Intelligent quality inspection system combined with quality inspection platform and implementation method thereof

Abstract

According to the intelligent quality inspection system combined with the quality inspection platform and the implementation method thereof, the functions of the quality inspection platform are combined, a path planning algorithm is adopted, and intelligent path planning and obstacle avoidance of a quality inspection robot in a detection area are achieved through a sensor, a camera and a laser system; by means of the high-precision measurement and real-time data processing capability of the quality inspection platform, real-time monitoring and data processing are directly carried out through a laser system and other sensors, so that the quality inspection operation is ensured to be carried out smoothly; through multi-sensor synchronous control of a quality inspection platform and an improved Golden Eagle algorithm, the path planning precision is improved to the same level as the measurement precision, and it is ensured that no extra error is introduced into obstacle avoidance motion; a'perception-decision 'closed-loop architecture is adopted, path turning is smoothed by using a B spline curve, continuous path curvature is guaranteed under response delay less than or equal to 50ms, and measurement jitter is eliminated; a multi-source data real-time fusion model based on a quality inspection platform is established, and the omission ratio of a complex environment is smaller than 0.1% by integrating the space pose, the material characteristic and the motion state of an obstacle.

Description

Technical Field

[0001] This invention relates to an intelligent quality inspection system and its implementation method that integrates a quality inspection platform, and particularly to an intelligent quality inspection system and its implementation method that integrates a quality inspection platform for wear protection of the crossbeam column and bag clamp of a vacuum lifting bag. Background Technology

[0002] With breakthroughs in high-precision optical measurement technology, intelligent quality inspection solutions centered on multi-sensor measurement systems in quality inspection platforms are reshaping the quality control system of the precision manufacturing industry. By integrating specialized fixture design, machine vision algorithms, and automated control modules, the quality inspection platform achieves closed-loop inspection of sub-micron dimensional tolerances and surface defects in complex parts, providing a revolutionary solution, especially for the high-volume, efficient inspection of micro-sized parts. The design of the parts inspection device makes the inspection process smoother and better facilitates the inspection of precision components. The use of the quality inspection system makes the parts inspection process more streamlined and accelerates the inspection speed. This system avoids the problem of missed inspections and reduces the possibility of damage to parts caused by human operation. Multi-station quick-change fixtures enable non-destructive clamping of various irregularly shaped parts. The fixture surface is treated with anodized matte finish, improving inspection efficiency by more than 300% compared to traditional tooling. Practical verification shows that this solution can reduce manual intervention by 90% and reduce the product scratch rate from 0.5% to below 0.02%, providing an efficient and reliable quality control solution for the precision manufacturing field.

[0003] Losses due to defects can reach 22% of the product's selling price. For example, in the automotive industry, in 2021, 15% of vehicles recalled in China were due to manufacturing defects. Traditional methods fail: Manual visual inspection suffers from inconsistent standards, high labor intensity, and a high rate of missed detections. Furthermore, high-speed production lines (such as coated steel plate production lines with speeds of 800 m / min) require detection algorithms to complete single-frame processing within 18 microseconds. Data scarcity constrains AI: AI quality inspection models rely on a large number of defect samples, but in actual production, samples of minute defects (such as cracked battery terminals or broken semiconductor solder joints) are scarce and the labeling cost is extremely high. Surface defect detection: Due to the multi-color and multi-texture characteristics of coated steel plates, traditional image processing (edge ​​extraction, threshold segmentation) is easily interfered with, requiring adaptive algorithms to improve robustness. Complex environment path planning: In dynamic industrial scenarios (such as hydroelectric turbine blade inspection and road inspection), fixed-path robots cannot cope with unknown obstacles, and the cumulative error of mechanical motion is >50 μm, affecting measurement accuracy.

