An AI method for removing impurities from bird's nest

Abstract

The application discloses a kind of bird's nest AI impurity removal methods, which is realized based on the picking equipment comprising XY linear motor motion platform, visual detection mechanism and double suction pen execution device.The method comprises the following steps: step S1 visual positioning stage, the coordinates and attitude of bird's nest and impurities are extracted;Step S2 alignment and impurity removal stage, drive the first suction pen to align impurities and use negative pressure suction, if pressure anomaly is monitored during the period, then automatically trigger the pipeline cleaning step comprising high-pressure back flushing and water flushing;Step S3 secondary imaging confirmation stage, closed-loop verification is carried out to determine whether the operation is qualified, and the impurity removal success rate is calculated to perform grading self-diagnosis and parameter optimization;Step S4 bird's nest transfer stage, for qualified bird's nest, use the second suction pen to perform flexible adsorption and transfer unloading based on vacuum degree closed-loop feedback.The application breaks through the limitation of traditional open-loop blind suction, greatly reduces the raw material loss rate and the missed detection rate, significantly improves the stability and intelligent operation level of continuous operation of the equipment.

Description

Technical Field

[0001] This invention belongs to the field of bird's nest impurity sorting technology, and specifically relates to an AI-based method for removing impurities from bird's nest. Background Technology

[0002] As a precious and high-end tonic, the sorting and selection process is crucial in the production of bird's nest, ensuring the final product's quality, purity, and taste. For a long time, the impurity removal process in bird's nest processing has relied heavily on manual labor using tweezers to pick out each nest individually. However, this purely manual method is not only extremely inefficient and costly, but also highly susceptible to visual fatigue and differences in experience, leading to high rates of missed and false detections. Furthermore, prolonged human contact with bird's nest can easily cause secondary cross-contamination.

[0003] To replace manual labor, the industry has begun to try introducing automated cleaning equipment. However, existing automated equipment often has the following drawbacks: First, most of them use traditional screw / belt drive modules, which have slow response speed and insufficient positioning accuracy, and cannot meet the needs of efficient bird's nest sorting; second, existing vision guidance equipment is mostly an open-loop control system, lacking a closed-loop feedback and verification mechanism between visual detection and grasping action, which cannot cope with the randomness of bird's nest posture, and the false negative rate is still high; third, it lacks intelligent status monitoring and adaptive adjustment capabilities. When there is slight blockage in the pipeline or parameter deviation, it cannot achieve self-diagnosis and self-cleaning, which greatly reduces the long-term stability and yield of the equipment.

[0004] Therefore, this invention provides an AI-based method for removing impurities from bird's nests, aiming to solve the problems of low positioning efficiency, unstable yield due to lack of closed-loop verification, and lack of abnormal self-diagnosis and self-cleaning mechanisms in existing automated bird's nest sorting equipment. Summary of the Invention

[0005] This invention provides an AI-based method for removing impurities from bird's nest, aiming to solve the problems pointed out in the background art.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: This application provides an AI-based method for removing impurities from bird's nest, which is implemented based on an AI-based bird's nest impurity sorting device. The device includes a bird's nest loading platform, an XY linear motor motion platform mounted on the platform, and a vision detection mechanism and a dual-suction pen execution device mounted on the XY linear motor motion platform. The dual-suction pen execution device integrates a first suction pen responsible for impurity removal and a second suction pen responsible for transfer. The method includes the following steps: Step S1: Visual positioning stage, the visual inspection mechanism images the bird's nest on the bird's nest loading platform, and the control system extracts the bird's nest position parameters, including bird's nest coordinates, impurity coordinates and impurity posture; Step S2: Alignment and impurity removal stage. According to the impurity coordinates, control the XY linear motor motion platform to drive the axis of the first suction pen to align with the center point of the impurity, and control the first suction pen to descend, and connect the negative pressure air source to suck up and remove the impurities. Step S3: Secondary imaging confirmation stage. After the impurity removal action is completed, the XY linear motor motion platform is controlled to drive the vision detection mechanism to move above the bird's nest area after impurity removal to perform secondary closed-loop imaging by the camera. The control system determines whether the picking operation of this cycle is qualified by comparing the image features and the number of impurities before and after impurity removal. If the picking operation for this cycle is deemed satisfactory, proceed to step S4; If the picking operation in this cycle is deemed unqualified, then step S1 is repeated based on the secondary imaging image. Step S4: Bird's nest transfer stage. Control the second suction pen to descend and adhere to the surface of the complete bird's nest body. Adjust the negative pressure suction according to the vacuum change fed back by the air pressure sensor to flexibly grab and transfer the bird's nest body into the corresponding quality grade feeding trough.

