Automatic feeding control system based on industrial vision

Abstract

The invention discloses an automatic feeding control system based on industrial vision, and belongs to the technical field of feeding control systems.The automatic feeding control system comprises a visual recognition module used for collecting image information such as the appearance, the size and the spatial position of a workpiece; the mechanical clamping module is used for grabbing, carrying and placing the workpiece according to the control instruction; the control operation module undertakes a core control task of the system and is used for coordinating cooperative work among the modules; the data processing and analyzing module is used for deeply processing and analyzing the image data acquired by the visual identification module and the data fed back by other sensors; the environment perception and compensation module is used for assisting the system to dynamically adjust working parameters of equipment, counteracting the influence of environment interference on the operation precision and automatically realizing compensation adjustment; the system can quickly and accurately identify the position and the posture of the workpiece in an industrial production environment with a high real-time requirement, and stably clamp the workpiece to carry out feeding operation.

Description

Technical Field

[0001] This invention relates to the field of material feeding control system technology, specifically an automatic material feeding control system based on industrial vision. Background Technology

[0002] As a key starting point in the production process, the performance of the automatic feeding system directly affects the efficiency and quality of the entire production line. The automatic feeding control system based on industrial vision, with the help of advanced machine vision technology, can quickly identify and accurately locate the position and posture of the workpiece, thereby guiding the robot or clamping equipment to complete the efficient and accurate feeding operation, which greatly improves the degree of production automation, effectively reduces manual intervention, and improves production efficiency.

[0003] For example, patent CN119225305A discloses a feeding control system for silicon steel sheet production, including a conveying unit, a sensing feedback unit, a sensing and imaging unit, a laser positioning unit, and a feeding control unit. The feeding control unit includes a communication interaction unit, a data analysis module, an image processing module, a control strategy module, a feeding management module, a fault monitoring module, and an execution report module. This system enables intelligent feeding control by calculating and analyzing changes during the feeding process, combining feeding requirements, and considering characteristics such as the size, shape, and surface quality of the silicon steel sheets. It controls the accuracy of feeding, and through real-time monitoring and adjustment of the feeding strategy, it can adapt to the feeding control of silicon steel sheets of different specifications, improving the overall efficiency of the production line. It reduces human error and unnecessary time consumption, lowers production costs, and improves production efficiency and product quality.

[0004] For example, patent CN117302901A discloses a feeding control method, a feeding control system, and a storage medium. The method includes: acquiring an identifier of a container containing materials and acquiring an image of the materials to be identified obtained by photographing them; acquiring feeding configuration information pre-associated with the identifier, and performing target recognition on the image to be identified using a preset recognition model to obtain the first type information of the materials; then controlling the operating status of the feeding equipment according to the feeding configuration information, the first type information, and the pre-stored station information of the designated station; when any information in the first type information and the station information does not match the feeding configuration information, the feeding equipment stops feeding. In this way, the accuracy of feeding can be improved and the problem of feeding errors can be mitigated.

[0005] For example, patent CN119820585B discloses a material feeding control system and control method based on a robotic arm. The method includes: establishing a first three-dimensional image of the robotic arm's environment; extracting a movement path feature vector based on the first three-dimensional image; determining the target position of the material based on the movement path feature vector; when the robotic arm reaches the target position of the material, preprocessing the first three-dimensional image and establishing a first coordinate system of the image; calculating the feature value of the material based on the first coordinate system of the image; extracting an image optimization factor based on the feature value; optimizing the spatial vector based on the image optimization factor to obtain the target vector of the material; when gripping the material, establishing a second coordinate system of the image based on a second three-dimensional image; extracting the coordinates of the gripping point where the material and the robotic arm come into contact based on the second coordinate system of the image; optimizing the coordinates of the gripping point based on the image coordinate optimization factor to obtain the feeding coordinates. This invention has excellent gripping and adjustment features; however, some existing systems rely on industrial vision positioning. When the workpiece surface has stains, reflections, or obstructions, or when the ambient light changes, the vision camera may not be able to accurately identify the workpiece position or posture, causing the robot arm to be unable to accurately grip the workpiece. In addition, after long-term use, gripping devices such as robot arms may deform or loosen, making it impossible to effectively control the gripping force on the workpiece. The workpiece may break under excessively tight gripping force, and may loosen and fall off under excessively loose gripping force.

[0006] To address the aforementioned issues, an innovative design is urgently needed based on the existing material feeding control system. Summary of the Invention

[0007] The purpose of this invention is to provide an automatic feeding control system based on industrial vision, in order to solve the problems mentioned in the background art. Some existing systems rely on industrial vision for positioning. When the workpiece surface has stains, reflections, obstructions, or changes in ambient light, the vision camera may not be able to accurately identify the position or posture of the workpiece, resulting in the robot arm being unable to accurately clamp the workpiece. In addition, after long-term use, the robot arm and other clamping devices may deform or loosen, resulting in the inability to effectively control the clamping force on the workpiece. The workpiece may be damaged under excessively tight clamping force, and may fall off under excessively loose clamping force.

[0008] To achieve the above objectives, the present invention provides the following technical solution: an automatic feeding control system based on industrial vision, the automatic feeding control system comprising: The visual recognition module is equipped with a high-resolution industrial camera and a professional image acquisition card. It is used to acquire image information such as the appearance, size and spatial position of the workpiece, filter out industrial environment interference, and generate high-definition and timely raw data, providing a reliable input basis for subsequent data processing.

[0009] The mechanical clamping module, including a high-precision robotic arm and a customized auxiliary gripper, performs the gripping, handling and placement of workpieces according to control commands, and also has an anti-collision protection mechanism.

[0010] The control and calculation module, which uses a high-performance industrial computer or programmable logic controller, undertakes the core control tasks of the system. It coordinates the collaborative work between various modules and performs precise calculations and decisions based on various feedback information.

[0011] The data processing and analysis module integrates professional data processing software and advanced algorithms to perform in-depth processing and analysis on image data collected by the visual recognition module and data from other sensors, extracting key feature information to provide decision support for the control and computing module.