[0004] Traditional systems (such as rail-guided CMMs) rely on preset mechanical paths and cannot dynamically avoid obstacles. When encountering environmental changes (such as temporary obstacles), manual adjustments are required, leading to inspection interruptions. Obstacle avoidance solutions like laser-guided AGVs operate independently of the measurement system, separating path planning from quality control. This results in insufficient motion trajectory accuracy (>50μm), failing to meet micrometer-level quality inspection requirements. While existing obstacle avoidance algorithms (such as basic RRT) support dynamic path generation, sharp turns / stops cause mechanical vibrations, distorting measurement data (especially in optical scanning scenarios). The phased processing architecture (perception → planning → measurement) suffers from asynchronous data bottlenecks, with sudden obstacle response delays exceeding 200ms and a false negative rate exceeding 3%. Vision, laser, and tactile sensors operate independently, and data is not integrated into a unified decision model. Obstacle recognition relies on a single information source (such as only contour or distance), leading to high misjudgment rates for complex workpieces.

[0005] Therefore, there is a particular need for an intelligent quality inspection system that integrates with a quality inspection platform and its implementation method to solve the aforementioned existing problems. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent quality inspection system and its implementation method that integrates a quality inspection platform. Addressing the shortcomings of existing technologies, this system improves the accuracy and efficiency of quality inspection through vision, sensors, and laser systems. By real-time monitoring, data analysis, and feedback of the inspection area, it achieves efficient and accurate quality inspection, avoiding missed detections and misjudgments, and improving the stability and precision of the quality inspection process.

[0007] The technical problem solved by this invention can be achieved by the following technical solutions:

[0008] In a first aspect, the present invention provides an intelligent quality inspection system integrated with a quality inspection platform, comprising a device end, an edge computing end, and a cloud platform end. The device end includes a quality inspection robot, which is connected to a 3D LiDAR module, an RGB-D camera module, an IMU inertial unit, and a robotic arm end-effector sensor. The edge computing end includes a real-time task scheduler, a dynamic path planning engine, a multi-machine collaborative controller, and a data preprocessing module. The cloud platform end includes an AI model training center, a detection rule base, historical data, and a path optimization algorithm library. The real-time task scheduler is connected to the dynamic path planning engine, and the AI ​​model training center is connected to the detection rule base. The AI ​​model training center updates the model and inputs it into the dynamic path planning engine. The historical data is connected to the path optimization algorithm library. The raw data from the 3D LiDAR module, the RGB-D camera module, and the IMU inertial unit are input into the data preprocessing module, and the data from the multi-machine collaborative controller is input into the data preprocessing module. The data preprocessing module inputs environmental feature data into the dynamic path planning engine, and the dynamic path planning engine inputs motion commands into the quality inspection robot.

[0009] Secondly, the present invention provides a method for implementing an intelligent quality inspection system combined with a quality inspection platform, comprising the following steps:

[0010] Step S1: Obtain detection area information; acquire accurate data of the detection area through the quality inspection platform, including the geometric information of existing obstacles and the target area, and perform path planning; make real-time adjustments based on high-precision measurement data to cope with possible environmental changes; utilize the automated processing capabilities of the quality inspection platform to further optimize path planning, making the quality inspection process smoother;

[0011] Step S2: Dealing with Unknown Obstacles and Adjusting Smooth Paths; In the actual inspection process, unknown obstacles or real-time changing environments may be encountered. Through the information from the laser sensors, cameras, and other sensors integrated into the quality inspection platform, obstacle features can be quickly extracted, and the path can be adjusted based on real-time feedback; Whether it is a known obstacle or a suddenly appearing obstacle, a smooth obstacle avoidance path is planned through intelligent algorithms, ensuring the continuous progress of the quality inspection task.

[0012] Step S3: Optimize the motion path of the quality inspection robot; through real-time data collection and analysis by the quality inspection platform, combined with the spatial pose and geometric information of obstacles, the optimal inspection path is calculated; the quality inspection robot will dynamically adjust its motion path based on this information to avoid obstacles and ensure stable operation in complex inspection environments until the task is successfully completed.