[0007] Furthermore, step S3 also includes the following steps: The control system determines the actual number of impurities successfully removed based on the difference between the initial number of impurities before impurity removal obtained in step S1 and the number of residual impurities after impurity removal obtained in step S3. The actual impurity removal success rate P is calculated by comparing the number of impurities actually successfully removed with the total number of impurity removal actions of the first suction pen in the corresponding period as recorded in the system log.

[0008] Furthermore, the control system is preset with a first-level yield threshold T1 and a second-level yield threshold T2, and T1 < T2; When the actual impurity removal success rate P satisfies T1≤P<T2, the control system determines it as a slight missed detection and automatically performs software parameter fine-tuning: lowering the confidence threshold of the recognition algorithm or increasing the exposure time of the visual inspection mechanism. When the actual impurity removal success rate P satisfies P < T1, the control system determines that the execution efficiency has decreased, and on the basis of fine-tuning the execution software parameters, it superimposes physical action compensation: increases the depth compensation of the first suction pen descent, or extends the holding time of the negative pressure air source suction.

[0009] Furthermore, in step S2, during the process of removing impurities by suction using a negative pressure air source, the pressure sensor in the air path of the first suction pen transmits the absolute negative pressure value in the airway in real time at a preset sampling rate. If the absolute negative pressure value exceeds the upper limit of the preset safety threshold range, the control system determines that the air path is blocked abnormally. The control system issues an interrupt command to forcibly suspend the current impurity removal cycle and automatically triggers a sudden cleaning step for the pipeline.

[0010] Furthermore, the timing sequence of the pipeline emergency cleaning step includes: First, shut off the negative pressure solenoid valve and move the dual suction pen actuator above the preset cleaning area; After a set delay time, the high-pressure air valve is opened to introduce high-pressure pulsed air into the air path of the first suction pen for backflushing. Close the high-pressure air valve and open the negative pressure bypass valve of the cleaning water circuit. Use negative pressure to draw pure water to perform a secondary flushing and washing of the inner wall of the air circuit. High-pressure pulsed air was introduced again to dry the water stains inside the air circuit, and then the impurity removal operation was resumed.

[0011] Furthermore, in step S1, the visual inspection mechanism includes a composite ring light source, which is composed of an outer ring shadowless LED white light source and an inner ring coaxial non-reflective micro blue light source.

[0012] Furthermore, the imaging process employs frequency-division flash control: First, control the outer ring shadowless LED white light source to perform a strobe exposure to obtain the macroscopic outline and cup-shaped image of the bird's nest body; Subsequently, the inner ring coaxial non-reflective micro blue light source is controlled to perform a secondary stroboscopic exposure to obtain a high-contrast image of the impurities. The control system overlays and fuses the images from the two exposures, and inputs them into a deep learning AI recognition model to output the location parameters of the bird's nest.

[0013] Furthermore, when extracting the posture of impurities, the deep learning AI recognition model classifies the impurity type by calculating the aspect ratio of the bounding rectangle of the impurity: if the aspect ratio is greater than a set threshold, it is determined to be a long strip fiber impurity; if the aspect ratio is less than the set threshold, it is determined to be a dotted mold spot impurity. The control system dynamically calls the suction time and negative pressure intensity parameters corresponding to the first suction pen in step S2 according to the classified impurity type.

