Self-adaptive visual identification and correction method for intelligent industrial robot

Abstract

The invention discloses a self-adaptive visual recognition and correction method for an intelligent industrial robot, and relates to the technical field of industrial automation and intelligent control, and the method comprises the steps: S1, regulating and controlling the frame rate of a visual collection module based on a real-time estimation result of the motion speed of an object on a production line, the method comprises the steps of S1, collecting image data in the object movement process and generating initial positioning information, S3, analyzing the deviation between the initial positioning information and a time delay parameter and generating a position correction instruction, and S4, adjusting a mechanical arm to act according to the position correction instruction so as to complete high-precision positioning operation. According to the self-adaptive visual recognition and correction method for the intelligent industrial robot, the visual collection frame rate is regulated and controlled according to the moving speed of an object on a production line, so that the problem that the positioning precision is reduced due to time delay is solved.

Description

Technical Field

[0001] The present invention relates to the field of industrial automation and intelligent control technologies, and particularly relates to an adaptive vision recognition and correction method for intelligent industrial robots. Background Art

[0002] In existing industrial automation production lines, intelligent industrial robots are widely used in tasks such as object recognition, grasping, and assembly. Among them, the vision recognition system is a key component for realizing the precise operation of industrial robots. The existing industrial robot vision systems usually adopt a fixed frame rate for image acquisition, and detect and locate objects on the production line through image processing algorithms.

[0003] However, due to the dynamic changes in the movement speeds of objects on the production line, it is difficult for a vision system with a fixed acquisition frame rate to adapt to the visual information acquisition requirements at different speeds in a timely manner. When the object speed is relatively fast, the image acquisition delay and processing time delay will increase significantly, resulting in inaccurate position information obtained by the industrial robot, thus affecting the positioning accuracy of the operation and the overall production efficiency. At present, some vision systems attempt to reduce the response time delay by improving the hardware performance or optimizing the image processing algorithm, but there are still the following deficiencies: on the one hand, the system cost increase brought by hardware upgrading is significant, and it cannot adapt to the acquisition requirements in different speed scenarios in real time; on the other hand, although algorithm optimization can reduce the processing time delay to a certain extent, due to the constant acquisition frame rate, the system still cannot dynamically adjust to cope with high-speed moving objects, ultimately resulting in a decrease in the robot's positioning accuracy and an increase in errors. Especially in complex working conditions or high-speed production lines, this delay problem is more prominent. Summary of the Invention

[0004] The purpose of the present invention is to provide an adaptive vision recognition and correction method for intelligent industrial robots, which regulates the vision acquisition frame rate according to the movement speed of objects on the production line to solve the problem of decreased positioning accuracy caused by time delay.

[0005] To achieve the above purpose, the present invention provides the following technical solution: an adaptive vision recognition and correction method for intelligent industrial robots, the method comprising:

[0006] S1. Regulate the frame rate of the vision acquisition module based on the real-time estimation result of the movement speed of objects on the production line;

[0007] S2. Collect image data during the movement of the object and generate initial positioning information, including selecting a reference object with a known size in the image, measuring the pixel distance of the reference object in the image, measuring the pixel distance of the target object in the image, calculating the physical distance of the target object in the actual space, and using the calculated actual distance of the target object as the initial positioning information of the robot system;

[0008] S3. Analyze the deviation between the initial positioning information and the delay parameter, and generate a position correction instruction, including determining the basic processing time generated according to different amounts of image data, analyzing the delay deviation of the initial positioning information, calculating the total delay value under the current image data, and generating a position correction instruction based on the total delay value under the current image data to ensure the operation accuracy of the robot manipulator;

[0009] S4. Adjust the manipulator movement according to the position correction instruction to complete the high-precision positioning operation, including determining the action point and reaction force of the target according to the position of the target object and the correction instruction, determining the acting force that the manipulator needs to apply, and controlling the manipulator to make adjustments according to the calculation result.

[0010] Preferably, S1 includes determining the change rate of the movement speed of the target object on the production line, and calculating the image sampling frequency of the visual acquisition module as time goes by.

[0011] Preferably, the specific steps for determining the change rate of the movement speed of the target object on the production line in S1 include: detecting and recording the speed of the moving object on the production line at the current moment in real time through a visual recognition system or a sensor; retrieving the speed of the object at the previous moment from the historical data of the system; confirming the time interval between the current moment and the previous moment; subtracting the speed value at the previous moment from the speed value at the current moment to obtain the speed change amount of the object during this period; dividing the speed change amount by the time interval between the two moments before and after to obtain the speed change rate per unit time.

[0012] Preferably, the specific steps for calculating the image sampling frequency of the visual acquisition module in S1 include determining the time elapsed from the start of sampling to the current moment according to the obtained speed change rate per unit time, multiplying the speed growth rate by the current estimated time to obtain the sampling frequency of the visual acquisition module, and using the calculated sampling frame rate as a control parameter to adjust the sampling speed of the visual acquisition module in real time to ensure that the image acquisition is synchronized with the movement speed of the object.

[0013] Preferably, the specific steps for calculating the physical distance of the target object in the actual space in S2 include first taking the actual distance value of the reference object in reality; then calculating the ratio of the measured distance of the target object in the image to the measured distance of the reference object in the image; and finally multiplying the actual distance value of the reference object by the above ratio to obtain the physical distance of the target object in the actual space.

[0014] Preferably, the specific steps of calculating the total time delay value under the current image data in S3 include: evaluating the complexity of the current image according to the processed image data, and quantitatively analyzing the complexity of the current image based on factors such as the number of target objects, background complexity, and lighting conditions; determining the basic processing time under ideal conditions, where the basic processing time is the shortest time required to complete image acquisition, recognition, and output of positioning information under the simplest image data conditions; determining the delay increment caused by the increase in image complexity according to the value of the image complexity, and the delay increment caused by the increase in image complexity is the product of the image complexity and the delay increment per unit complexity; adding the basic processing time and the delay increment caused by the increase in image complexity to obtain the total time delay value under the current image data; using the calculated total time delay value as a time delay compensation parameter, and adjusting the positioning information and motion control instructions according to the compensation parameter.