[0013] In a preferred embodiment of the present invention, in step S1, the laser rangefinder point cloud data, high-resolution industrial camera RGB image and depth information, and tactile sensor pressure / vibration data of the quality inspection platform are fused in real time. The laser rangefinder point cloud data is subjected to point cloud filtering for noise reduction / downsampling. The high-resolution industrial camera RGB image and depth information are subjected to image preprocessing, distortion correction / enhancement. The tactile sensor pressure / vibration data is synchronized based on hardware triggering / software time. After hand-eye calibration, the relative pose of the document transmitter is determined. All data are transformed to the world coordinate system through coordinate transformation. Vision and laser fusion generate a high-precision point cloud with color. Semantic segmentation and annotation identify and classify obstacles. Tactile data fusion verifies the material and accurate contour. Filtering algorithms are used to predict motion, dynamically update and track, construct a dynamic three-dimensional environment model, and output a high-precision 3D voxel map, including obstacle categories and contours, material types, and other physical properties.

[0014] In a preferred embodiment of the present invention, in step S2, the spatial pose of the sudden obstacle is extracted in real time by a laser sensor and a vision system, triggering local path rapid replanning (RRT* algorithm variant), maintaining a safe distance of ≥10mm and generating a smooth detour path.

[0015] In a preferred embodiment of the present invention, in step S2, the high-precision point cloud map (accuracy ≤ 5μm) generated by the quality inspection platform is converted into a gridded motion space for path planning, thereby realizing the transfer of measurement accuracy to motion accuracy.

[0016] In a preferred embodiment of the present invention, in step S3, an improved strategy and a B-spline curve smoothing optimization module are introduced into the traditional Golden Eagle algorithm: the environmental grid map is dynamically updated according to real-time sensor data; and a smooth trajectory with continuous curvature is automatically generated at path turning points (eliminating sudden stop jitter).

[0017] In a preferred embodiment of the present invention, in step S3, the sub-millimeter positioning accuracy (≤0.1mm) of the quality inspection platform and the online environmental compensation module (temperature / vibration) are used to ensure the stability of the robot's motion trajectory and avoid measurement jitter.

[0018] In a preferred embodiment of the present invention, in step S3, the robot end effector integrates a standardized sensor interface, which supports the switching of optical / laser / contact probes in seconds according to task requirements, and is automatically calibrated by the quality inspection platform.

[0019] In a preferred embodiment of the present invention, in step S3, the B-spline path is ensured to meet the measurement stability requirements by using a curvature change rate threshold constraint (e.g., ≤0.05rad / m) to avoid data distortion caused by robot acceleration and deceleration.

[0020] This invention relates to an intelligent quality inspection system and its implementation method that integrates a quality inspection platform. Compared with existing technologies, this system combines the functions of the quality inspection platform and employs a path planning algorithm. Through sensors, cameras, and laser systems, it enables the quality inspection robot to intelligently plan paths and avoid obstacles within the inspection area, overcoming the inaccuracies caused by path deviations, environmental interference, or measurement errors in traditional quality inspection processes. Leveraging the high-precision measurement and real-time data processing capabilities of the quality inspection platform, real-time monitoring and data processing are directly performed through the laser system and other sensors, ensuring smooth quality inspection operations. Through multi-sensor synchronous control of the quality inspection platform and an improved Golden Eagle algorithm, the path planning accuracy is improved to the same level as the measurement accuracy (≤5μm), ensuring that obstacle avoidance does not introduce additional errors. A "perception-decision" closed-loop architecture is adopted, utilizing B-spline curves to smooth path transitions, ensuring continuous path curvature (rate of change ≤0.05rad / m) with a response delay of ≤50ms, eliminating measurement jitter. A real-time multi-source data fusion model based on the quality inspection platform is established, comprehensively considering the spatial pose, material characteristics, and motion state of obstacles, achieving a false negative rate of <0.1% in complex environments, thus achieving the objectives of this invention.