[0012] The environmental sensing and compensation module is equipped with various types of environmental sensors, including light, temperature, and humidity sensors, to monitor changes in ambient light, temperature, and humidity in real time. The data is fed back to the control and calculation module to help the system dynamically adjust the equipment's operating parameters, offset the impact of environmental interference on operating accuracy, and automatically achieve compensation and adjustment.

[0013] Preferably, the high-resolution industrial camera organically combines a visible light camera with an infrared thermal imager; wherein, the visible light camera can capture rich color, texture and detail information, and in an ideal environment with sufficient light, it can accurately acquire the surface features and geometry of the workpiece, providing basic data for visual recognition; the infrared thermal imager images based on the thermal radiation characteristics of the object itself, and can stably acquire the contour and position information of the workpiece in low light, when the workpiece surface has stains and the workpiece is partially occluded.

[0014] Preferably, the high-precision robotic arm adopts a multi-axis linkage structure, equipped with an industrial-grade servo motor and a precision reducer, achieving a positioning accuracy of ±0.01mm and a repeatability accuracy within ±0.005mm. It can flexibly adjust its posture in three-dimensional space to adapt to the workpiece transfer needs of different workstations. The customized auxiliary gripper is specifically designed according to the workpiece's material, shape, and surface characteristics. Specific gripping methods include: pneumatic grippers, driven by compressed air, with fast response speed; the gripping force is flexibly adjustable within the range of 0.1-100N via a pressure regulating valve, suitable for various applications. It is suitable for handling medium to heavy regular workpieces of metal and plastic weighing 1-50kg; the electric gripper, driven by a servo motor, has a position control accuracy of ±0.02mm and a closed-loop force control function, which adjusts the clamping parameters in real time. It is suitable for electronic components, micro gear precision parts, and thin-walled metal workpieces that are easily deformed, and supports rapid switching between multiple types of workpieces; the vacuum suction cup works on the principle of negative pressure adsorption, without contacting the edge of the workpiece. It is suitable for flat parts with thin metal plates and glass substrates, as well as easily damaged parts such as ceramic sheets and plastic films. There is no mechanical stress during the adsorption process, which effectively protects the integrity of the workpiece surface.

[0015] Preferably, the anti-collision protection mechanism is based on the accurate capture of collision risk signals, specifically including: a force feedback sensor, integrated into the end joint of the robot or the connection between the auxiliary gripper and the robot, to detect the axial force, radial force and torque experienced by the robot during its movement in real time. When the robot comes into contact with an obstacle, the force will instantly exceed a preset threshold, and the sensor will immediately convert the force signal into an electrical signal and transmit it to the control and calculation module; a laser ranging / visual obstacle avoidance sensor, with laser ranging sensors installed at key nodes in the robot's movement path and at the end of the robot, which can scan the distance of objects in the surrounding 30-180° range in real time. When the visual detection shows that the distance between the obstacle and the robot is less than the safety threshold, an early warning signal is immediately triggered; and travel limit switches, with mechanical or photoelectric limit switches installed at the movement limit positions of each axis of the robot. When the robot is about to reach its physical limit due to command deviation or malfunction, the switch is triggered, forcibly cutting off the motion drive signal of that axis to avoid overload damage to the mechanical structure.

[0016] Preferably, the high-performance industrial computer is equipped with a multi-core industrial-grade processor, ECC fault-tolerant memory, and high-speed solid-state drive, possessing powerful data parallel computing capabilities and multi-task processing capabilities, simultaneously processing massive amounts of data from multiple modules such as visual recognition, mechanical control, and environmental compensation; the programmable logic controller adopts an anti-interference design, has a wide operating temperature range and high reliability, supports mixed digital / analog input and output, and realizes logic control through ladder diagrams and structured text industrial programming languages, adapting to standardized material feeding scenarios with fixed processes and fast response requirements, and is suitable for long-term continuous production environments.

[0017] Preferably, when the control and calculation module receives a collision risk signal, it executes different response strategies according to the risk level, specifically including: Level 1 response: when the laser ranging sensor detects an obstacle but no actual contact occurs, the module immediately adjusts the robot's movement path and avoids the risk without interrupting the feeding process by replanning the trajectory in real time; Level 2 response: when the force feedback sensor detects that the force is close to the threshold and the limit switch triggers a warning, the module immediately issues a deceleration command to reduce the robot's movement speed; if the force continues to rise to the threshold and the obstacle is not avoided, an emergency stop is triggered, and all axes of the robot are immediately de-energized and locked to prevent the collision force from increasing further; Level 3 response: in the event of a minor collision, after triggering the emergency stop, the module automatically records the time, location, and force of the collision, displays an alarm prompt on the system interface, and the robot automatically returns to a safe position, waiting for staff to troubleshoot the fault and restart after confirming safety.

[0018] Preferably, the data processing and analysis module establishes a multi-channel data access mechanism to simultaneously receive two types of core data, as follows: one type includes high-resolution workpiece image data uploaded by the visual recognition module, with a data volume of tens to hundreds of MB per second, containing information on the workpiece's appearance, size, position, and posture; the other type includes real-time data fed back by various sensors, covering force sensor data from the mechanical clamping module, temperature and humidity / light data from the environmental perception module, and position data from the stroke sensor, with data types including analog and digital quantities, and a sampling frequency of 100-1000Hz.

