of buds and / or defect regions. Buds are typical small targets compared to the seed potato body, occupying only a few pixels in the image. If the anchor box configuration of conventional target detection is used, buds are easily missed or misclassified as background. The small target detection layer adjusts the number of channels and enhances image features by applying a 1x1 convolutional layer to the feature map. It also adjusts the size or proportion of the anchor boxes in the feature map to fit the size of the small target. Further adjustments to the number of channels and enhancements to image features are achieved by adding 3x3 or 5x5 convolutional layers. The final output is the detection result of the small target, including the bounding box coordinates of the feature and the probability of the feature category. Through these structural improvements, the model can accurately capture tiny buds and minor damage or rotten areas on the surface of the seed potato, effectively improving the representation ability and recognition accuracy of small target features.

[0053] When multiple surface images are input into the trained target detection model, the model performs independent inference detection on each image. Specifically, the backbone network of the model performs multi-level feature extraction on the input images, the neck network enhances and fuses features at different scales through a weighted bidirectional feature pyramid module and a small target detection layer, and the head network outputs the detection results. For each input image, the model outputs detection boxes for the body region, the bud location, and the defect region. The body region detection box is marked with a rectangle indicating the overall position and boundary of the seed potato in the image, the bud location detection box is marked with a rectangle indicating the precise coordinates of the bud in the image, and the defect region detection box is marked with a rectangle indicating the position and extent of the damaged or rotten area in the image. Each detection box is accompanied by a corresponding confidence score, which reflects the model's confidence in the presence of the corresponding target within the detection box.

[0054] After receiving the detection results from each image, the industrial control computer performs multi-angle detection result fusion processing. Since the same feature may appear simultaneously in multiple images acquired from different angles, the fusion processing requires feature matching and redundancy removal of the detection boxes from images from different angles. Multiple detection boxes corresponding to the same physical feature in different images are merged and associated to ultimately form a complete feature detection result set for the seed potato, including uniquely identified body region detection box information, one or more bud / eye location detection box information, and one or more defect region detection box information. After fusion processing, the industrial control computer transmits the coordinate data and confidence information of the body detection box, bud / eye detection box, and defect detection box to the subsequent attitude calculation module, size estimation module, and instruction generation module for further data analysis and control decisions. The entire image acquisition and target detection and recognition process is completed in a very short time as the seed potato passes through the visual inspection station, achieving efficient, accurate, and automated visual perception of continuous seed potatoes on the conveyor line, providing a comprehensive and reliable decision-making data foundation for subsequent attitude determination, size grading, and diversion conveying.

[0055] Step S140: Based on the detection frames of the body area and the bud position, determine the orientation angle of the bud relative to the seed potato body. Estimate the physical size grade of the seed potato based on the detection frames of the body area. Based on the orientation angle, defect area, and size grade, generate a control instruction set. The control instruction set is used to control the adjustment of the bud of the seed potato to a preset target posture and to control the conveying of the seed potato to the discharge channel corresponding to the size grade and / or defect area. The specific content is as follows:

[0056] In step S140, after acquiring the body region detection box, bud position detection box, and defect region detection box output by the target detection model, the industrial control computer first performs the calculation of the bud's orientation angle relative to the seed potato body. Specifically, the attitude calculation module extracts the center point coordinates of the body region detection box, which represents the approximate geometric center position of the seed potato body in the image plane. Simultaneously, it extracts the center point coordinates of the bud position detection box, which represents the precise position of the bud in the image plane. Subsequently, the attitude calculation module uses the center point of the body region detection box as the reference origin and calculates the relative position vector of the bud detection box's center point relative to this reference origin. This relative position vector includes the offset distance and direction of the bud relative to the seed potato body in both the horizontal and vertical directions. After obtaining the relative position vector, the attitude calculation module further calculates the rotation angle deviation value of the bud relative to a preset reference direction based on this vector. This angle deviation value is the orientation angle of the bud relative to the seed potato body. When there are multiple bud detection boxes in a surface image, the attitude calculation module calculates the angle deviation value of each bud relative to the center point of the body, and selects the optimal bud as the reference for attitude determination according to preset rules. For example, the angle value corresponding to the bud detection box with the highest confidence is selected, or the weighted average of multiple bud angle values ​​is selected, so as to obtain the final orientation angle information.