[0021] The features of the present invention can be clearly understood by referring to the drawings and the following detailed description of preferred embodiments. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the intelligent quality inspection system combined with the quality inspection platform of the present invention;

[0023] Figure 2 This is a flowchart illustrating the implementation method of the intelligent quality inspection system combined with the quality inspection platform of the present invention.

[0024] Figure 3 This is a flowchart illustrating step S1 of the present invention;

[0025] Figure 4 This is a flowchart illustrating step S2 of the present invention;

[0026] Figure 5 This is a flowchart illustrating the Golden Eagle Search algorithm of the present invention;

[0027] Figure 6 This is a flowchart illustrating the improved Golden Eagle Search algorithm of the present invention. Detailed Implementation

[0028] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below with reference to specific illustrations.

[0029] Example

[0030] like Figure 1 As shown, the intelligent quality inspection system of the present invention, which integrates a quality inspection platform, includes a device end, an edge computing end, and a cloud platform end. The device end includes a quality inspection robot, which is connected to a 3D LiDAR module, an RGB-D camera module, an IMU inertial unit, and a robotic arm end-effector sensor. The edge computing end includes a real-time task scheduler, a dynamic path planning engine, a multi-machine collaborative controller, and a data preprocessing module. The cloud platform end includes an AI model training center, a detection rule base, historical data, and a path optimization algorithm library. The real-time task scheduler is connected to the dynamic path planning engine, and the AI ​​model training center is connected to the detection rule base. The AI ​​model training center updates the model and inputs it into the dynamic path planning engine. The historical data is connected to the path optimization algorithm library. The raw data from the 3D LiDAR module, the RGB-D camera module, and the IMU inertial unit are input into the data preprocessing module, and the data from the multi-machine collaborative controller is also input into the data preprocessing module. The data preprocessing module inputs environmental feature data into the dynamic path planning engine, and the dynamic path planning engine inputs motion commands into the quality inspection robot.

[0031] like Figure 2 As shown, the intelligent quality inspection system and its implementation method combining a quality inspection platform of the present invention include the following steps:

[0032] Step S1: Obtain detection area information; acquire accurate data of the detection area through the quality inspection platform, including the geometric information of existing obstacles and the target area, and perform path planning; make real-time adjustments based on high-precision measurement data to cope with possible environmental changes; utilize the automated processing capabilities of the quality inspection platform to further optimize path planning, making the quality inspection process smoother;

[0033] Step S2: Dealing with Unknown Obstacles and Adjusting Smooth Paths; In the actual inspection process, unknown obstacles or real-time changing environments may be encountered. Through the information from the laser sensors, cameras, and other sensors integrated into the quality inspection platform, obstacle features can be quickly extracted, and the path can be adjusted based on real-time feedback; Whether it is a known obstacle or a suddenly appearing obstacle, a smooth obstacle avoidance path is planned through intelligent algorithms, ensuring the continuous progress of the quality inspection task.

[0034] Step S3: Optimize the motion path of the quality inspection robot; through real-time data collection and analysis by the quality inspection platform, combined with the spatial pose and geometric information of obstacles, the optimal inspection path is calculated; the quality inspection robot will dynamically adjust its motion path based on this information to avoid obstacles and ensure stable operation in complex inspection environments until the task is successfully completed.

[0035] In step S1, the laser rangefinder point cloud data, high-resolution industrial camera RGB image and depth information, and tactile sensor pressure / vibration data from the quality inspection platform are fused in real time. The laser rangefinder point cloud data undergoes point cloud filtering for noise reduction / downsampling. The high-resolution industrial camera RGB image and depth information undergo image preprocessing, distortion correction / enhancement. The tactile sensor pressure / vibration data is synchronized based on hardware triggering / software time. After hand-eye calibration, the relative pose of the document feeder is determined. All data are transformed to the world coordinate system through coordinate transformation. Vision and laser fusion generate a high-precision point cloud with color. Semantic segmentation and annotation identify and classify obstacles. Tactile data fusion verifies material and precise contours. Filtering algorithms are used to predict motion, dynamically update and track, construct a dynamic 3D environment model, and output a high-precision 3D voxel map, including obstacle categories, contours, material types, and other physical properties.