[0019] Preferably, the data processing and analysis module relies on advanced algorithms to perform in-depth analysis of the preprocessed data, accurately extracting key features to support decision-making. The advanced algorithms are applied as follows: a workpiece feature extraction algorithm, for image data, uses a template matching algorithm to quickly locate the workpiece position, and combines it with a contour extraction algorithm to calculate the workpiece length, width, and hole diameter parameters; for workpieces with posture deviations, the Hough transform and minimum bounding rectangle algorithm are used to analyze their rotation angle and tilt, providing a precise basis for posture adjustment for the mechanical clamping module; a deep learning object detection algorithm is used to identify workpiece surface defects; and a multi-data fusion algorithm combines the workpiece position data extracted from the image with sensor feedback. The system integrates the position data of the robotic arm, eliminates mechanical and visual errors through coordinate calibration to improve workpiece positioning accuracy, combines force sensor data and workpiece material information, and uses fuzzy control algorithms to determine whether the clamping force is suitable. It also integrates temperature and humidity data with image clarity parameters to establish a correlation model between environmental parameters and image quality, providing a reference for the visual recognition module to adjust exposure parameters. The dynamic trend analysis algorithm performs time-series analysis on historical data of workpiece size deviation, clamping force fluctuation, and environmental parameter changes, predicts data change trends through moving average and exponential smoothing algorithms, and classifies and statistically analyzes workpiece type and size through clustering algorithms, providing data support for production scheduling.

[0020] Preferably, the environmental perception and compensation module is configured with three key sensors—light, temperature, and humidity—based on the core influencing factors of the industrial environment to ensure the comprehensiveness and accuracy of parameter monitoring. Specific applications are as follows: The light sensor employs industrial-grade high-sensitivity photodiodes and CMOS image sensors to capture real-time changes in ambient light intensity. This provides accurate feedback on light uniformity and intensity parameters to improve the imaging quality of the visual recognition module, preventing overexposure in strong light and underexposure in weak light from affecting image feature extraction. The temperature sensor uses platinum resistance thermometers and thermocouples to monitor the workshop environment temperature. The sensor is placed close to the robotic arm motor and camera lens to collect localized heating temperatures. The humidity sensor uses capacitive humidity sensing elements to monitor real-time changes in workshop air humidity, providing early warnings of the impact of moisture on the insulation performance of electrical equipment, as well as potential risks of oxidation on metal workpiece surfaces and deformation of plastic workpieces.

[0021] Preferably, the automated compensation applications in the environmental perception and compensation module are as follows: Visual recognition module compensation: When the light sensor reports low light intensity, the control module instructs the camera to increase exposure time and aperture; when excessive light causes glare, it instructs the camera to activate backlight compensation and adjust the light source angle to ensure image clarity and feature recognition, avoiding workpiece positioning deviation due to lighting issues; Mechanical clamping module compensation: When the temperature sensor reports rising robot motor temperature, the control module automatically corrects the robot's motion parameters based on a pre-stored thermal deformation compensation model; when humidity is too high, it instructs the gripper to increase clamping force to prevent workpiece slippage due to surface dampness; System overall adaptation adjustment: When humidity consistently exceeds 85% RH, the control module, in addition to adjusting the clamping force, also instructs the equipment's cooling fan to increase its speed to accelerate airflow within the electrical cabinet, preventing moisture condensation from affecting the circuit; when environmental parameters exceed the compensation range, it triggers system load reduction and suspends material feeding until the environment returns to a safe range, ensuring equipment lifespan and production safety.

[0022] Compared with the prior art, the beneficial effects of the present invention are: By leveraging multimodal vision fusion technology and intelligent image recognition algorithms, the system can effectively overcome these interference factors and stably control the positioning accuracy within ±0.5mm, greatly improving the accuracy of the robot arm in recognizing the position and posture of the workpiece and ensuring the precise execution of the gripping action.

[0023] The system uses built-in high-precision pressure and displacement sensors to monitor clamping force and jaw displacement in real time and accurately. It also uses advanced PID control algorithms for intelligent regulation, keeping the clamping force deviation within ±0.2N. It can precisely adjust the clamping force according to the material, shape and size of different workpieces, achieving stable and reliable gripping of workpieces. This effectively avoids workpiece damage and production accidents caused by improper clamping force, and significantly improves the quality and safety of loading operations.

[0024] Sensor monitoring feedback and intelligent control algorithms can monitor the working status of the clamping equipment in real time, promptly detect and correct abnormal clamping force caused by equipment deformation, loosening and other problems; when a slight deformation of the gripper is detected, the system will automatically adjust the control parameters to compensate for the change in clamping force caused by the deformation, ensuring that the clamping force is always within a suitable range, and avoiding production interruption and workpiece loss caused by equipment failure. Attached Figure Description

[0025] Figure 1 This is a block diagram of the automatic feeding control system of the present invention.

[0026] Figure 2 This is a flowchart of the automatic feeding control system of the present invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0028] This application provides an automatic feeding control system based on industrial vision. The automatic feeding control system includes: a vision recognition module, which is equipped with a high-resolution industrial camera and a professional image acquisition card to acquire image information such as the appearance, size, and spatial position of the workpiece, filter out industrial environmental interference, and generate high-definition, high-time-efficiency raw data, providing a reliable input basis for subsequent data processing; a mechanical clamping module, including a high-precision robotic arm and a customized auxiliary gripper, which performs gripping, handling, and placement operations on the workpiece according to control commands, while also having an anti-collision protection mechanism; and a control and calculation module, which uses a high-performance industrial computer or a programmable logic controller to undertake the core control tasks of the system. The system coordinates the collaborative work between various modules and performs precise calculations and decisions based on various feedback information. The data processing and analysis module integrates professional data processing software and advanced algorithms to perform in-depth processing and analysis on image data collected by the visual recognition module and data from other sensors, extracting key feature information to provide decision support for the control and calculation module. The environmental perception and compensation module is equipped with various types of environmental sensors, including light, temperature, and humidity sensors, to monitor changes in environmental parameters such as light, temperature, and humidity in real time. It feeds the data back to the control and calculation module to help the system dynamically adjust equipment operating parameters, offset the impact of environmental interference on operating accuracy, and automatically achieve compensation and adjustment.

[0029] Example 1, please refer to Figures 1-2 The present invention provides the following technical solution: an automatic feeding control system based on industrial vision, comprising an automatic feeding control system based on industrial vision, wherein the automatic feeding control method includes the following steps: S1. The visual recognition module adopts multimodal visual fusion technology, which organically combines a visible light camera and an infrared thermal imager to collect images of the workpiece from all directions. The collected image data is quickly transmitted to the data processing and analysis module.