[0057] During transport along the flat conveyor belt, seed potatoes may exhibit arbitrary and random postures, with their buds facing in any direction. If the orientation of the buds is not uniformly adjusted, the subsequent cutting and planting process can easily result in buds being hit or pieces containing no buds, severely impacting the germination rate and seedling quality. Therefore, accurately obtaining the orientation angle of the buds relative to the seed potato body is the prerequisite and foundation for subsequent posture correction. This orientation angle information will be sent to the posture correction execution mechanism via a control command set, providing an angular deviation basis for uniformly adjusting the seed potato buds to the preset target posture.

[0058] While calculating the bud eye orientation angle, the size estimation module simultaneously performs the estimation of the seed potato's physical size level based on the pixel size of the body region detection box and the camera calibration parameters. During the target detection model training phase, the training dataset includes size reference markers placed on the same plane as the seed potatoes. These reference markers serve as standard references for known actual sizes, and the model establishes a mapping relationship between pixel size and physical size by recognizing these reference markers. During the online inference phase, the size estimation module calls pre-stored camera calibration parameters, including camera intrinsic and extrinsic parameters, such as focal length, principal point coordinates, distortion coefficients, and the camera's installation height and angle relative to the transport plane. Using these calibration parameters, the size estimation module converts the pixel width and height of the body region detection box in the image into the length and width dimensions of the seed potato in actual physical space, obtaining the actual physical size data of the seed potato. After obtaining the physical size data, the size estimation module compares it with the system's preset size grading thresholds. Specifically, the preset size grading thresholds include a first size threshold and a second size threshold, with the first size threshold being less than the second size threshold. When the physical dimensions of the seed potatoes are less than or equal to the first size threshold, they are classified as small-sized; when the physical dimensions are greater than the first size threshold but less than or equal to the second size threshold, they are classified as medium-sized; and when the physical dimensions are greater than the second size threshold, they are classified as large-sized. Different sizes of seed potatoes require different parameters from equipment such as seed metering machines and cutting machines during subsequent planting and processing. Therefore, they need to be graded and sorted according to size to allow subsequent processing equipment to use differentiated operating parameters for different sizes of seed potatoes, improving processing quality and equipment adaptability.

[0059] After calculating the bud eye orientation and estimating the physical size, the industrial control computer further generates a control instruction set based on the orientation, defect area, and size. This control instruction set includes attitude correction instructions and graded diversion instructions, which are the core basis for driving the subsequent actuators. Specifically, when the industrial control computer determines that the seed potato has a rotten area based on the defect area detection frame, it indicates that the seed potato has lost its basic conditions for use as a seed potato and cannot be used for subsequent planting and processing, thus generating a defect rejection instruction. Similarly, when the damaged area within the defect area detection frame exceeds the system's preset damaged area threshold, it indicates that the seed potato's skin is severely damaged, making it highly susceptible to soil-borne diseases or loss of germination ability during subsequent storage or planting, thus also generating a defect rejection instruction. The defect rejection instruction includes the seed potato's identification and location information, used to control the diversion actuator to push the seed potato off the main conveyor line and into the waste channel when it arrives at the diversion station. In the judgment logic, the defect rejection instruction has a higher priority than the posture correction instruction and the grading and diversion instruction. That is, when the seed potato is judged to be rotten or severely damaged, the system directly skips the posture correction and grading and diversion process and does not perform subsequent posture adjustment and specification grading operations. Instead, it directly rejects it as waste material to avoid wasting execution resources on invalid actions.