[0015] Preferably, the specific steps of determining the force that the robotic arm needs to apply in S4 include: determining the reverse force generated by the target object on the robotic arm; measuring the actual distance between the force application point of the target object and its support point; measuring the actual distance between the position where the robotic arm applies the force and its fixed support point; multiplying the reverse force generated by the target object by the acting distance of the object to obtain the torque generated by the target object; dividing the torque of the target object by the acting distance of the robotic arm to obtain the force that the robotic arm needs to apply; adjusting the output of the robotic arm according to the force calculated above.

[0016] Preferably, S1 further includes estimating the frame rate of the visual acquisition module. Specifically, according to a preset proportionality coefficient, a ratio operation is performed on the maximum movement speed of the object on the production line and the actual movement speed of the object currently measured to calculate the frame rate; then, the initial position of the object on the production line is detected by a sensor; next, the measured object speed and the current position information are input into the control module to calculate the adjustment range of the frame rate; if the adjustment range exceeds the preset threshold, a real-time frame rate control mechanism is triggered to dynamically adjust the frame rate of the visual acquisition module; if the adjustment range does not exceed the threshold, the original frame rate remains unchanged.

[0017] Preferably, S1 further includes first using the Kalman filtering method to estimate the acceleration of the object on the production line; then, updating the parameters in the object motion speed model according to the estimated acceleration value; next, introducing an acceleration change constraint condition to limit the difference between the updated acceleration value and the acceleration value at the previous moment; finally, calculating the average motion speed within the target area in combination with the updated acceleration value, and setting the frame rate of the visual acquisition module according to the average speed.

[0018] Preferably, in S1, according to the estimated acceleration value, the parameters in the object motion speed model are updated, including calculating a new acceleration value by weighted calculation of the acceleration value at the previous moment and the currently measured object motion speed. When performing weighted calculation, a smoothing factor is used, and the value range of the smoothing factor is 70%-95%.

[0019] As can be seen from the above technical solutions, the present invention has the following beneficial effects:

[0020] The adaptive vision recognition and correction method for intelligent industrial robots regulates the frame rate of the vision acquisition module based on the real-time estimation result of the object motion speed on the production line, acquires image data during the object motion process and generates initial positioning information, analyzes the deviation between the initial positioning information and the delay parameter, generates a position correction instruction, and adjusts the manipulator action according to the position correction instruction to complete high-precision positioning operations. It breaks through the technical limitation of the fixed acquisition frame rate of the existing industrial robot vision system and can regulate the image sampling frequency of the vision acquisition module in real time according to the dynamic change of the object motion speed on the production line. Through this method, the synchronous response between the vision acquisition system and the object motion state is realized, the real-time performance of image acquisition and the accuracy of data processing are greatly improved, the positioning accuracy and operation stability of the robot system are significantly improved, the problems of image acquisition delay and processing lag caused by the high-speed motion of the target object are effectively solved, and it is ensured that the industrial robot can continuously obtain clear and accurate image information under different speed conditions. Compared with the prior art that relies on fixed frame rates and high-performance hardware, the present invention does not require a significant increase in hardware costs, reduces system upgrade and operation and maintenance costs, has higher economy and practicability, improves the operation accuracy and reaction speed of industrial robots in complex environments, effectively suppresses the positioning errors caused by image acquisition and processing delays, significantly reduces the operation deviation rate, and improves the application performance of the robot system on high-speed dynamic production lines. It can adapt to different types of industrial production lines and various working scenarios. Its dynamic adjustment strategy not only improves production efficiency but also enhances the system's adaptability to complex working conditions and multi-target operation environments, significantly improving the automation operation level and market competitiveness of intelligent industrial robots. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] As shown Figure 1 in the figure, the present invention provides a technical solution: an adaptive vision recognition and correction method for an intelligent industrial robot, the method comprising:

[0024] S1. Adjust the frame rate of the vision acquisition module based on the real-time estimation result of the moving speed of the object on the production line;

[0025] S2. Collect the image data during the movement of the object and generate initial positioning information, including selecting a reference object with a known size in the image, measuring the pixel distance of the reference object in the image, measuring the pixel distance of the target object in the image, calculating the physical distance of the target object in the actual space, and using the calculated actual distance of the target object as the initial positioning information of the robot system;

[0026] S3. Analyze the deviation between the initial positioning information and the delay parameter, and generate a position correction instruction, including determining the basic processing time generated according to different amounts of image data, analyzing the delay deviation of the initial positioning information, calculating the total delay value under the current image data, and generating a position correction instruction according to the total delay value under the current image data to ensure the operation accuracy of the robot manipulator;

[0027] S4. Adjust the actions of the manipulator according to the position correction instruction to complete the high-precision positioning operation, including determining the action point and reaction force of the target according to the position of the target object and the correction instruction, determining the acting force that the manipulator needs to apply, and controlling the manipulator to make adjustments according to the calculation result.