[0030] In this embodiment, the high-resolution industrial camera organically combines a visible light camera and an infrared thermal imager. The visible light camera can capture rich color, texture, and detail information, and in ideal environments with sufficient light, it can accurately acquire the surface features and geometry of the workpiece, providing basic data for visual recognition. The infrared thermal imager images based on the thermal radiation characteristics of the object itself, and can stably acquire the contour and position information of the workpiece even in low light, when the workpiece surface is dirty, or when the workpiece is partially occluded.

[0031] It should be noted that during the data fusion stage, the system employs an attention-based fusion algorithm. This algorithm dynamically allocates weights by evaluating the importance of features from different modalities, deeply fusing the advantageous features of visible light and infrared images to generate a comprehensive image containing all necessary information. For example, when stains are detected on the workpiece surface causing key features in the visible light image to become blurred, the algorithm automatically increases the weight of the corresponding region's features in the infrared image, thereby ensuring accurate judgment of the workpiece's position and orientation. This significantly enhances the system's visual recognition capabilities in complex environments and effectively overcomes the recognition challenges caused by stains, reflections, occlusions, and changes in lighting.

[0032] S2, the data processing and analysis module performs a series of complex processing steps on the image data, including noise reduction, enhancement, and feature extraction, to identify key information such as the position, posture, and size of the workpiece, and sends this information to the control and calculation module. Based on the received workpiece information, the data processing and analysis module combines preset feeding process parameters and the real-time status of the mechanical clamping module to calculate the motion trajectory and clamping force of the mechanical clamping module through precise algorithms.

[0033] In this embodiment, the high-performance industrial computer is equipped with a multi-core industrial-grade processor, ECC fault-tolerant memory, and a high-speed solid-state drive. It has powerful data parallel computing capabilities and multi-task processing capabilities, and can simultaneously process massive amounts of data from multiple modules such as visual recognition, mechanical control, and environmental compensation. The programmable logic controller adopts an anti-interference design, has a wide operating temperature range and high reliability, supports mixed digital / analog input and output, and realizes logic control through ladder diagrams and structured text industrial programming languages. It is suitable for standardized material feeding scenarios with fixed processes and fast response requirements, and is suitable for long-term continuous production environments.

[0034] In this embodiment, when the control and calculation module receives a collision risk signal, it executes different response strategies according to the risk level. Specifically, these include: Level 1 response: When the laser ranging sensor detects an obstacle but no actual contact occurs, the module immediately adjusts the robot's movement path and avoids the risk without interrupting the loading process by replanning the trajectory in real time; Level 2 response: When the force feedback sensor detects that the force is close to the threshold and the limit switch triggers a warning, the module immediately issues a deceleration command to reduce the robot's movement speed; if the force continues to rise to the threshold and the obstacle is not avoided, an emergency stop is triggered, and all axes of the robot are immediately de-energized and locked to prevent the collision force from increasing further; Level 3 response: In the event of a minor collision, after triggering the emergency stop, the module automatically records the time, location, and force magnitude of the collision, displays an alarm prompt on the system interface, and the robot automatically returns to a safe position, waiting for staff to troubleshoot the fault and restart after confirming safety.

[0035] It should be noted that, at the collaborative control level, the control and computing module establishes a real-time data interaction link with modules such as vision recognition, mechanical clamping, and environmental perception through an industrial communication bus, enabling precise collaboration among the modules. For example, after receiving the workpiece coordinate data uploaded by the vision recognition module, it immediately sends motion path instructions to the mechanical clamping module and simultaneously triggers the environmental perception module to provide feedback on the current temperature, humidity, and light parameters, ensuring the position compensation accuracy of the robotic arm during grasping. In multi-station parallel loading scenarios, the module can dynamically allocate the task priority of each robotic arm to avoid motion conflicts between stations and ensure an orderly and efficient loading process.

[0036] It should be noted that in the computation and decision-making stage, the control computation module performs hierarchical computation and decision-making based on various feedback information (image feature data, robot force sensor data, and environmental parameter data): For control commands with high real-time requirements, such as robot emergency stop and clamping force adjustment, hardware-level computation provides a fast response, ensuring millisecond-level decision latency; for complex scenarios, such as workpiece posture deviation and sudden changes in ambient light, deep computation is performed in conjunction with preset algorithm models (such as deviation compensation algorithms and adaptive control algorithms) to output the optimal adjustment strategy. For example, when the ambient temperature rises and causes robot positioning deviation, the module automatically corrects the robot motion parameters based on temperature sensor data and a pre-stored thermal deformation compensation model to ensure placement accuracy; when the vision recognition module reports workpiece surface defects, the module can immediately decide to pause the feeding process, trigger an alarm, and record defect data, providing a basis for subsequent quality traceability.

[0037] S3. Control commands are transmitted to the mechanical gripping module. The robot arm and gripper move precisely to the workpiece position according to the commands, gripping the workpiece with an appropriate clamping force. During the gripping process, the force sensor built into the mechanical gripping module monitors the clamping force in real time and feeds the data back to the control calculation module. If the clamping force deviates from the preset range, the control calculation module will adjust the control commands in a timely manner to ensure that the clamping force is always at an appropriate level.

[0038] In this embodiment, the high-precision robotic arm adopts a multi-axis linkage structure, equipped with an industrial-grade servo motor and a precision reducer. Its positioning accuracy reaches ±0.01mm, and its repeatability is consistently within ±0.005mm. It can flexibly adjust its posture in three-dimensional space to adapt to the workpiece transfer needs of different workstations. The customized auxiliary gripper is specifically designed according to the workpiece's material, shape, and surface characteristics. Specific gripping methods include: pneumatic grippers, driven by compressed air, with fast response speed, and the gripping force can be flexibly adjusted within the range of 0.1-100N via a pressure regulating valve. It is suitable for medium to heavy regular workpieces of metal and plastic weighing 1-50kg; the electric gripper is driven by a servo motor and has a position control accuracy of ±0.02mm and closed-loop force control function, which adjusts the clamping parameters in real time. It is suitable for electronic components, micro gear precision parts, and thin-walled metal workpieces that are easily deformed, and supports quick switching between multiple types of workpieces; the vacuum suction cup works by using the principle of negative pressure adsorption, without contacting the edge of the workpiece. It is suitable for flat parts with thin metal plates and glass substrates, as well as easily damaged parts such as ceramic sheets and plastic films. There is no mechanical stress in the adsorption process, which effectively protects the integrity of the workpiece surface.