[0060] When the industrial control computer does not detect any rotten areas in the seed potato and the damaged area does not exceed a preset threshold, it indicates that the seed potato is qualified and ready for use. At this point, the attitude correction and grading process begins. The industrial control computer generates an attitude correction command based on the previously calculated bud orientation angle. This command includes target rotation direction and target rotation angle data. The attitude correction command controls the attitude correction actuator to rotate and adjust the seed potato's bud from its current orientation angle to the preset target attitude. The preset target attitude is a standard bud orientation pre-set by the system according to the requirements of the subsequent cutting and planting process, such as an upward-facing bud or a sideways bud. Different planting equipment and cutting processes may have different requirements for bud orientation, and the system can flexibly set this target attitude parameter according to actual production needs. The rotation direction and rotation angle data carried in the attitude correction command are calculated from the angular deviation between the current bud orientation angle and the preset target attitude. The actuator drives the servo motor to rotate according to the specified direction and angle, thereby precisely adjusting the bud to the target orientation.

[0061] Meanwhile, the industrial control computer generates corresponding grading and diversion commands based on the previously estimated physical size grades of the seed potatoes. When the seed potato size grade is small, the grading and diversion command controls the diversion mechanism to transport the seed potato to the small-size discharge channel; when the seed potato size grade is medium, the grading and diversion command controls the diversion mechanism to transport the seed potato to the medium-size discharge channel; and when the seed potato size grade is large, the grading and diversion command controls the diversion mechanism to transport the seed potato to the large-size discharge channel. The discharge channels for different size grades are respectively connected to the corresponding subsequent collection bins or subsequent pre-planting conveying equipment, so that seed potatoes of different size grades can enter their respective independent processing flow paths, facilitating the use of differentiated processing parameters by subsequent equipment for different sizes. The attitude correction command and the grading and diversion command together constitute a complete control command set, which covers all the driving parameters required for seed potato attitude adjustment and flow direction control.

[0062] After the control instruction set is generated, the industrial control computer sends attitude correction instructions and graded distribution instructions to the lower-level controller via the industrial communication protocol. The lower-level controller, acting as the execution layer control device, is responsible for driving the attitude correction and distribution mechanisms to perform corresponding actions according to the instructions. Specifically, the attitude correction mechanism is located at the attitude correction station downstream of the vision inspection station. This station is equipped with a rotating roller group and a servo drive motor. The rotating roller group carries the seed potatoes and drives them to rotate along their own axis under the drive of the servo motor. After receiving the attitude correction instruction, the lower-level controller parses the rotation direction and angle data contained in the instruction, and controls the servo motor to drive the rotating roller group to rotate in the specified direction and angle, adjusting the seed potato buds to the preset target posture, thus achieving precise attitude correction and orientation sorting of the seed potatoes. The distribution mechanism is located in the distribution conveying section downstream of the attitude correction station. This section is equipped with multiple sets of pneumatic pusher plates or swing arms for distribution and multiple discharge channels, including discharge channels for small-sized seed potatoes, medium-sized seed potatoes, large-sized seed potatoes, and waste materials. After receiving the grading and diversion instructions and the defect rejection instructions, the lower-level controller parses the channel selection information contained in the instructions and controls the corresponding pneumatic pusher or swing arm diversion mechanism to act when the seed potatoes arrive at the diversion station, pushing the seed potatoes into the discharge channel corresponding to the size grade and / or defect area, thus completing the grading and diversion conveying operation. Throughout the process, the generation and issuance of control instructions are carried out in an orderly manner based on the rhythm of the seed potatoes passing through the visual inspection station, ensuring that the seed potatoes running continuously and at high speed on the conveyor line can receive timely and accurate control instructions, thereby achieving fully automated and highly efficient seed potato orientation sorting and grading diversion conveying.