[0028] The core of this embodiment lies in achieving high-precision recognition and real-time positioning control of target objects in a dynamic environment through adaptive visual recognition and correction technology. First, in step S1, the motion speed of objects on the production line is estimated in real time. The system obtains the motion parameters of objects in real time through speed sensors installed on the production line or through image sequence analysis technology. According to the detected motion speed, the frame rate of the visual acquisition module is dynamically adjusted to ensure that clear and effective image data can still be obtained under different speed conditions. For example, when the object's motion speed is high, the acquisition frame rate is increased to avoid image blurring and ensure the integrity and timeliness of image information. In step S2, the visual acquisition module obtains continuous image data and identifies the set reference object in the image through image processing algorithms. The reference object has a known physical size. By measuring its pixel distance in the image, the system can calculate the proportional coefficient between pixels and the actual distance. Subsequently, the pixel size of the target object in the image is detected and measured, and based on the above proportional coefficient, its position coordinates and size information in the actual space are calculated. This calculation process uses a geometric correction model based on computer vision to ensure the accuracy and consistency of the target object positioning data. In step S3, the system conducts comprehensive analysis based on the initial positioning information, combined with the time delay parameters introduced during image acquisition, processing, and transmission. By statistically analyzing the basic processing time caused by different amounts of image data, the system establishes a dynamic time delay model to calculate the total time delay value of the currently acquired image in real time. This time delay value includes image acquisition delay, processing delay, and network transmission delay. The actual position of the target object is predicted and corrected using the time delay value to generate accurate correction instructions. The correction instructions are sent to the robot manipulator execution unit through the control system. In step S4, the robot control system adjusts the actions of the manipulator according to the correction instructions. First, the position of the action point of the target object is determined, and the reaction force parameters during the operation are analyzed based on the physical model. The system further calculates the magnitude and direction of the force that the manipulator needs to apply according to the material characteristics, shape, and force application requirements of the target object. The control unit drives the manipulator to perform fine adjustment operations based on the calculation results. This adjustment process can achieve real-time feedback correction through closed-loop control to ensure the force control accuracy and dynamic response ability during the operation, and improve the stability and flexibility of the manipulator in a complex operating environment. In addition, to further improve the robustness and adaptability of the system, this embodiment also introduces multi-sensor information fusion technology. By combining the data of force sensors, visual sensors, and position sensors, comprehensive perception and dynamic adjustment of the target object state are achieved. The overall architecture design ensures the visual recognition accuracy and mechanical operation accuracy of the robot system on a high-speed dynamic production line, and improves the intelligent operation ability of industrial robots under different complex working conditions.

[0029] In this embodiment, by estimating the real-time motion speed of objects on the production line and dynamically regulating the frame rate of the vision acquisition module, the accuracy and timeliness of image acquisition are improved. A pixel ratio model is established using a reference object to achieve high-precision positioning of the target object in the actual space. Through comprehensive analysis of the processing delays of different data volumes and automatic correction of delay deviations, the positioning error caused by system delay is significantly reduced, and the operation accuracy and stability of the industrial robot manipulator are remarkably improved. In addition, based on the analysis of the action point and reaction force, precise control of the manipulator's movements is achieved, further enhancing the adaptability and robustness of the robot system and improving the automation level and application breadth of intelligent industrial robots.

[0030] S1 includes determining the rate of change of the motion speed of the target object on the production line and calculating the image sampling frequency of the vision acquisition module as time progresses. In this embodiment, based on Claim 1, the frame rate regulation mechanism of the vision acquisition module in step S1 is further optimized. Specifically, by detecting and analyzing the rate of change of the motion speed of the target object on the production line, that is, the change amplitude of the target object's speed per unit time, the image sampling frequency of the vision acquisition module is dynamically calculated. The system obtains the rate of change of the target object's speed through speed sensors installed on the production line or in the robot's working area, or by using vision recognition algorithms to process consecutive image frames. The system predicts the motion trend of the target object at future moments based on the real-time obtained rate of change of speed data. During the progression in the time dimension, the calculation of the sampling frequency uses a dynamic adjustment algorithm, such as a predictive control theory-based or adaptive scheduling algorithm, enabling the vision acquisition module to automatically adjust the sampling frequency according to different speed change situations. When it is detected that the rate of change of the target object's speed is large, that is, the object's motion state is unstable or there are frequent accelerations and decelerations, the system increases the image sampling frequency to ensure that more real-time image information is collected and reduce the recognition error caused by information loss or blurring; conversely, when the rate of change of speed is small and the object's motion state is stable, the image sampling frequency is appropriately reduced to save system resources and reduce the data processing pressure. This adaptive regulation method enhances the system's response ability to the dynamic motion of the target object and improves the efficiency and accuracy of overall image processing.

[0031] By introducing the real-time detection and analysis of the rate of change of the target object's motion speed, this embodiment can more accurately reflect the dynamic motion state of the target object and avoid problems of insufficient or redundant information acquisition caused by a fixed sampling frequency. The adaptive adjustment of the image sampling frequency not only improves the system's ability to recognize and position the target in a high-speed dynamic environment but also effectively reduces the computational load and energy consumption during image acquisition and processing, optimizes the system resource allocation, and further improves the working efficiency and stability of the robot system. In addition, this method can also effectively improve the robot's adaptability to changes in the production line, broadening its application scenarios and practical value.

[0032] The specific steps for determining the change rate of the motion speed of the target object on the production line in S1 include: detecting and recording the speed of the moving object on the production line at the current moment in real time through a vision recognition system or a sensor; retrieving the motion speed of the object at the previous moment from the historical data of the system; confirming the time interval between the current moment and the previous moment; subtracting the speed value at the previous moment from the speed value at the current moment to obtain the speed change amount of the object during this period; dividing the speed change amount by the time interval between the two moments to obtain the speed change rate per unit time.

[0033] The specific formula for determining the change rate of the motion speed of the target object on the production line is:

[0034] where r represents the change rate of the motion speed of the target object, v t represents the motion speed of the object at the current moment, v t-Δt represents the motion speed of the object at the previous moment, and Δt represents the time interval between the current moment and the previous moment.