[0036] In step S2, the spatial pose of sudden obstacles is extracted in real time by using a laser sensor and a vision system, triggering local path rapid replanning (RRT* algorithm variant), maintaining a safe distance of ≥10mm and generating a smooth detour path.

[0037] In step S2, the high-precision point cloud map (accuracy ≤ 5μm) generated by the quality inspection platform is converted into a gridded motion space for path planning, thereby realizing the transfer of measurement accuracy to motion accuracy.

[0038] The quality inspection platform acquires 3D data of the workpiece surface using a high-precision structured light scanner or laser interferometer. After preprocessing such as point cloud denoising and multi-point point cloud registration (accuracy ≤ 5μm), a high-precision point cloud map is generated. Subsequently, the point cloud is converted into a 3D grid map through voxelization, and each grid is assigned attributes (such as obstacles, free regions), while semantic information such as distance field is calculated. Finally, the path planning algorithm combines the geometric constraints of the grid map with the kinematic constraints of the robot to generate a collision-free and smooth optimized trajectory, realizing a closed-loop transfer from measurement accuracy to motion accuracy.

[0039] In step S3, an improved strategy and a B-spline curve smoothing optimization module are introduced into the traditional Golden Eagle algorithm: the environmental grid map is dynamically updated based on real-time sensor data; and a smooth trajectory with continuous curvature is automatically generated at path turning points (eliminating sudden stop jitter).

[0040] Improvements to the Golden Eagle Search Algorithm

[0041] (1) Golden Eagle Search Algorithm

[0042] The Golden Eagle Search (GEO) algorithm is based on the behavior of golden eagles adjusting their speed at different stages of predation. In GEO, each golden eagle selects a target prey, which is the optimal solution currently found by the flock. Each search agent in the Golden Eagle algorithm remembers its own optimal solution and randomly selects a target prey from the collective memory in each iteration. It then calculates its own attack and cruise vectors based on this prey to update its memory. To enhance the diversity and exploration capabilities of the search, a one-to-one mapping strategy is proposed, allowing each search agent to randomly select a prey from the memory of another search agent as a target in the current iteration. It is important to note that the closest or furthest prey is not necessarily the selected one. In this scheme, each prey in memory can only be assigned or mapped by one search agent, which then performs attack and cruise operations on it. The golden eagle's attack vector can be calculated using formula (1).

[0043]

[0044] It's the eagle's attack vector. This is the best location (for prey) the eagle has visited so far. The eagle's current location.

[0045] The golden eagle cruises its prey, calculated using formula (2).

[0046]

[0047] It is a normal vector. It is a variable vector. It is any point on the hyperplane. We will X i As any point on the hyperplane, let A i As the normal vector of the hyperplane, it can be shown Golden Eagle Cruise Vector Diagram. This can be obtained from (3).

[0048]

[0049] The attack vector is: The design variable vector is: The location of the selected prey is:

[0050] Use the following procedure to find a random k-dimensional target point. Located on the Golden Eagle cruise hyperplane, we obtain i. The steps are as follows:

[0051] 1) Randomly select a variable as the fixed variable. We will denote the index of the selected variable as k.

[0052] 2) Assign random values ​​to all variables except the kth variable (because the kth variable is fixed).

[0053] 3) Use equation (4) to find the value of the fixed variable.

[0054]

[0055] c k It is the p-th target point of k c Element C, a j It is an element of the j-th attack vector d is the parameter in the formula. It is the k-th attack vector element k is the index of a fixed variable. A random destination point on the cruise hyperplane was found. Equation (5) represents the destination on the "cruise hyperplane".