[0063] Based on steps S110 to S140 above, in this embodiment of the invention, seed potatoes are controlled to be transported to the visual inspection station at intervals in response to the feeding start signal. Surface images are acquired and input into the target detection model. The seed potato body, bud position, and defect area are identified simultaneously. The orientation angle of the bud relative to the seed potato body is calculated based on the coordinates of the center point of the body detection frame and the bud detection frame. At the same time, the physical size level is estimated based on the body detection frame. Finally, the orientation angle, defect area, and size level are fused to generate a control instruction set to drive subsequent execution actions. This achieves integrated intelligent closed-loop control of the entire process of seed potato "visual perception - posture calculation - comprehensive judgment - execution control". This method completely overcomes the shortcomings of traditional solutions, such as a single perception dimension and a disconnect between detection and execution. It avoids the problem of damaging the buds during subsequent cutting due to neglecting the orientation of the buds, and prevents defective seed potatoes from mixing into qualified materials and affecting planting quality. It can perfectly adapt to seed potato materials with different varieties, sizes, and appearances, greatly improving the accuracy of seed potato orientation and sorting and the level of automation in diversion and transportation. It effectively reduces the cost of manual intervention and comprehensively improves the efficiency and reliability of seed potato pretreatment operations. In turn, it solves the technical problem that existing technologies cannot simultaneously achieve accurate perception of seed potato bud orientation, automatic posture correction, and integrated closed-loop control of defect size grading and diversion.

[0064] The seed potato orientation sorting and diversion conveying method of the present invention acquires surface images of seed potatoes at a visual inspection station, including: when a seed potato is detected to have arrived at the visual inspection station, triggering the acquisition of surface images of the seed potato; wherein, when acquiring surface images, a ring-shaped shadowless light source is provided, and multiple surface images of the seed potato are acquired from different angles, the multiple surface images are respectively input into the target detection model, and the detection results are fused together.

[0065] Furthermore, the target detection model is a detection model trained based on the improved YOLOv11 neural network algorithm, and the training dataset used for training is a set of seed potato images pre-labeled with seed potato bodies, eyes, damage defects, rotten areas, and size reference marks; the improved YOLOv11 neural network algorithm introduces a weighted bidirectional feature pyramid module in the feature fusion network. The feature pyramid of the weighted bidirectional feature pyramid module includes feature maps of different scales. The weighted bidirectional feature pyramid module is configured to adaptively weight and fuse input features of different scales through learnable weight parameters.

[0066] Furthermore, a small target detection layer is added to the improved YOLOv11 neural network algorithm. This small target detection layer is configured to adjust the size of the anchor frame to fit the size of the bud and / or defect area.

[0067] Furthermore, based on the detection frame of the body region and the detection frame of the bud position, the orientation angle of the bud relative to the seed potato body is determined, including: extracting the center point coordinates of the detection frame of the body region and the center point coordinates of the detection frame of the bud position; calculating the relative position vector of the bud center point relative to the body center point; and determining the angle deviation value of the bud relative to the preset reference direction based on the relative position vector, as the orientation angle.

[0068] Furthermore, a set of control instructions is generated based on the azimuth angle, defect area, and size level, including: when a defect is detected that the defect area contains a rotten area or the damaged area exceeds a preset threshold, a defect rejection instruction is generated, which is used to control the pushing of seed potatoes to the waste channel; when no defect is detected that the defect area contains a rotten area or the damaged area exceeds a preset threshold, an attitude correction instruction is generated based on the azimuth angle, and a graded diversion instruction is generated based on the size level.

[0069] Furthermore, the attitude correction command is used to control the rotation of the servo drive mechanism to adjust the seed potato's buds to a preset target attitude.

[0070] Furthermore, after generating the control instruction set, the process also includes: acquiring real-time speed data during the transport process; dynamically compensating and correcting the execution time of the attitude correction action based on the real-time speed data; and updating the control instruction set based on the corrected execution time of the action.

[0071] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0072] This invention also provides a seed potato orientation sorting and diversion conveying device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0073] Figure 2 According to one embodiment of the present invention, a seed potato orientation sorting and diversion conveying device includes:

[0074] Response module 201 is used to respond to the feeding start signal and control the interval delivery of seed potatoes to be processed to the vision inspection station;

[0075] The acquisition module 202 is used to acquire the surface image of the seed potato at the visual inspection station, input the surface image into the pre-trained target detection model, identify the body region, bud position and defect region of the seed potato, and output the detection box of the body region, the detection box of the bud position and the detection box of the defect region.