[0035] This embodiment further refines the process of determining the change rate of the motion speed of the target object on the production line, ensuring the accuracy and real-time performance of the sampling frequency adjustment. First, a vision recognition system or an external sensor installed on the production line (such as a laser velocimeter, an infrared sensor, or a millimeter-wave radar) continuously detects the moving target object on the production line and records the motion speed data at the current moment in real time. This speed data can be obtained through continuous frame image analysis (such as inter-frame displacement calculation) or sensor data output and stored in the system database. Subsequently, the system automatically retrieves the historical speed data of the target object at the previous moment from the database to form a pair of speed sampling values at adjacent moments. The system confirms the time interval between the current moment and the previous moment, that is, the time sampling interval. By calculating the difference between the speed value at the current moment and the speed value at the previous moment, the speed change amount of the target object in this time interval is obtained. The system then divides this change amount by the time interval to calculate the instantaneous speed change rate of the target object. This real-time calculation process runs automatically in the system control unit at a fixed sampling period to ensure the acquisition of continuous speed change rate data. The system dynamically adjusts the image sampling frequency of the vision acquisition module according to this data. When it detects a significant increase in the speed change rate, the system increases the image sampling frequency to capture more detailed motion state information; when the speed change rate is small, it appropriately reduces the sampling frequency to optimize the use of computing resources. In this way, the system realizes efficient vision acquisition control based on the dynamic behavior of the target object, ensuring the accuracy of subsequent positioning and operation.

[0036] This embodiment realizes the fine-grained dynamic analysis of the motion state of the target object by introducing a calculation method of the speed change rate based on historical data comparison. Compared with the traditional single-moment speed sampling method, this method can more accurately identify the acceleration, deceleration or uniform speed state of the object, so as to dynamically optimize the image sampling frequency according to actual needs. It improves the ability of the visual acquisition module to respond to the system dynamics, and significantly enhances the adaptability of the robot system to high-speed production lines and complex environments. This solution effectively improves the accuracy of target recognition and positioning, reduces the error rate, and optimizes the overall resource allocation and energy consumption control of the system. In addition, this method has good scalability and versatility, is applicable to various industrial production line environments, and is of great significance for improving the level of production automation.

[0037] The specific steps of calculating the image sampling frequency of the visual acquisition module in S1 include determining the time elapsed from the start of sampling to the current moment according to the obtained speed change rate per unit time, multiplying the speed growth rate by the current estimated time to obtain the sampling frequency of the visual acquisition module, and using the calculated sampling frame rate as a control parameter to adjust the sampling speed of the visual acquisition module in real time to ensure that the image acquisition is synchronized with the motion speed of the object.

[0038] The specific formula for calculating the image sampling frequency of the visual acquisition module is: A = r × t;

[0039] Among them, A represents the image sampling frequency of the visual acquisition module, r represents the motion speed change rate of the target object, and t represents time.

[0040] On the basis of the foregoing, this embodiment further refines the calculation method of the image sampling frequency of the visual acquisition module. First, the system obtains the rate of change of the motion speed of the target object on the production line through the visual recognition system or external sensors, that is, the rate of increase or decrease of the speed per unit time. This rate of change of speed is obtained by continuously sampling the speed values of the target object at different time points based on the difference algorithm or a more complex prediction model. Subsequently, the system records in real time the time value elapsed from the start of image sampling to the current moment during the acquisition process, that is, the cumulative sampling time T. Multiply the currently obtained rate of change of speed by the time T to obtain the dynamic sampling frequency at the current moment. This sampling frequency value characterizes the optimal sampling speed required by the image acquisition system under the changing motion state of the target object. The sampling frequency is passed as a dynamic control parameter to the visual acquisition module, and the control system adjusts its sampling frame rate in real time. The control system automatically increases or decreases the image sampling rate of the visual acquisition module according to the calculation result, so as to ensure that the time resolution of image acquisition can adapt to the real-time motion state of the target object. When the target object accelerates, the system increases the sampling frame rate and reduces the image sampling time interval to prevent image blurring or positioning errors of the target object due to rapid movement; when the motion of the target object tends to be stable or decelerates, the system reduces the sampling frequency to optimize the system resource utilization efficiency. In addition, to ensure that image acquisition is synchronized with the motion speed of the target object, the system also introduces a closed-loop feedback mechanism to dynamically optimize the sampling frequency according to the feedback of the acquired image quality and motion state, so that the visual acquisition is always in the optimal state, improving the overall recognition accuracy and data processing efficiency.

[0041] By dynamically calculating the sampling frequency based on the rate of change of speed and the cumulative time, this embodiment realizes the synchronous control of the visual acquisition module and the motion state of the target object. Compared with the traditional fixed sampling method, it can significantly improve the system response sensitivity and adaptability, and ensure the integrity and clarity of image acquisition of the target object under different motion states. This method optimizes the sampling control logic, reduces image acquisition delay and processing redundancy, and improves the overall performance and working efficiency of the system. Especially in high-speed production lines or industrial scenarios with frequent changes in motion states, this embodiment effectively enhances the processing ability of the industrial robot system for complex dynamic environments, and significantly improves the accuracy and stability of automated operations.

[0042] The specific steps for calculating the physical distance of the target object in the actual space in S2 include: First, obtain the actual distance value of the reference object in reality; then, calculate the ratio of the measured distance of the target object in the image to the measured distance of the reference object in the image; finally, multiply the actual distance value of the reference object by the above ratio to obtain the physical distance of the target object in the actual space.

[0043] The specific formula for calculating the physical distance of the target object in the actual space is:

[0044] Among them, B1 represents the actual distance of the reference object, B2 represents the physical distance of the target object in the actual space, c1 represents the measured distance of the reference object in the image, and c2 represents the measured distance of the target object in the image.