[0056]

[0057] Now that the target point has been determined, we can calculate the cruise vector for the Golden Eagle. The elements of the target point obtained in iteration i (t) are random numbers between 0 and 1.

[0058] The Golden Eagle's shifting includes both attack and vector. We define the Golden Eagle's step size vector i in the iteration as shown in equation (6).

[0059]

[0060] It is the attack coefficient after t iterations. It is the cruise coefficient after t iterations, and its magnitude is affected by the attack and cruise. and It is a random vector whose elements are in the interval [0,1]. and It is an Euclidean function of the attack vector and the cruise vector, and is calculated using formula (7).

[0061]

[0062] The position of the Golden Eagle at iteration t+1 is the step size vector plus the position at iteration t, and its calculation formula is shown in equation (8).

[0063]

[0064] If the golden eagle's new location *i* is better than its remembered location, the eagle's memory is updated accordingly. Otherwise, the memory remains intact, but the eagle will remain at the new location. In each new iteration, each golden eagle randomly selects another eagle from the population to circle around its best access location, calculates the attack vector, calculates the cruise vector, and finally calculates the step size vector and the new location for the next iteration. This loop continues until any termination condition is met.

[0065] Geosynchronous orbit uses p a and p c From exploration to development. The algorithm starts from low p a and high p c As the iteration progresses, p a It increases gradually, while p c The values ​​decrease gradually. The initial and final values ​​of the two parameters are defined by the user. Intermediate values ​​can be calculated using the linear transition shown in equation (9).

[0066]

[0067] t represents the current iteration, and T represents the maximum number of iterations. and The initial and final values ​​of the aggression tendency (p) a ), which are: and The initial and final values ​​of the cruise tendency (p) c The flowchart of the Golden Eagle Search algorithm is as follows: Figure 2 .

[0068] (2) Logistic chaotic mapping

[0069] The Logistic chaotic map is a nonlinear dynamical system that can generate sequences with chaotic properties. Its formula is (10).

[0070] x n+1 = r·x n (1-x n (10)

[0071] Where r is the control parameter, x is the initial value of the mapping, and n represents the number of iterations of the mapping. n It is the value after the nth iteration.

[0072] During population initialization, if individuals are too concentrated or too dispersed, it may affect the algorithm's search efficiency and convergence speed. To address this issue, we incorporate a Logistic chaotic map into population initialization. In each evolutionary iteration, the Logistic chaotic map generates random numbers to perturb the positions within the population, increasing its diversity and randomness. This allows the population to better adapt to different environments and improves the algorithm's global and local search capabilities. The purpose of adding a Logistic chaotic map during the initialization of the Golden Eagle Search algorithm is to increase the algorithm's randomness and diversity, thereby improving its search efficiency and convergence speed.

[0073] (3) Quasi-oppositional learning

[0074] Quasi-opposites learning, by introducing the concept of quasi-opposites, helps algorithms find the optimal solution faster. Specifically, quasi-opposites help the algorithm better explore the search space, thus discovering better solutions. During each optimization, the current solution and its opposite solution are compared simultaneously, and the better one is selected as the new solution. In the Golden Eagle Search algorithm, quasi-opposites learning is used to enhance the Golden Eagle's search ability and convergence speed. Specifically, when a Golden Eagle reaches its latest position, i.e., the optimal position, other Golden Eagles generate opposite points based on the optimal position and choose whether to update their own positions according to the fitness function. This allows Golden Eagles to approach prey more quickly and may also discover better solutions. The opposite solution for the Golden Eagle's position t+1 in the iteration can then be obtained.