[0014] Furthermore, in step S4, the specific process of the bird's nest transfer stage is as follows: When the second suction pen descends and adheres to the main body of the bird's nest, the negative pressure flow is slowly increased through the electric proportional valve; When the pressure sensor detects that the vacuum level in the sealed cavity formed by the suction cup and the surface of the bird's nest reaches the first preset value, it controls the second suction pen to stop descending; The electric proportional valve continues to increase the suction force. When the vacuum reaches the second preset value, the PID controller dynamically adjusts the opening of the proportional valve to lock the negative pressure value at the second preset value, and then controls the XY linear motor motion platform to perform the transfer.

[0015] Furthermore, the XY linear motor motion platform includes a Y-axis linear motor module and an X-axis linear motor module that are perpendicular to each other and both adopt a linear motor direct drive structure with no mechanical backlash; the first pen and the second pen are both equipped with independent Z-axis lifting drive motors, which are arranged side by side in physical layout and can achieve staggered or synchronous lifting and cooperative operation through the control system.

[0016] Compared with the prior art, the present invention has the following technical effects: 1. Compared to traditional automated sorting equipment, which often uses open-loop blind suction and cannot confirm whether impurities have been truly removed, this invention achieves result verification through secondary closed-loop imaging, fundamentally preventing missed defects from flowing into the next process. Furthermore, by setting primary and secondary yield thresholds, this invention endows the equipment with automatic diagnostic capabilities: the system can automatically determine whether the current missed inspection is a minor issue caused by light or algorithm parameter deviations, or a "decreased execution efficiency" due to prolonged pneumatic actuator operation, and automatically fine-tunes the feature exposure parameters or physical pressure depth accordingly. This adaptive evolution mechanism greatly reduces the equipment's maintenance threshold and ensures long-term yield stability.

[0017] 2. Impurities in semi-dried bird's nest, such as mold spots and fine down, are highly sticky and often cause gradual blockage of the tubing or air leakage due to poor sealing during suction. This invention automatically suspends the current impurity removal cycle and executes a self-cleaning sequence by real-time monitoring of the instantaneous spikes or drops in absolute pressure in the airway, without manual intervention. This foolproof mechanism completely solves the problem of continuous batch missed detections caused by suction power reduction due to localized blockage, effectively avoiding secondary contamination or misoperation of expensive bird's nest raw materials. Attached Figure Description

[0018] Figure 1 This is a flowchart of an AI-based method for removing impurities from bird's nest, as described in this invention. Figure 2 This is an isometric view of the overall structure of the AI-based bird's nest impurity sorting machine in the AI-based bird's nest impurity removal method of the present invention.

[0019] In the picture: 1. Y-axis linear motor module; 2. X-axis linear motor module; 3. Mounting bracket; 4. Dual pen suction actuator; 5. Bird's nest loading platform; 6. Visual inspection agency. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present application and with reference to the accompanying drawings.

[0021] This embodiment provides an AI-based method for removing impurities from bird's nest, which relies on an AI-based bird's nest impurity sorting machine. For example... Figure 2 As shown, the equipment mainly consists of a bird's nest loading platform 5, an XY linear motor motion platform, a mounting frame 3 slidingly supported on the platform, and a dual-pen suction device 4 and a vision inspection mechanism 6 integrated on the mounting frame 3. The XY linear motor motion platform includes a Y-axis module 1 and an X-axis module 2 that are perpendicular to each other and both adopt a backlash-free direct drive structure, ensuring extremely high dynamic response and micron-level positioning accuracy. The dual-pen suction device 4 includes a first suction pen and a second suction pen that are independently controlled and equipped with independent Z-axis motors.