[0045] This embodiment details the calculation method of the physical distance of the target object in the actual space in step S2, aiming to improve the robot system's ability to accurately obtain the position of the target object. First, when collecting images, the system needs to introduce a reference object with a known size. This reference object can be a standard calibration device with clear geometric parameters, such as a standard length scale, a calibration block with a fixed size, or a cylinder with a known diameter. By pre-entering the actual distance value of this reference object in the actual space into the system, a basic mapping relationship between the real space and the image pixel coordinates is established. Secondly, the vision acquisition module detects the contour boundaries of the reference object and the target object in the image through image processing algorithms, and measures the pixel distances of the two in the image plane. The measured distance of the target object in the image and the measured distance of the reference object in the image are obtained. The system calculates the ratio of the two, and this ratio reflects the size proportional relationship of the target object relative to the reference object. Finally, multiply the actual distance value of the known reference object by the above ratio to obtain the physical distance of the target object in the real space, that is. This calculation process simplifies the complex three-dimensional reconstruction or calibration process, and through direct conversion based on relative proportions, quickly realizes the determination of the physical distance of the target object. The obtained distance value is an important part of the initial positioning information of the robot system, providing accurate spatial data support for subsequent position correction and robotic arm motion control.

[0046] By adopting the method of proportional conversion based on the reference object, this embodiment significantly improves the simplicity and practicality of calculating the actual distance of the target object. Compared with the traditional ranging methods based on complex camera internal parameter calibration or three-dimensional point cloud reconstruction, this method is more efficient and easier to implement, especially suitable for the rapid deployment and instant calculation requirements in dynamic production line scenarios. This method does not require frequent calibration of the camera, reducing the system maintenance cost and complexity. At the same time, by using a known reference object to achieve accurate conversion, the accuracy of the spatial positioning of the target object is improved, thereby enhancing the accuracy and reliability of the robot system in operations such as grasping, assembly, and detection. This method has good stability and applicability and can be widely applied to different industrial production environments.

[0047] The specific steps for calculating the total delay value under the current image data in S3 are as follows: Based on the processed image data, evaluate the complexity of the current image. The complexity of the current image is quantitatively analyzed according to factors such as the number of target objects, background complexity, and lighting conditions; Determine the basic processing time under ideal conditions. The basic processing time is the shortest time required to complete image acquisition, recognition, and output of positioning information under the simplest image data conditions; According to the value of the image complexity, determine the delay increment caused by the increase in image complexity. The delay increment caused by the increase in image complexity is the product of the image complexity and the delay increment per unit complexity; Add the basic processing time to the delay increment caused by the increase in image complexity to obtain the total delay value under the current image data; Use the calculated total delay value as a delay compensation parameter, and adjust the positioning information and motion control instructions according to the compensation parameter.

[0048] The specific formula for calculating the total delay value under the current image data is: D = g + h × E;

[0049] Where, D represents the total delay value under the current image data, g represents the basic processing time, h represents the delay increment per unit complexity, and E represents the complexity of the current image data.

[0050] This embodiment aims to improve the time-delay compensation ability of the vision recognition system of intelligent industrial robots. By dynamically calculating the total time-delay value of image processing, it ensures the real-time performance and accuracy of robot positioning and operation. First, the system analyzes the complexity of each piece of acquired image data in step S3. This complexity assessment comprehensively considers the number of target objects, background complexity, and current lighting conditions. The more target objects there are, the longer it takes to analyze the image; when the background complexity is high, the algorithm needs to perform more feature extraction and denoising processes; when there are large differences in lighting conditions, the system needs to increase image enhancement or dynamic exposure compensation operations, thus increasing the overall processing load. The system quantifies the image complexity factors through a preset parameter model or a machine learning-based image feature evaluation model to form a complexity assessment value. Secondly, the system sets the basic processing time under ideal conditions based on the experimental results in the test calibration stage or under ideal conditions. This basic processing time is the shortest processing time required for the system to process the simplest image (such as a single target object, uniform background, ideal lighting), from image acquisition, target recognition to outputting positioning information. On the basis of obtaining the image complexity, the system calculates the delay increment caused by the increase in image complexity according to the unit complexity delay increment. The unit complexity delay increment can be dynamically adjusted through experiments or an adaptive algorithm according to the system processing ability and algorithm efficiency. Finally, the basic processing time and the delay increment are added together to obtain the total time-delay value for the current image data. This total time-delay value is input into the robot control system as a time-delay compensation parameter to correct the real-time positioning information and motion control instructions of the target object. By compensating for the time lag in current image acquisition and processing, the system achieves precise positioning of dynamic moving targets and manipulator operation control, reduces the positioning deviation caused by system time delay, and improves the real-time response ability and operation accuracy of the robot system.

[0051] Through the quantitative evaluation of image complexity and the calculation of delay increment, this embodiment effectively solves the problem of processing time fluctuations caused by the complexity of different image data, ensuring that the robot system has high timeliness in positioning and controlling target objects. Compared with traditional fixed-delay estimation or static model compensation methods, this method introduces a dynamic image analysis and time-delay adaptive calculation mechanism, significantly improving the adaptability and robustness of the system in complex production environments. It further optimizes the real-time control and fine operation level of the robot manipulator on the dynamic production line, reduces the product defect rate caused by time-delay errors, and improves the overall production efficiency and process stability.

[0052] The specific steps for determining the acting force that the robotic arm needs to apply in S4 include: determining the reverse acting force generated by the target object on the robotic arm; measuring the actual distance between the force application point of the target object and its support point; measuring the actual distance between the position where the robotic arm applies the acting force and its fixed support point; multiplying the reverse acting force generated by the target object by the acting distance of the object to obtain the torque generated by the target object; dividing the torque of the target object by the acting distance of the robotic arm to obtain the acting force that the robotic arm needs to apply; and adjusting the output of the robotic arm according to the acting force calculated above.

[0053] The specific formula for determining the acting force that the robotic arm needs to apply is:

[0054] Among them, X1 represents the acting force that the robotic arm needs to apply, X2 represents the reverse acting force generated by the target object, Y1 represents the actual distance between the force application point of the robotic arm and its fixed support point, and Y2 represents the actual distance between the force application point of the target object and its support point.