[0039] In this embodiment, the anti-collision protection mechanism is based on the accurate capture of collision risk signals, specifically including: a force feedback sensor, integrated into the end joint of the robot or the connection between the auxiliary gripper and the robot, to detect the axial force, radial force and torque experienced by the robot during its movement in real time. When the robot comes into contact with an obstacle, the force will instantly exceed a preset threshold, and the sensor will immediately convert the force signal into an electrical signal and transmit it to the control and calculation module; a laser ranging / visual obstacle avoidance sensor, with laser ranging sensors installed at key nodes in the robot's movement path and at the end of the robot, which can scan the distance of objects in the surrounding 30-180° range in real time. When the visual detection shows that the distance between the obstacle and the robot is less than the safety threshold, an early warning signal is immediately triggered; and travel limit switches, with mechanical or photoelectric limit switches installed at the movement limit positions of each axis of the robot. When the robot is about to reach its physical limit due to command deviation or malfunction, the switch is triggered, forcibly cutting off the motion drive signal of that axis to avoid overload damage to the mechanical structure.

[0040] In practical implementation, the anti-collision protection mechanism also enhances its applicability and stability, adapting to complex industrial environments through the following design features: Dynamic threshold adjustment: The system can automatically adjust the force feedback and distance thresholds based on the workpiece weight and movement speed. For example, when gripping a 10kg metal workpiece, the force feedback threshold is set to 150N, the movement speed is relatively slow, and the safety distance is set to 30mm; when gripping a 0.1kg electronic component, the threshold is set to 50N, the speed is relatively fast, and the safety distance is set to 10mm, avoiding false triggering or insufficient protection caused by fixed thresholds; Redundant backup design: Key sensors employ dual-path backup. When one sensor fails, the other can still operate normally, preventing the anti-collision mechanism from being paralyzed due to sensor failure. At the same time, the control and calculation module has a built-in independent anti-collision control program, separate from the main loading program, ensuring that the anti-collision function can still operate normally even if the main program malfunctions; Anti-interference processing: Shielded cables are used for sensor signal transmission to avoid signal misjudgment caused by electromagnetic interference in the industrial environment. At the same time, the module filters the sensor data, identifying only continuous and stable risk signals, reducing false triggering caused by factors such as vibration and dust.

[0041] In this embodiment, the data processing and analysis module establishes a multi-channel data access mechanism to simultaneously receive two types of core data, as follows: one type includes high-resolution workpiece image data uploaded by the vision recognition module, with a data volume of tens to hundreds of MB per second, containing information on the workpiece's appearance, size, position, and posture; the other type includes real-time data fed back by various sensors, covering force sensor data from the mechanical clamping module, temperature and humidity / light data from the environmental perception module, and position data from the stroke sensor. The data types include analog and digital quantities, and the sampling frequency reaches 100-1000Hz.

[0042] In this embodiment, the data processing and analysis module relies on advanced algorithms to perform in-depth analysis of the preprocessed data, accurately extracting key features to support decision-making. The advanced algorithms are applied as follows: A workpiece feature extraction algorithm is used for image data, employing a template matching algorithm to quickly locate the workpiece position, combined with a contour extraction algorithm to calculate the workpiece's length, width, and aperture dimensions; for workpieces with posture deviations, Hough transform and minimum bounding rectangle algorithms are used to analyze their rotation angle and tilt, providing a precise basis for the mechanical clamping module to adjust its posture; a deep learning object detection algorithm is used to identify workpiece surface defects; and a multi-data fusion algorithm combines the workpiece position data extracted from the image with sensor data... The system integrates the position data of the robotic arm, eliminates mechanical and visual errors through coordinate calibration to improve workpiece positioning accuracy, combines force sensor data and workpiece material information, and uses fuzzy control algorithms to determine whether the clamping force is suitable. It also integrates temperature and humidity data with image clarity parameters to establish a correlation model between environmental parameters and image quality, providing a reference for the vision recognition module to adjust exposure parameters. The dynamic trend analysis algorithm performs time-series analysis on historical data of workpiece size deviation, clamping force fluctuation, and environmental parameter changes, predicts data change trends through moving average and exponential smoothing algorithms, and classifies and statistically analyzes workpiece type and size through clustering algorithms, providing data support for production scheduling.

[0043] In practice, for the raw data received, the data processing and analysis module performs preprocessing operations using professional data processing software to eliminate interference factors. For image data, Gaussian filtering and median filtering are used to remove dust and light spot noise from the industrial environment. Histogram equalization and illumination compensation are used to correct image deviations caused by uneven lighting. Edge detection (such as the Canny algorithm) is then used to enhance the workpiece contour features, clearing obstacles for subsequent feature extraction. For sensor data, sliding window filtering and Kalman filtering are used to remove instantaneous pulse interference (such as force fluctuations caused by mechanical vibration). Data calibration is used to correct sensor zero-point drift to ensure that the data accuracy meets the analysis requirements.