[0076] The calculation module 203 is used to determine the orientation angle of the bud relative to the seed potato body based on the detection frame of the body area and the detection frame of the bud position, estimate the physical size grade of the seed potato based on the detection frame of the body area, and generate a control instruction set based on the orientation angle, defect area and size grade. The control instruction set is used to control the adjustment of the bud of the seed potato to a preset target posture and to control the conveying of the seed potato to the discharge channel corresponding to the size grade and / or defect area.

[0077] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0078] According to one embodiment of the present invention, an electronic device is also provided, comprising: a memory storing an executable program; and a processor for running the program, wherein the program executes the above-described seed potato orientation sorting and diversion transportation method during operation.

[0079] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0080] Step S1: In response to the feeding start signal, control the interval delivery of the seed potatoes to be processed to the vision inspection station;

[0081] Step S2: Obtain the surface image of the seed potato at the visual inspection station, input the surface image into the pre-trained target detection model, identify the body region, bud position and defect region of the seed potato, and output the detection box of the body region, the detection box of the bud position and the detection box of the defect region.

[0082] Step S3: Based on the detection frame of the body area and the detection frame of the bud position, determine the orientation angle of the bud relative to the seed potato body. Estimate the physical size grade of the seed potato based on the detection frame of the body area. Based on the orientation angle, defect area and size grade, generate a control instruction set. The control instruction set is used to control the adjustment of the bud of the seed potato to the preset target posture and to control the conveying of the seed potato to the discharge channel corresponding to the size grade and / or defect area.

[0083] According to one embodiment of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the storage medium is located to perform the above-described seed potato orientation sorting and diversion conveying method.

[0084] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps:

[0085] Step S1: In response to the feeding start signal, control the interval delivery of the seed potatoes to be processed to the vision inspection station;

[0086] Step S2: Obtain the surface image of the seed potato at the visual inspection station, input the surface image into the pre-trained target detection model, identify the body region, bud position and defect region of the seed potato, and output the detection box of the body region, the detection box of the bud position and the detection box of the defect region.

[0087] Step S3: Based on the detection frame of the body area and the detection frame of the bud position, determine the orientation angle of the bud relative to the seed potato body, estimate the physical size grade of the seed potato based on the detection frame of the body area, and generate a control instruction set based on the orientation angle, defect area and size grade. The control instruction set is used to control the adjustment of the bud of the seed potato to the preset target posture and to control the conveying of the seed potato to the discharge channel corresponding to the size grade and / or defect area.

[0088] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0089] According to one embodiment of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the above-described seed potato orientation sorting and diversion transportation method.

[0090] Optionally, in this embodiment, the above-mentioned computer program product can be configured as a computer program that performs the following steps:

[0091] Step S1: In response to the feeding start signal, control the interval delivery of the seed potatoes to be processed to the vision inspection station;

[0092] Step S2: Obtain the surface image of the seed potato at the visual inspection station, input the surface image into the pre-trained target detection model, identify the body region, bud position and defect region of the seed potato, and output the detection box of the body region, the detection box of the bud position and the detection box of the defect region.

[0093] Step S3: Based on the detection frame of the body area and the detection frame of the bud position, determine the orientation angle of the bud relative to the seed potato body, estimate the physical size grade of the seed potato based on the detection frame of the body area, and generate a control instruction set based on the orientation angle, defect area and size grade. The control instruction set is used to control the adjustment of the bud of the seed potato to the preset target posture and to control the conveying of the seed potato to the discharge channel corresponding to the size grade and / or defect area.

[0094] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0095] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0096] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between units or modules may be electrical or other forms.

[0097] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0098] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0099] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0100] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for directional sorting and diversion of seed potatoes, characterized in that, include: In response to the feeding start signal, the control system will periodically transport the seed potatoes to be processed to the vision inspection station; The surface image of the seed potato at the visual inspection station is acquired, and the surface image is input into a pre-trained target detection model to identify the body region, eye position and defect region of the seed potato, and output the detection box of the body region, the detection box of the eye position and the detection box of the defect region. Based on the detection frame of the body area and the detection frame of the bud position, the orientation angle of the bud relative to the seed potato body is determined. Based on the detection frame of the body area, the physical size grade of the seed potato is estimated. Based on the orientation angle, the defect area, and the size grade, a control instruction set is generated. The control instruction set is used to control the adjustment of the bud of the seed potato to a preset target posture and to control the conveying of the seed potato to the discharge channel corresponding to the size grade and / or the defect area.