[0055] This embodiment details the calculation method of the force applied by the robotic arm in step S4, ensuring the stability and reliability of the robotic arm during high-precision positioning operations. First, the system determines the reaction force generated by the target object during contact with the robotic arm through sensors or physical modeling analysis. This reaction force is the reaction of the object to the robotic arm due to external forces and can be measured by force sensors, strain gauges, or dynamic simulation systems. Subsequently, the system measures the actual distance between the force application point of the target object (i.e., the point where the external force acts) and the support point (i.e., the support position of the object's fixed or contact surface). This distance reflects the lever arm length of the torque generated by the target object and is an important parameter for calculating the torque generated by the target object. Further measure the actual distance from the position where the robotic arm applies the force (such as robotic fingers, grippers, etc.) to its fixed support point (usually the base or joint axis of the robotic arm). This distance determines the lever arm length of the torque generated by the robotic arm. After obtaining the above data, the system multiplies the reaction force generated by the target object by its acting distance to calculate the torque of the target object on the outside world. To maintain the torque balance of the system and achieve precise control, the robotic arm needs to apply a corresponding force. This force is calculated by dividing the torque generated by the target object by the acting distance of the robotic arm. Dynamically adjust the output control parameters of the robotic arm according to the calculated force, drive the servo motor or hydraulic actuator to apply the corresponding force, and realize the stable clamping or operation of the target object, ensuring that the robotic arm maintains attitude stability and operation accuracy in a complex dynamic environment. The entire calculation and adjustment process realizes real-time feedback through a closed-loop control system, ensuring accurate force application and rapid response, and improving the overall operation performance of the robot system. By using the torque balance principle to calculate the force applied by the robotic arm, this embodiment effectively improves the force control accuracy and dynamic stability of intelligent industrial robots in high-precision operations. Compared with the traditional method of setting the force application value based on experience or static models, this method introduces a real-time data measurement and dynamic calculation mechanism, which can automatically adjust the force application according to the changes of the target object and the operating environment, reducing errors and operation failure rates. It improves the compliance and force control ability of the system, enabling the robotic arm to have stronger adaptability and safety when handling target objects of different materials, shapes, and weights. It further optimizes the production process and improves the comprehensive performance in application scenarios such as automated assembly, precision operation, and safe interaction.

[0056] S1 also includes estimating the frame rate of the visual acquisition module. Specifically, according to a preset proportionality coefficient, a ratio operation is performed on the maximum movement speed reached by the object on the production line and the actual movement speed of the object currently measured to calculate the frame rate. Then, the initial position of the object on the production line is detected by a sensor. Next, the measured object speed and the current position information are input into the control module to calculate the adjustment range of the frame rate. If the adjustment range exceeds the preset threshold, a real-time frame rate control mechanism is triggered to dynamically adjust the frame rate of the visual acquisition module. If the adjustment range does not exceed the threshold, the original frame rate remains unchanged.

[0057] Based on the S1 step, this implementation further improves the estimation and control mechanism of the frame rate of the visual acquisition module. First, the system detects the current actual speed V1 of the moving object on the production line in real time through a sensor or a visual recognition system. According to the maximum movement speed V max set by the system, using a preset proportionality coefficient K1, the ratio R = V max / V1 is calculated to preliminarily estimate the sampling frame rate F required by the visual acquisition module. This ratio reflects the proportion of the current object movement state relative to the maximum processing capacity of the system to ensure that the visual system acquisition frequency conforms to the change in the object movement speed. Subsequently, the system obtains the initial position information P0 of the target object through sensors (such as laser rangefinders, infrared sensors, visual sensors, etc.) installed in the production line or the robot working area. This information and the real-time measured speed data V1 are input into the control module. The control module calculates the adjustment range ΔF of the sampling frame rate F compared to the current set value based on a kinematic model or a prediction algorithm. The calculation of the adjustment range ΔF is based on the currently detected change in movement speed and the change in object spatial displacement. By dynamically analyzing and predicting the subsequent movement trend, it is determined whether it is necessary to perform real-time adjustment of the sampling frame rate. If the calculated adjustment range ΔF exceeds the system preset threshold T (i.e., ΔF > T), the real-time frame rate control mechanism is triggered. The system dynamically adjusts the sampling speed of the visual acquisition module according to the newly calculated sampling frame rate F, increasing or decreasing the image acquisition frequency to ensure that the image acquisition is always synchronized with the movement state of the target object, avoiding image blurring, recognition errors, or information loss. If the adjustment range ΔF does not exceed the preset threshold T (i.e., ΔF ≤ T), the current sampling frame rate remains unchanged to reduce system resource consumption and control complexity. This real-time control process is implemented through a closed-loop control system, continuously optimizing the sampling frame rate by combining motion state feedback to ensure the efficient response of the system to the movement state of the target object on the production line and the stability of the image acquisition quality.

[0058] By calculating the visual acquisition frame rate based on the ratio of the maximum movement speed to the actual movement speed of the target object, this embodiment realizes the dynamic optimization control of the sampling frequency, effectively improving the sensitivity and response speed of the system to the state changes of moving objects on the production line. By introducing an adjustment amplitude judgment and threshold control mechanism, the problem of frequent adjustment of the sampling frame rate due to small fluctuations is avoided, reducing the system control complexity and resource occupancy, and improving the system stability and energy consumption management efficiency. The overall solution enhances the image acquisition ability and data processing efficiency of the robot system in a high-speed dynamic operating environment, and improves the automation control level and operation reliability of intelligent industrial robots.

[0059] S1 also includes first using the Kalman filtering method to estimate the acceleration of the object on the production line; then, based on the estimated acceleration value, updating the parameters in the object motion speed model; next, introducing an acceleration change constraint condition to limit the difference between the updated acceleration value and the acceleration value at the previous moment; finally, combining the updated acceleration value, calculating the average motion rate in the target area, and setting the frame rate of the visual acquisition module according to this average rate.