[0022] like Figure 1 As shown, the AI-based method for removing impurities from bird's nest according to the present invention includes the following specific steps: Step S1, Visual Localization Stage: The visual inspection unit 6 employs a composite ring light source, consisting of an outer ring of shadowless LED white light (color temperature 5500K-6500K) and an inner ring of coaxial, non-reflective micro-blue light (wavelength 450nm-470nm). To overcome the interference of the translucent crystal reflection of the bird's nest on identification, the system uses frequency-division flash control: first, the outer ring of white light is used for a single flash exposure to obtain the macroscopic outline and shape information of the bird's nest; within 50 milliseconds, the inner ring of micro-blue light is used for a second flash exposure. The micro-blue light has extremely strong light absorption characteristics for black down, wood chips, and dark mold spots, which can significantly improve the contrast of impurities.

[0023] The control system fuses the images from the two exposures and inputs them into a recognition model based on a deep learning AI convolutional neural network. The model extracts the two-dimensional coordinates of the impurities and classifies their orientation by calculating the aspect ratio of the bounding rectangle of the impurity: if the aspect ratio is greater than 3, it is identified as a long, thin fiber; if the aspect ratio is less than 1.5, it is identified as a dotted mold spot. Based on this, the system dynamically calls upon the subsequent suction time and negative pressure intensity parameters of the first suction pen.

[0024] Step S2, Alignment and Noise Removal Stage: After receiving the coordinates, the control system drives the XY linear motor motion platform to translate at high speed, so that the first suction pen aligns with the center point of the impurity. Then the first suction pen descends and connects the negative pressure to suck up the impurity.

[0025] During this process, the system is equipped with an air path anomaly monitoring and a passive triggering mechanism for sudden cleaning. The control system has a preset normal negative pressure safety threshold range, such as -40kPa to -60kPa. A high-precision pressure sensor in the first suction pen's air path transmits the absolute pressure in the air path in real time. If the absolute negative pressure value is greater than -80kPa, it is determined that viscous impurities are clogging the suction pen; if the absolute negative pressure value is less than -20kPa and close to normal pressure, it is determined that the airflow is short-circuited or leaking; when an airflow short-circuit or leak is determined, the system issues a warning alarm to remind the operator to check the air path. When the system determines that sticky impurities are clogging the suction pen, it issues an interrupt command and triggers a sudden cleaning sequence for the pipeline: First, the negative pressure solenoid valve is shut off, and the dual suction pen actuator 4 is moved above the preset cleaning area; after a delay of 0.1 seconds, the high-pressure air valve is opened to introduce 0.6MPa-0.8MPa high-pressure pulsed air for backflushing for 0.5 seconds; then, the negative pressure bypass valve of the cleaning water circuit is opened, and 20ml of pure water is drawn in for 0.5 seconds to rinse the inner wall of the air circuit a second time; finally, high-pressure air is introduced again for 0.5 seconds to dry the inside of the pipe. After the entire self-cleaning cycle is completed, the impurity removal operation resumes.

[0026] Step S3, Secondary Imaging Confirmation Stage: After the impurity removal process is completed, the control XY platform moves the vision inspection mechanism 6 to above the bird's nest for closed-loop secondary imaging. The control system compares the image features before and after impurity removal to determine whether the bird's nest is qualified.

[0027] To achieve intelligent graded self-diagnosis and parameter adjustment of the equipment, the system retrieves the initial number of impurities N1 and the number of residual impurities N2 identified, obtains the actual number of impurities successfully removed N0, and calculates the actual impurity removal success rate P (P=N0 / M) by combining the total number of impurity removal actions M of the first suction pen in the corresponding cycle.

[0028] The system presets a primary yield threshold T1 (e.g., 80%) and a secondary yield threshold T2 (e.g., 90%). When 80% ≤ P < 90%, the system determines "minor missed detection" and automatically performs software fine-tuning: increasing the exposure time by 10%. When P < 80%, the system determines "decreased execution efficiency" and superimposes hardware action compensation in addition to software parameter tuning: automatically increasing the depth compensation of the first suction pen's Z-axis downward pressure (increasing by 0.5-1.0 mm) or extending the suction holding time by 200 milliseconds. Through closed-loop parameter tuning, the equipment can adaptively overcome ambient light attenuation or slight fatigue of pneumatic components.