[0044] It should be noted that the data processing and analysis module transforms the extracted key feature information into structured decision suggestions, which are transmitted to the control and computing module in real time via the industrial bus. Specific outputs include: precise control parameters, such as the workpiece's three-dimensional coordinates (X / Y / Z axis positions) and attitude angles (rotation angle, tilt angle), guiding the robot to adjust its motion trajectory; workpiece dimensional deviation data (such as the difference between actual and standard dimensions), assisting the control module in determining whether the clamping position needs adjustment; environmental compensation parameters, such as requiring a 0.005mm offset in robot positioning for every 1°C increase in temperature, supporting the system's dynamic correction accuracy; abnormal warning signals, immediately outputting an abnormal signal when workpiece surface defects, dimensional deviations, abnormal clamping force, or excessive environmental parameters are detected, triggering the control module to perform operations such as pausing loading and issuing alarms to prevent unqualified workpieces from flowing into the next stage; and optimization suggestion data, outputting equipment optimization suggestions based on historical data analysis, such as checking the vision camera's focal length or adjusting the clamping pressure when the loading success rate at a certain station is below 95%, or replacing the vacuum suction cup model when a certain type of workpiece frequently falls off the clamp, helping the system continuously improve its operating efficiency and stability.

[0045] S4. After gripping the workpiece, the mechanical clamping module moves the workpiece to the designated loading position according to the predetermined motion trajectory to complete the loading operation. The environmental perception and compensation module continuously monitors changes in environmental parameters. Once it detects that parameters such as ambient light or temperature exceed the set threshold, it will send a compensation signal to the control and calculation module. The control and calculation module will adjust the image acquisition parameters of the vision recognition module and the control commands of the mechanical clamping module accordingly based on the compensation signal to eliminate the influence of environmental factors on the system operation and ensure the accuracy and stability of the loading operation.

[0046] In this embodiment, the environmental perception and compensation module is configured with three key sensors—light, temperature, and humidity—based on the core influencing factors of the industrial environment to ensure the comprehensiveness and accuracy of parameter monitoring. Specific applications are as follows: The light sensor employs industrial-grade high-sensitivity photodiodes and CMOS image sensors to capture changes in ambient light intensity in real time. This provides accurate feedback on light uniformity and intensity parameters to improve the imaging quality of the visual recognition module, preventing overexposure in strong light and underexposure in weak light from affecting image feature extraction. The temperature sensor uses platinum resistance thermometers and thermocouples to monitor the workshop environment temperature. The sensor is placed close to the robotic arm motor and camera lens to collect localized heating temperatures. The humidity sensor uses a capacitive humidity sensing element to monitor changes in workshop air humidity in real time, providing early warnings of the impact of moisture on the insulation performance of electrical equipment, as well as potential risks of oxidation on metal workpiece surfaces and deformation of plastic workpieces.

[0047] In this embodiment, the automated compensation application in the environmental perception and compensation module is as follows: Visual recognition module compensation: When the light sensor reports low light intensity, the control module instructs the camera to increase exposure time and aperture. When excessive light causes glare, the module instructs the camera to activate backlight compensation and adjust the light source angle to ensure image clarity and feature recognition, avoiding workpiece positioning deviation due to lighting issues. Mechanical clamping module compensation: When the temperature sensor reports rising robot motor temperature, the control module automatically corrects the robot's motion parameters based on a pre-stored thermal deformation compensation model. When humidity is too high, the module instructs the gripper to increase clamping force to prevent workpiece slippage due to surface dampness. System overall adaptation adjustment: When humidity consistently exceeds 85%RH, the control module, in addition to adjusting clamping force, also instructs the equipment's cooling fan to increase speed to accelerate airflow within the electrical cabinet, preventing moisture condensation from affecting the circuitry. When environmental parameters exceed the compensation range, the system is triggered to reduce load and suspend material feeding until the environment returns to a safe range, ensuring equipment lifespan and production safety.

[0048] It should be noted that the environmental perception and compensation module establishes a real-time data link with the control and computing module through an industrial communication protocol. The data transmission delay is controlled within 10-50ms to ensure that changes in environmental parameters can be quickly captured. Specifically: data acquisition and packaging: the sensor collects parameter data at a preset sampling frequency. After processing by the module's built-in signal conditioning circuit, it is converted into a standard digital signal and then packaged by the data packaging unit into a protocol frame that the control and computing module can recognize, containing information such as parameter type, value, and acquisition timestamp; abnormal data early warning: the module has built-in threshold judgment logic. When the detected parameter exceeds the preset safety range, it will immediately trigger a local audible and visual early warning and send an abnormal alarm frame with higher priority than normal data to the control and computing module to ensure that the system prioritizes environmental risks.

[0049] Example 2: To better understand the above technical solution, the following will provide a detailed explanation of the specific implementation method using actual examples. To provide efficient and accurate solutions for various practical application scenarios, an automatic feeding control system based on industrial vision is proposed. The automatic feeding control system includes: The visual recognition module is equipped with a high-resolution industrial camera and a professional image acquisition card. It is used to acquire image information such as the appearance, size and spatial position of the workpiece, filter out industrial environment interference, and generate high-definition and timely raw data, providing a reliable input basis for subsequent data processing.

[0050] The mechanical clamping module, including a high-precision robotic arm and a customized auxiliary gripper, performs the gripping, handling and placement of workpieces according to control commands, and also has an anti-collision protection mechanism.

[0051] The control and calculation module, which uses a high-performance industrial computer or programmable logic controller, undertakes the core control tasks of the system. It coordinates the collaborative work between various modules and performs precise calculations and decisions based on various feedback information.

[0052] The data processing and analysis module integrates professional data processing software and advanced algorithms to perform in-depth processing and analysis on image data collected by the visual recognition module and data from other sensors, extracting key feature information to provide decision support for the control and computing module.

[0053] The environmental sensing and compensation module is equipped with various types of environmental sensors, including light, temperature, and humidity sensors, to monitor changes in ambient light, temperature, and humidity in real time. The data is fed back to the control and calculation module to help the system dynamically adjust the equipment's operating parameters, offset the impact of environmental interference on operating accuracy, and automatically achieve compensation and adjustment.

[0054] This system innovatively adopts multimodal vision fusion technology, organically combining a visible light camera with an infrared thermal imager. The visible light camera can capture rich color, texture, and detail information, and in ideal environments with sufficient light, it can accurately acquire the surface features and geometry of the workpiece, providing basic data for visual recognition. The infrared thermal imager, on the other hand, images based on the object's own thermal radiation characteristics. Even in low light conditions, when the workpiece surface is dirty or partially obscured, it can stably acquire the outline and position information of the workpiece, effectively making up for the shortcomings of the visible light camera.