[0075] This paper proposes an algorithm based on a quasi-opposition learning strategy, which improves the convergence speed and accuracy of the algorithm while avoiding excessive evaluation. The quasi-opposition learning strategy is a method that accelerates the search process by utilizing the relationship between the current solution and its opposite solution. In this paper, this strategy is applied only in the Golden Eagle position update stage to fully leverage its advantages. Specifically, after each position update, the opposite solution of the current solution is obtained using the quasi-opposition learning method and compared with the current optimal position. If the opposite solution is better than the current optimal position, it is replaced with the new optimal position; otherwise, the current optimal position remains unchanged. This ensures a good starting point for each iteration. When the maximum number of iterations is reached, the best individual in the population is output as the algorithm's result. This method effectively avoids getting trapped in local optima while maintaining global search capability, and does not increase the number of evaluations excessively. Therefore, by employing a quasi-opposition learning strategy in the Golden Eagle position update part, the efficiency and accuracy of the algorithm can be effectively improved, making it more suitable for solving practical problems.

[0076] In step S3, the sub-millimeter positioning accuracy (≤0.1mm) and online environmental compensation module (temperature / vibration) of the quality inspection platform are used to ensure the stability of the robot's motion trajectory and avoid measurement jitter.

[0077] Based on a high-precision sensor network (laser, temperature, vibration), an accurate mathematical model is established through precise calibration. On a hardware real-time software platform, multi-source data is fused and real-time calculations are performed. Finally, through feedforward compensation or external closed-loop control, the robot's motion commands are fine-tuned to counteract the interference caused by robot body errors, thermal deformation, and external vibrations, so that its actual motion trajectory in the task space is stabilized within the sub-millimeter range.

[0078] In step S3, the robot end effector integrates a standardized sensor interface, which supports the switching of optical / laser / contact probes in seconds according to task requirements, and is automatically calibrated by the quality inspection platform.

[0079] The standardized sensor interface integrated at the robot end effector is an automated quick-change system (such as the ATIQC series or Schunk MSE series) with multiple media (electric / pneumatic / data) channels. It supports optical / laser / contact probes to complete physical locking and media connection within seconds, and achieves automatic calibration of the tool coordinate system through integrated calibration algorithms.

[0080] In step S3, the curvature change rate threshold constraint (e.g., ≤0.05rad / m) is used to ensure that the B-spline path meets the measurement stability requirements and avoid data distortion caused by robot acceleration and deceleration.

[0081] By using nonlinear optimization algorithms (such as sequential quadratic programming SQP) with B-spline curve control points as variables, the global constraint of curvature change rate (dκ / ds) ≤ 0.05 rad / m is discretized into inequality constraints on a large number of sampling points while minimizing the path fitting error. The smooth path that finally meets the measurement stability requirements is then solved through iterative calculation.

[0082] This invention relates to an intelligent quality inspection system and its implementation method that integrates a quality inspection platform. For known inspection areas, the system uses high-precision measurements from the platform to plan paths, avoiding missed inspections caused by path errors or inaccurate data in traditional quality inspection systems. For unknown targets or complex environments, it can perceive and plan obstacle avoidance paths in real time. By reading information from cameras, laser sensors, and other sensors integrated into the quality inspection platform, obstacle features are extracted. Based on these features and distance data, advanced algorithms are used to analyze and adjust the path in real time, ensuring seamless integration of the quality inspection process.

[0083] The present invention relates to an intelligent quality inspection system and its implementation method that combines a quality inspection platform. Through the precision control algorithm of the quality inspection platform, combined with the improved Golden Eagle algorithm, the system performs path planning for the quality inspection robot, optimizes the robot's movement path in different environments, and ensures that obstacles are avoided to the greatest extent during the inspection process. The system automatically adjusts the path to ensure that the robot can smoothly reach the inspection parts in dynamic environments, thus ensuring the efficiency and accuracy of the inspection process.

[0084] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as defined by the appended claims and their equivalents.