[0029] Step S4, Bird's Nest Transfer Stage: Once step S3 determines that the sample is qualified, control the second suction pen to move and descend as a whole to adhere to the surface of the complete bird's nest body.

[0030] This stage employs flexible suction control: as the second suction pen descends, the negative pressure flow is slowly increased via an electro-proportional valve, while a pressure sensor monitors the vacuum level within the sealed chamber in real time. When the vacuum level reaches a first preset value (e.g., -15 kPa, indicating the suction nozzle is in contact but not generating significant pulling force), the Z-axis motor stops pressing down to prevent crushing the bird's nest. The proportional valve continues to slowly increase the suction force, and when the vacuum level reaches a second preset value (e.g., -35 kPa, this setting ensures the pulling force is slightly greater than the weight of the bird's nest but far less than the critical tearing force of the bird's nest fibers), the PID controller dynamically adjusts the opening of the proportional valve to maintain a constant air pressure at this negative pressure value. Subsequently, the Z-axis motor rises, and the XY linear motor performs high-speed transfer. This mechanism completely avoids physical damage to the bird's nest caused by instantaneous high negative pressure.

[0031] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A bird's nest AI impurity removal method, based on a bird's nest impurity AI sorting device, the device including a bird's nest loading platform (5), an XY linear motor motion platform mounted on the platform, and a vision detection mechanism (6) and a dual suction pen execution device (4) mounted on the XY linear motor motion platform; the dual suction pen execution device (4) integrates a first suction pen responsible for impurity removal and a second suction pen responsible for transfer; Its features are, The method includes the following steps: Step S1: Visual positioning stage, the bird's nest on the bird's nest loading platform (5) is imaged by the visual detection mechanism (6), and the bird's nest position parameters are extracted by the control system. The bird's nest position parameters include bird's nest coordinates, impurity coordinates and impurity posture. Step S2: Alignment and impurity removal stage. According to the impurity coordinates, control the XY linear motor motion platform to drive the axis of the first suction pen to align with the center point of the impurity, and control the first suction pen to descend, and connect the negative pressure air source to suck up and remove the impurities. Step S3: Secondary imaging confirmation stage. After the impurity removal action is completed, the XY linear motor motion platform is controlled to drive the vision detection mechanism (6) to move above the bird's nest area after impurity removal to perform secondary closed-loop imaging of the camera. The control system determines whether the periodic picking operation is qualified by comparing the image features and the number of impurities before and after impurity removal. If the picking operation for this cycle is deemed satisfactory, proceed to step S4; If the picking operation in this cycle is deemed unqualified, then step S1 is repeated based on the secondary imaging image. Step S4: Bird's nest transfer stage, control the second suction pen to descend and adhere to the complete bird's nest body surface, and finely adjust the negative pressure suction according to the vacuum degree change fed back by the air pressure sensor to flexibly grab and transfer the bird's nest body to the corresponding quality grade feeding tank; Step S3 further includes the following steps: The control system calculates the actual number of impurities successfully removed based on the difference between the initial number of impurities before impurity removal obtained in step S1 and the number of residual impurities after impurity removal obtained in step S3; the actual number of impurities successfully removed is calculated as a ratio to the total number of impurity removal actions of the first suction pen in the corresponding period as recorded in the system log, to obtain the actual impurity removal success rate P. The control system has a preset first-level yield threshold T1 and a second-level yield threshold T2, and T1 < T2; when the actual impurity removal success rate P satisfies T1 ≤ P < T2, the control system determines it as a slight missed detection and automatically performs software parameter fine-tuning: lowering the confidence threshold of the recognition algorithm or increasing the exposure time of the visual inspection mechanism (6); when the actual impurity removal success rate P satisfies P < T1, the control system determines it as a decrease in execution efficiency, and superimposes physical action compensation on the basis of performing software parameter fine-tuning: increasing the depth compensation amount of the first suction pen descent, or extending the holding time of the negative pressure air source suction.