[0055] To further improve recognition accuracy and speed, the system introduces a deep learning-based intelligent image recognition algorithm with a convolutional neural network (CNN) as the core architecture. By constructing multiple convolutional and pooling layers, it performs stepwise feature extraction on the input fused image. The convolutional kernels in the convolutional layers can automatically learn local features in the image, such as edges and corners, while the pooling layers downsample the feature map, reducing the amount of data and improving computational efficiency while retaining key features. To enable the model to accurately identify workpieces of different types and postures, a large-scale and diverse workpiece image dataset was used for training. The dataset includes workpiece images in normal conditions, as well as image samples with various complex conditions such as stains, reflections, and varying degrees of occlusion. During training, data augmentation techniques such as random cropping, rotation, and adding noise were used to expand the diversity of the dataset, enabling the model to learn the characteristics of workpieces under various changing conditions and enhance the model's generalization ability. In addition, a transfer learning strategy is introduced. The CNN model of this system is initialized using model parameters pre-trained on a large-scale general image dataset. Based on this, the model is fine-tuned for workpiece recognition tasks in industrial material handling scenarios, so that the model can quickly adapt to specific industrial visual inspection requirements. This greatly reduces training time and data requirements, while improving recognition accuracy and speed, ensuring that the system can quickly and accurately identify the position and posture of workpieces in industrial production environments with high real-time requirements.

[0056] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "connected" and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0057] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An automatic feeding control system based on industrial vision, characterized in that, The automatic feeding control system includes: The visual recognition module is equipped with a high-resolution industrial camera and a professional image acquisition card. It is used to acquire image information such as the appearance, size and spatial position of the workpiece, filter out industrial environment interference, and generate high-definition and timely raw data, providing a reliable input basis for subsequent data processing. The mechanical clamping module, including a high-precision robotic arm and a customized auxiliary gripper, performs the gripping, handling and placement of workpieces according to control commands, and also has an anti-collision protection mechanism. The control and calculation module, which uses a high-performance industrial computer or programmable logic controller, undertakes the core control tasks of the system, coordinates the collaborative work between various modules, and performs precise calculations and decisions based on various feedback information. The data processing and analysis module integrates professional data processing software and advanced algorithms to perform in-depth processing and analysis on the image data collected by the visual recognition module and the data fed back by other sensors, extract key feature information, and provide decision support for the control and computing module. The environmental sensing and compensation module is equipped with various types of environmental sensors, including light, temperature, and humidity sensors, to monitor changes in ambient light, temperature, and humidity in real time. The data is fed back to the control and calculation module to help the system dynamically adjust the equipment's operating parameters, offset the impact of environmental interference on operating accuracy, and automatically achieve compensation and adjustment.

2. The automatic feeding control system based on industrial vision according to claim 1, characterized in that: The high-resolution industrial camera organically combines a visible light camera with an infrared thermal imager; Among them, visible light cameras can capture rich color, texture and detail information. In ideal environments with sufficient light, they can accurately acquire the surface features and geometry of workpieces, providing basic data for visual recognition. Infrared thermal imagers image based on the thermal radiation characteristics of objects themselves. They can stably acquire the contour and position information of workpieces even in low light, when the workpiece surface is dirty, or when the workpiece is partially obscured.

3. The automatic feeding control system based on industrial vision according to claim 1, characterized in that: The high-precision robotic arm adopts a multi-axis linkage structure, equipped with an industrial-grade servo motor and a precision reducer. Its positioning accuracy can reach ±0.01mm, and its repeatability is stable within ±0.005mm. It can flexibly adjust its posture in three-dimensional space to adapt to the workpiece transfer needs of different workstations. Customized auxiliary grippers are specifically designed based on the workpiece's material, shape, and surface characteristics. Specific gripping methods include: The pneumatic gripper, driven by compressed air, has a fast response speed. The clamping force can be flexibly adjusted in the range of 0.1-100N through the pressure regulating valve, and is suitable for medium to heavy regular workpieces of metal and plastic weighing 1-50kg. The electric gripper, driven by a servo motor, has a position control accuracy of ±0.02mm and a closed-loop force control function. It can adjust the gripping parameters in real time and is suitable for electronic components, micro gear precision parts, as well as thin-walled metal workpieces that are easily deformed. It supports rapid switching between multiple types of workpieces. Vacuum suction cups work on the principle of negative pressure adsorption, without needing to contact the edge of the workpiece. They are suitable for flat parts with thin metal plates and glass substrates, as well as easily damaged parts such as ceramic sheets and plastic films. The adsorption process is stress-free, effectively protecting the integrity of the workpiece surface.

4. The automatic feeding control system based on industrial vision according to claim 1, characterized in that: The collision avoidance protection mechanism is based on the accurate capture of collision risk signals, specifically including: Force feedback sensors are integrated into the end joint of the robot or the connection between the auxiliary gripper and the robot. They detect the axial force, radial force and torque that the robot experiences during its movement in real time. When the robot comes into contact with an obstacle, the force will instantly exceed the preset threshold. The sensor immediately converts the force signal into an electrical signal and transmits it to the control and computing module. Laser ranging / visual obstacle avoidance sensors are installed at key nodes in the robot's movement path and at the end of the robot. They can scan the distance of objects in the surrounding 30-180° range in real time. When the robot detects that the distance between an obstacle and the robot is less than the safety threshold, an early warning signal is immediately triggered. Travel limit switches are mechanical or photoelectric limit switches installed at the extreme positions of each axis of the robot. When the robot is about to reach its physical limit due to command deviation or malfunction, the switch is triggered, forcibly cutting off the motion drive signal of that axis to avoid overload damage to the mechanical structure.