[0060] Based on the S1 step, this embodiment further optimizes the frame rate regulation mechanism of the visual acquisition module by introducing the Kalman filtering method to perform high-precision estimation and dynamic correction of the acceleration of moving objects on the production line. First, the system collects the motion state data of the target object, including basic information such as position and speed, through sensors or a visual recognition system. Subsequently, the Kalman filtering algorithm is used to dynamically process the collected data to eliminate interference such as sensing errors and system noise, and obtain a smooth and accurate acceleration estimation value Z1. The Kalman filter performs recursive prediction and state update of the acceleration of the target object based on the state space model and the measurement noise model to ensure the stability and robustness of the estimated data. The system uses the estimated acceleration value Z1 to update the relevant parameters in the object motion speed model in real time, including the speed prediction value, acceleration state quantity, etc., to improve the description accuracy of the motion model for the dynamic behavior of the target object. After the model update is completed, the system introduces an acceleration change constraint condition to limit the difference ΔZ between the current estimated acceleration Z1 and the acceleration Z0 at the previous moment. This constraint condition usually adopts an absolute value limit or a dynamic window filtering strategy to ensure that the acceleration change amplitude is within a reasonable range, avoiding misjudgment or control anomalies caused by sudden interference or abnormal data, and improving the robustness and control stability of the system. Finally, combining the updated acceleration value, the system calculates the average motion rate V of the object in the target area. avg This average rate is calculated by comprehensively analyzing the acceleration and speed data of the target object in a continuous time period, reflecting the overall motion trend of the target object. The system sets the frame rate of the visual acquisition module according to V. avgSet or dynamically adjust the sampling frame rate F of the visual acquisition module to ensure that the image acquisition frequency matches the actual motion state of the target object, improving the timeliness and effectiveness of image acquisition. This processing flow effectively improves the accuracy and real-time performance of the estimation of the motion state of objects on the production line through a recursive prediction and real-time feedback mechanism, ensuring the synchronization and response ability of the visual acquisition module in a dynamic environment.

[0061] By dynamically estimating the acceleration of the target object using the Kalman filtering method, this embodiment significantly improves the system's perception accuracy and prediction ability for changes in the motion state of moving objects on the production line. Compared with the traditional static velocity model, the Kalman filter can effectively eliminate system noise and sensing errors, improving the stability and reliability of the estimation of motion parameters. Introducing acceleration change constraint conditions further ensures the system's ability to suppress abnormal data and avoids false triggering and misregulation problems. Calculate the average motion rate within the target area based on the acceleration value and adjust the sampling frame rate of the visual acquisition module accordingly, achieving precise control of the sampling frequency and ensuring the optimization of image acquisition quality and system resource utilization. This method enhances the adaptability and control accuracy of the intelligent industrial robot system to a high-speed dynamic production environment, effectively improving the automation operation efficiency and system stability.

[0062] In S1, according to the estimated acceleration value, update the parameters in the object motion speed model, including calculating a new acceleration value by weighted calculation of the acceleration value at the previous moment and the currently measured object motion speed. When performing weighted calculation, a smoothing factor is used, and the value range of the smoothing factor is 70% - 95%.

[0063] This embodiment further improves the method for dynamically updating the parameters of the object motion speed model in step S1. Based on the Kalman filter, the system introduces a weighted smoothing calculation mechanism for updating the acceleration value of the target object to improve the stability and response sensitivity of state estimation. First, the system obtains the currently measured actual motion speed V1 of the target object through the Kalman filter algorithm or other sensing data analysis methods, and combines it with the acceleration estimation value Z0 at the previous moment. The system uses a weighted calculation method to comprehensively consider the difference between the acceleration Z0 at the previous moment and the currently measured motion speed V1, and calculates a new acceleration value Z1. The weighted calculation formula is: Z1 = α × Z0 + (1 - α) × ΔV / Δt;

[0064] Among them, α is a smoothing factor, with a value range of 70% to 95%, which is flexibly set according to application requirements and the complexity of the motion state. ΔV is the difference between the velocity V1 measured at the current moment and the velocity V0 at the previous moment, and Δt is the time interval. The introduction of the smoothing factor realizes an effective balance between the historical acceleration value and the current velocity change. By increasing the value of α, the system can enhance the retention of the historical state and reduce the drastic fluctuations caused by noise or abnormal data; when the value of α is decreased, the system enhances the sensitivity to the current state change and the ability to quickly respond to dynamic environmental changes. The updated acceleration value Z1 is used to real-time correct the relevant parameters in the object motion velocity model, including state information such as predicted velocity and position, and provides a highly reliable data basis for subsequent control links such as the calculation of the average rate of the target area and the setting of the frame rate of the visual acquisition module. Through smoothed weighted update, the system ensures that it still has a stable and accurate acceleration estimation ability in a noisy environment, effectively improving the system's tracking ability for the motion state of high-speed dynamic change objects.

[0065] By introducing the weighted calculation and smoothing factor mechanism, this embodiment significantly improves the stability and robustness of the acceleration estimation value. Compared with the traditional method of directly calculating acceleration using the current velocity difference, this method can effectively filter out the abnormal effects brought by high-frequency noise and data mutations, and avoid the system from frequently adjusting the sampling frame rate or control strategy due to short-term data anomalies. The flexible setting range (70% - 95%) of the smoothing factor ensures that the system can achieve an adaptive balance between sensitivity and stability according to specific application scenarios, improving the overall response performance of the system. This optimization mechanism improves the control accuracy and stability of the intelligent industrial robot system in complex production environments, and is particularly suitable for application scenarios with high-speed dynamic operations or high requirements for position and velocity control accuracy, further enhancing the reliability of system automation control and data processing.