2. The seed potato orientation sorting and diversion conveying method according to claim 1, characterized in that, Acquiring a surface image of the seed potato at the visual inspection station includes: When the seed potato is detected to have arrived at the visual detection station, the surface image of the seed potato is captured. The surface images are acquired using a ring-shaped shadowless light source, and multiple surface images of the seed potato are acquired from different angles. These multiple surface images are then input into the target detection model, and the detection results are fused together.

3. The seed potato orientation sorting and diversion conveying method according to claim 1, characterized in that, The target detection model is a detection model trained based on the improved YOLOv11 neural network algorithm, and the training dataset used for training is a set of seed potato images pre-labeled with seed potato body, buds, damage defects, rotten areas and size reference marks; The improved YOLOv11 neural network algorithm introduces a weighted bidirectional feature pyramid module into the feature fusion network. The feature pyramid of the weighted bidirectional feature pyramid module includes feature maps of different scales. The weighted bidirectional feature pyramid module is configured to adaptively weight and fuse input features of different scales through learnable weight parameters.

4. The seed potato orientation sorting and diversion conveying method according to claim 3, characterized in that, The improved YOLOv11 neural network algorithm adds a small target detection layer, which is configured to adjust the size of the anchor frame to fit the size of the bud and / or the defect area.

5. The seed potato orientation sorting and diversion conveying method according to claim 1, characterized in that, Based on the detection frame of the main body area and the detection frame of the bud position, the orientation angle of the bud relative to the seed potato body is determined, including: Extract the center point coordinates of the detection frame of the body region and the center point coordinates of the detection frame of the bud position; Calculate the relative position vector of the bud center point with respect to the body center point; Based on the relative position vector, the angular deviation value of the bud relative to the preset reference direction is determined, and used as the azimuth angle.

6. The seed potato orientation sorting and diversion conveying method according to claim 1, characterized in that, Based on the azimuth angle, the defect area, and the size level, a control instruction set is generated, including: When a defect is detected that the defect area contains a rotten area or the damaged area exceeds a preset threshold, a defect rejection instruction is generated. The defect rejection instruction is used to control the pushing of the seed potato to the waste channel. When no defect is detected that contains a rotten area or a damaged area exceeding a preset threshold, an attitude correction command is generated based on the azimuth angle, and a graded diversion command is generated based on the size level.

7. The seed potato orientation sorting and diversion conveying method according to claim 6, characterized in that, The posture correction command is used to control the rotation of the servo drive mechanism to adjust the buds of the seed potato to the preset target posture.

8. The seed potato orientation sorting and diversion conveying method according to claim 1, characterized in that, After generating the control instruction set, the method further includes: Acquire real-time speed data during the conveying process; Based on the real-time speed data, the timing of the attitude correction action is dynamically compensated and corrected. The control instruction set is updated based on the corrected action execution time.

9. A seed potato orientation sorting and diversion conveying device, characterized in that, include: The response module is used to respond to the feeding start signal and control the interval delivery of the seed potatoes to be processed to the vision inspection station; The acquisition module is used to acquire the surface image of the seed potato at the visual inspection station, input the surface image into a pre-trained target detection model, identify the body region, bud position and defect region of the seed potato, and output the detection box of the body region, the detection box of the bud position and the detection box of the defect region. The calculation module is used to determine the orientation angle of the bud relative to the seed potato body based on the detection frame of the body area and the detection frame of the bud position, estimate the physical size grade of the seed potato based on the detection frame of the body area, and generate a control instruction set based on the orientation angle, the defect area and the size grade. The control instruction set is used to control the adjustment of the bud of the seed potato to a preset target posture and to control the conveying of the seed potato to the discharge channel corresponding to the size grade and / or the defect area.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device on which the storage medium is located to perform the seed potato orientation sorting and diversion conveying method of any one of claims 1 to 8.

Images