5. The automatic feeding control system based on industrial vision according to claim 1, characterized in that: The high-performance industrial computer is equipped with a multi-core industrial-grade processor, ECC fault-tolerant memory and high-speed solid-state drive, and has powerful data parallel computing and multi-task processing capabilities. It can simultaneously process massive amounts of data from multiple modules such as visual recognition, mechanical control and environmental compensation. The programmable logic controller (PLC) features an anti-interference design, a wide operating temperature range, and high reliability. It supports mixed digital / analog input / output and achieves logic control through ladder diagrams and structured text industrial programming languages. It is suitable for standardized material feeding scenarios with fixed processes and fast response requirements, and is applicable to long-term continuous production environments.

6. The automatic feeding control system based on industrial vision according to claim 1, characterized in that: When the control and calculation module receives a collision risk signal, it executes different response strategies according to the risk level, specifically including: In Level 1 Response, when the laser rangefinder detects an obstacle but does not make actual contact, the module immediately adjusts the robot's movement path and avoids risks without interrupting the loading process by replanning the trajectory in real time. In the second-level response, when the force feedback sensor detects that the force is close to the threshold and the limit switch triggers an early warning, the module immediately issues a deceleration command to reduce the movement speed of the robot arm; if the force continues to rise to the threshold and the obstacle is not avoided, an emergency stop is triggered, and all axes of the robot arm are immediately de-energized and locked to prevent the collision force from increasing further. In a Level 3 response scenario, upon the occurrence of a minor collision, the module will automatically record the time, location, and force of the collision after triggering an emergency stop. An alarm will then pop up on the system interface, and the robotic arm will automatically retract to a safe position, awaiting personnel to troubleshoot the fault and confirm safety before restarting.

7. The automatic feeding control system based on industrial vision according to claim 1, characterized in that: The data processing and analysis module establishes a multi-channel data access mechanism to simultaneously receive two types of core data, as detailed below: One type includes high-resolution workpiece image data uploaded by the visual recognition module, with a data volume of tens to hundreds of MB per second, containing information on the workpiece's appearance, size, position, and posture; Another category includes real-time data from various sensors, covering force sensor data from the mechanical clamping module, temperature / humidity / light data from the environmental sensing module, and position data from the travel sensor. The data types include analog and digital quantities, with sampling frequencies ranging from 100 to 1000 Hz.

8. The automatic feeding control system based on industrial vision according to claim 1, characterized in that: The data processing and analysis module relies on advanced algorithms to perform in-depth analysis of the preprocessed data and accurately extract key features that support decision-making. The advanced algorithms are applied as follows: The workpiece feature extraction algorithm, for image data, uses a template matching algorithm to quickly locate the workpiece position, and combines a contour extraction algorithm to calculate the workpiece length, width, and hole diameter parameters; for workpieces with posture deviation, the Hough transform and minimum bounding rectangle algorithm are used to analyze their rotation angle and tilt, providing a precise posture adjustment basis for the mechanical clamping module; and a deep learning target detection algorithm is used to identify workpiece surface defects. The multi-data fusion algorithm integrates the workpiece position data extracted from the image with the robot position data fed back by the sensor. It eliminates mechanical and visual errors through coordinate calibration, thereby improving the workpiece positioning accuracy. It combines force sensor data and workpiece material information and uses a fuzzy control algorithm to determine whether the clamping force is suitable. It also integrates temperature and humidity data with image clarity parameters to establish a correlation model between environmental parameters and image quality, providing a reference for the visual recognition module to adjust exposure parameters. The dynamic trend analysis algorithm performs time-series analysis on historical data of workpiece size deviation, clamping force fluctuation, and environmental parameter changes. It predicts data change trends through moving average and exponential smoothing algorithms, and classifies and statistically analyzes workpiece types and sizes through clustering algorithms, providing data support for production scheduling.

9. The automatic feeding control system based on industrial vision according to claim 1, characterized in that: The environmental sensing and compensation module is equipped with three key sensors—light, temperature, and humidity—based on the core influencing factors of the industrial environment to ensure comprehensive and accurate parameter monitoring. Specific applications are as follows: The light sensor uses an industrial-grade high-sensitivity photodiode and CMOS image sensor to capture changes in ambient light intensity in real time. It accurately feeds back the uniformity and intensity parameters of light to improve the imaging quality of the visual recognition module, avoiding overexposure in strong light and underexposure in weak light from affecting image feature extraction. Temperature sensors, using platinum resistance and thermocouple sensors, are used to monitor the ambient temperature in the workshop. The sensors are placed close to the robot motor and camera lens to collect localized heating temperatures. The humidity sensor uses a capacitive humidity sensing element to monitor changes in workshop air humidity in real time, provide early warning of the impact of moisture on the insulation performance of electrical equipment, and warn of potential risks such as oxidation of metal workpieces and deformation of plastic workpieces.

10. An automatic feeding control system based on industrial vision according to claim 1, characterized in that: The automated compensation application in the environmental perception and compensation module is as follows: The visual recognition module compensates for low light intensity by instructing the camera to increase exposure time and aperture when the light sensor indicates low light intensity. Conversely, if the light intensity is too strong and causes reflection, the module instructs the camera to activate backlight compensation and adjust the light source angle to ensure image clarity and feature recognition, thus avoiding workpiece positioning deviations due to lighting issues. The mechanical clamping module compensates for the temperature sensor feedback when the temperature of the robot arm motor rises. The control module automatically corrects the robot arm motion parameters based on the pre-stored thermal deformation compensation model. When the humidity is too high, the clamping device is instructed to increase the clamping force to prevent the workpiece from slipping and falling off due to the wet surface. The system is adapted and adjusted as a whole. When the humidity is consistently higher than 85% RH, the control module, in addition to adjusting the clamping force, also instructs the equipment cooling fan to increase its speed to accelerate air circulation in the electrical cabinet, preventing water vapor condensation from affecting the circuit. When the environmental parameters exceed the compensation range, the system is triggered to reduce the load and suspend feeding until the environment returns to a safe range, ensuring the life of the equipment and production safety.

Images