[0066] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive vision recognition and correction method for intelligent industrial robots, characterized in that, The method includes: S1. Adjust the frame rate of the visual acquisition module based on the real-time estimation result of the moving speed of the object on the production line; S2. Collect the image data during the movement of the object and generate the initial positioning information, including selecting a reference object with a known size in the image, measuring the pixel distance of the reference object in the image, measuring the pixel distance of the target object in the image, calculating the physical distance of the target object in the actual space, and using the calculated actual distance of the target object as the initial positioning information of the robot system; S3. Analyze the deviation between the initial positioning information and the delay parameter, and generate a position correction instruction, including determining the basic processing time generated according to different amounts of image data, analyzing the delay deviation of the initial positioning information, calculating the total delay value under the current image data, and generating a position correction instruction according to the total delay value under the current image data to ensure the operation accuracy of the robot manipulator; S4. Adjust the actions of the manipulator according to the position correction instruction to complete the high-precision positioning operation, including determining the action point and reaction force of the target according to the position of the target object and the correction instruction, determining the acting force that the manipulator needs to apply, and controlling the manipulator to make adjustments according to the calculation result.

2. An adaptive vision recognition and correction method for an intelligent industrial robot according to claim 1, characterized in that: The S1 includes determining the change rate of the moving speed of the target object on the production line, and calculating the image sampling frequency of the visual acquisition module as time goes by.

3. An adaptive vision recognition and correction method for an intelligent industrial robot according to claim 2, characterized in that: The specific steps for determining the change rate of the moving speed of the target object on the production line in the S1 include: detecting and recording the speed of the moving object on the production line at the current moment in real time through the visual recognition system or sensor; retrieving the moving speed of the object at the previous moment from the historical data of the system; confirming the time interval between the current moment and the previous moment; subtracting the speed value at the previous moment from the speed value at the current moment to obtain the speed change amount of the object during this period; dividing the speed change amount by the time interval between the two moments before and after to obtain the speed change rate per unit time.

4. An adaptive vision recognition and correction method for an intelligent industrial robot according to claim 3, characterized in that: The specific steps for calculating the image sampling frequency of the visual acquisition module in the S1 include determining the time elapsed from the start of sampling to the current moment according to the obtained speed change rate per unit time, multiplying the speed growth rate by the current estimated time to obtain the sampling frequency of the visual acquisition module, and using the calculated sampling frame rate as a control parameter to adjust the sampling speed of the visual acquisition module in real time to ensure that the image acquisition is synchronized with the moving speed of the object.

5. An adaptive vision recognition and correction method for an intelligent industrial robot according to claim 1, characterized in that: The specific steps for calculating the physical distance of the target object in the actual space in the S2 include first, taking the actual distance value of the reference object in reality; then, calculating the ratio of the measured distance of the target object in the image to the measured distance of the reference object in the image; finally, multiplying the actual distance value of the reference object by the above ratio to obtain the physical distance of the target object in the real space.

6. An adaptive vision recognition and correction method for an intelligent industrial robot according to claim 1, characterized in that: The specific steps for calculating the total time delay value under the current image data in S3 are as follows: According to the processed image data, evaluate the complexity of the current image. The complexity of the current image is quantitatively analyzed based on factors such as the number of target objects, background complexity, and lighting conditions; Determine the basic processing time under ideal conditions. The basic processing time is the shortest time required to complete image acquisition, recognition, and output of positioning information under the simplest image data conditions; According to the value of the image complexity, determine the delay increment caused by the increase in image complexity. The delay increment caused by the increase in image complexity is the product of the image complexity and the delay increment per unit complexity; Add the basic processing time and the delay increment caused by the increase in image complexity to obtain the total time delay value under the current image data; Use the calculated total time delay value as a time delay compensation parameter, and adjust the positioning information and motion control instructions according to the compensation parameter.

7. An adaptive vision recognition and correction method for an intelligent industrial robot according to claim 1, characterized in that: The specific steps for determining the force that the robotic arm needs to apply in S4 are as follows: Determine the reverse force generated by the target object on the robotic arm; Measure the actual distance between the force application point of the target object and its support point; Measure the actual distance between the position where the robotic arm applies the force and its fixed support point; Multiply the reverse force generated by the target object by the action distance of the object to obtain the torque generated by the target object; Divide the torque of the target object by the action distance of the robotic arm to obtain the force that the robotic arm needs to apply; Adjust the output of the robotic arm according to the force calculated above.

8. An adaptive vision recognition and correction method for an intelligent industrial robot according to claim 1, characterized in that: S1 also includes estimating the frame rate of the visual acquisition module. Specifically, according to a preset proportional coefficient, perform a ratio operation on the maximum movement speed of the object on the production line and the actual movement speed of the object measured currently to calculate the frame rate. Then, detect the initial position of the object on the production line through a sensor; Next, input the measured object speed and current position information into the control module to calculate the adjustment range of the frame rate; If the adjustment range exceeds the preset threshold, trigger the real-time frame rate control mechanism to dynamically adjust the frame rate of the visual acquisition module. If the adjustment range does not exceed the threshold, maintain the original frame rate unchanged.

9. An adaptive vision recognition and correction method for an intelligent industrial robot according to claim 1, characterized in that: S1 also includes first using the Kalman filtering method to estimate the acceleration of the object on the production line; Then, based on the estimated acceleration value, update the parameters in the object motion speed model; Next, introduce an acceleration change constraint condition to limit the difference between the updated acceleration value and the acceleration value at the previous moment; Finally, combine the updated acceleration value to calculate the average motion rate in the target area, and set the frame rate of the visual acquisition module according to this average rate.

10. An adaptive vision recognition and correction method for an intelligent industrial robot according to claim 9, characterized in that: Updating the parameters in the object motion speed model according to the estimated acceleration value in S1 includes obtaining a new acceleration value by performing a weighted calculation on the acceleration value at the previous moment and the actual movement speed of the object measured currently. When performing the weighted calculation, a smoothing factor is used, and the value range of the smoothing factor is 70% - 95%.

Images