2. The AI-based method for removing impurities from bird's nest according to claim 1, characterized in that, In step S2, during the process of using a negative pressure air source to suction and remove impurities, the pressure sensor in the air path of the first suction pen transmits the absolute negative pressure value in the air passage in real time at a preset sampling rate. If the absolute negative pressure value exceeds the upper limit of the preset safety threshold range, the control system determines that the air path is blocked abnormally. The control system issues an interrupt command to forcibly suspend the current impurity removal cycle and automatically triggers a sudden cleaning step for the pipeline.

3. The AI-based method for removing impurities from bird's nest according to claim 2, characterized in that, The timing sequence of the pipeline emergency cleaning step includes: First, shut off the negative pressure solenoid valve and move the double suction pen actuator (4) above the preset cleaning area; After a set delay time, the high-pressure air valve is opened to introduce high-pressure pulsed air into the air path of the first suction pen for backflushing. Close the high-pressure air valve and open the negative pressure bypass valve of the cleaning water circuit. Use negative pressure to draw pure water to perform a secondary flushing and washing of the inner wall of the air circuit. High-pressure pulsed air was introduced again to dry the water stains inside the air circuit, and then the impurity removal operation was resumed.

4. The AI ​​method for removing impurities from bird's nest according to claim 3, characterized in that, In step S1, the visual inspection mechanism (6) includes a composite ring light source, which is composed of an outer ring shadowless LED white light source and an inner ring coaxial non-reflective micro blue light source.

5. The AI ​​method for removing impurities from bird's nest according to claim 4, characterized in that, In step S1, the imaging process employs frequency-division flash control: First, control the outer ring shadowless LED white light source to perform a strobe exposure to obtain the macroscopic outline and cup-shaped image of the bird's nest body; Subsequently, the inner ring coaxial non-reflective micro blue light source is controlled to perform a secondary stroboscopic exposure to obtain a high-contrast image of the impurities. The control system overlays and fuses the images from the two exposures, and inputs them into a deep learning AI recognition model to output the location parameters of the bird's nest.

6. The AI-based method for removing impurities from bird's nest according to claim 5, characterized in that, When extracting the posture of impurities, the deep learning AI recognition model classifies the impurity type by calculating the aspect ratio of the bounding rectangle of the impurity: if the aspect ratio is greater than a set threshold, it is determined to be a long strip fiber impurity; if the aspect ratio is less than the set threshold, it is determined to be a dotted mold spot impurity. The control system dynamically calls the suction time and negative pressure intensity parameters corresponding to the first suction pen in step S2 according to the classified impurity type.

7. The AI ​​method for removing impurities from bird's nest according to claim 6, characterized in that, In step S4, the specific process of the bird's nest transfer stage is as follows: When the second suction pen descends and adheres to the main body of the bird's nest, the negative pressure flow is slowly increased through the electric proportional valve; When the pressure sensor detects that the vacuum level in the sealed cavity formed by the suction cup and the surface of the bird's nest reaches the first preset value, it controls the second suction pen to stop descending; The electric proportional valve continues to increase the suction force. When the vacuum reaches the second preset value, the PID controller dynamically adjusts the opening of the proportional valve to lock the negative pressure value at the second preset value, and then controls the XY linear motor motion platform to perform the transfer.

8. The method for removing impurities from bird's nest using AI according to claim 7, characterized in that, The XY linear motor motion platform includes a Y-axis linear motor module (1) and an X-axis linear motor module (2) that are perpendicular to each other and both adopt a linear motor direct drive structure with no mechanical backlash; the first pen and the second pen are both equipped with independent Z-axis lifting drive motors, which are arranged side by side in physical layout and can achieve staggered or synchronous lifting and cooperative operation through the control system.

Images