Claims

1. An implementation method of an intelligent quality inspection system combined with a quality inspection platform, characterized in that, Comprise the following steps: Step S1, obtain the detection area information; Obtain the accurate data of the detection area through the quality inspection platform, including the existing obstacles, the geometric information of the target area, and plan the path; According to the high-precision measurement data, real-time adjustment is carried out to cope with possible environmental changes; Utilize the automatic processing capability of the quality inspection platform, further optimize the path planning, so that the quality inspection process is more smooth; Step S2, deal with unknown obstacles and smooth path adjustment; In the actual detection process, unknown obstacles or real-time changing environment will be encountered, through the information of laser sensor, camera and other sensors integrated by the quality inspection platform, the obstacle characteristics can be quickly extracted, and the path can be adjusted according to the real-time feedback; Whether it is a known obstacle or a sudden obstacle, a smooth obstacle avoidance path is planned through intelligent algorithm, and the continuous progress of the quality inspection task is ensured; Step S3, optimize the motion path of the quality inspection robot; Through real-time data acquisition and analysis by the quality inspection platform, combined with the spatial pose and geometric information of the obstacle, the optimal detection path is calculated; The quality inspection robot will dynamically adjust the motion path according to these information, avoid obstacles and ensure stable operation in complex detection environment, until the task is successfully completed.

2. The method of claim 1, wherein the method further comprises: In step S1, the laser ranging sensor, high-resolution industrial camera and tactile sensor data of the quality inspection platform are fused in real time to construct a dynamic three-dimensional environment model, realize millisecond-level identification of obstacle characteristics, and the obstacle characteristics include position, contour and material. 3.The implementation method of the intelligent quality inspection system combined with the quality inspection platform according to claim 1, wherein, In step S2, the spatial pose of the sudden obstacle is extracted in real time by the laser sensor and vision system, the local path is quickly re-planned, the safety distance is maintained ≥10mm, and the smooth bypass path is generated.

4. The method of claim 1, wherein the method further comprises: In step S2, the high-precision point cloud map generated by the quality inspection platform is converted into a rasterized motion space for path planning, realizing the transfer of measurement accuracy to motion accuracy. 5.The intelligent quality inspection system integrated with a quality inspection platform and the implementation method thereof according to claim 1, wherein, In step S3, the improved strategy and B-spline curve smoothing optimization module are introduced in the traditional JY algorithm: dynamically update the environment grid map according to real-time sensor data; The turning point of the path automatically generates a smooth trajectory with continuous curvature, eliminating sudden stop jitter.

6. The method of claim 1, wherein the method further comprises: In step S3, relying on the submillimeter level positioning accuracy and online environment compensation module of the quality inspection platform, the stability of the robot motion trajectory is ensured, and measurement jitter is avoided.

7. The method of claim 1, wherein the method further comprises: In step S3, the robot end integrates a standardized sensor interface, supports optical / laser / contact probe switching within seconds according to task requirements, and is automatically calibrated by the quality inspection platform.

8. The method of claim 1, wherein the method further comprises: In step S3, through the curvature change rate threshold constraint, it is ensured that the B-spline path meets the measurement stability requirements, and the data distortion caused by robot acceleration and deceleration is avoided.

9. An intelligent inspection system implementing the combination quality inspection platform of any of the implementation methods of claims 1 to 8. It includes device end, edge computing end and cloud platform end, the device end includes quality inspection robot, the quality inspection robot is respectively connected with 3D laser radar module, RGB-D camera module, IMU inertial unit and mechanical arm end sensor; The edge computing end includes real-time task scheduler, dynamic path planning engine, multi-machine cooperative controller and data preprocessing module; The cloud platform end includes AI model training center, detection rule library, historical data and path optimization algorithm library; The real-time task scheduler is connected with the dynamic path planning engine, the AI model training center is connected with the detection rule library, the AI model training center inputs the model after being updated into the dynamic path planning engine, and the historical data are connected with the path optimization algorithm library; the original data of the 3D laser radar module, the RGB-D camera module and the IMU inertial unit are respectively input into the data preprocessing module, the data of the multi-machine cooperative controller are input into the data preprocessing module, the environment feature data are input into the dynamic path planning engine by the data preprocessing module, and the dynamic path planning engine inputs the motion instruction into the quality inspection robot.

Images