Machine vision driven intelligent bed assembly quality monitoring method and system

Abstract

The invention discloses a machine vision-driven intelligent bed assembly quality monitoring method and system, and relates to the technical field of image processing, and the method comprises the steps: collecting the three-dimensional surface image data of an intelligent bed assembly through machine vision, and generating an assembly image data set; topological structure data of an intelligent bed frame are obtained, spatial registration is carried out in combination with the assembly body image data set, and three-dimensional twin structure parameters are constructed; traversing the three-dimensional twin structure parameters to perform assembly positioning, extracting an assembly feature vector group to perform quality identification, and drawing a quality defect map; and performing graded early warning based on the quality defect map, executing component calibration operation according to multi-stage early warning information, and constructing an assembly quality evaluation report of the intelligent bed. The technical problem of low assembly quality evaluation accuracy of the intelligent bed in the prior art is solved, and the technical effect of improving the assembly quality evaluation accuracy of the intelligent bed is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to a method and system for monitoring the assembly quality of an intelligent bed driven by machine vision. Background Art

[0002] Existing methods for evaluating the assembly quality of intelligent beds mainly rely on manual inspection or simple visual inspection. These methods cannot provide sufficient accuracy and consistency and are easily affected by human factors. In addition, most existing methods for evaluating assembly quality rely on static images or two-dimensional image analysis and cannot effectively capture the three-dimensional spatial characteristics of the assembly, resulting in inaccurate evaluation of assembly quality under complex assembly processes. Summary of the Invention

[0003] This application provides a method and system for monitoring the assembly quality of an intelligent bed driven by machine vision, which are used to solve the technical problem of low accuracy in evaluating the assembly quality of intelligent beds in the prior art.

[0004] In view of the above problems, this application provides a method and system for monitoring the assembly quality of an intelligent bed driven by machine vision.

[0005] In the first aspect of this application, a method for monitoring the assembly quality of an intelligent bed driven by machine vision is provided. The method includes:

[0006] Collecting three-dimensional surface image data of the intelligent bed assembly through machine vision to generate an assembly image dataset; obtaining topological structure data of the intelligent bed frame, performing spatial registration in combination with the assembly image dataset to construct three-dimensional twin structure parameters; traversing the three-dimensional twin structure parameters for assembly positioning, extracting an assembly feature vector group for quality identification, and drawing a quality defect map, wherein, mapping the defect probability distribution map to the three-dimensional twin structure parameters according to the defect confidence score matrix for quality identification, and constructing the quality defect map; performing hierarchical early warning based on the quality defect map, performing component calibration operations according to multi-level early warning information, and constructing an assembly quality evaluation report for the intelligent bed, wherein, parsing the quality defect map to generate an RGB coding matrix, performing real-time grading according to the RGB coding matrix, and constructing multi-level early warning information, and the multi-level early warning information includes a multi-level early warning instruction set.

[0007] In a possible implementation manner, collecting three-dimensional surface image data of the intelligent bed assembly through machine vision to generate an assembly image dataset includes: setting a plurality of trigger sensors based on the assembly line conveyor belt, and when it is detected that the assembly of the intelligent bed enters the shooting area, synchronously activating the multi-modal imaging module of the multi-spectral industrial camera through the plurality of trigger sensors; continuously frame-synchronously collecting the assembly through the multi-modal imaging module to obtain a composite image sequence; dynamically enhancing the composite image sequence to generate the assembly image dataset.

[0008] In a possible implementation, the process of obtaining the topological structure data of the intelligent bed frame includes: deploying three-dimensional line laser scanning devices on both sides of the assembly station to synchronously scan the intelligent bed frame from multiple perspectives to generate multi-perspective scan data; performing surface feature recognition on the intelligent bed frame based on the multi-perspective scan data to construct a three-dimensional coordinate system, where the three-dimensional coordinate system includes multi-perspective initial point cloud data; performing rough registration based on the multi-perspective initial point cloud data to determine multiple outlier point cloud data, removing the multi-perspective initial point cloud data according to the multiple outlier point cloud data to generate multi-perspective point cloud data; mapping the multi-perspective point cloud data to the three-dimensional coordinate system, determining the coordinates of multiple key connection nodes, and constructing the topological structure data of the intelligent bed frame according to the coordinates of the multiple key connection nodes.

[0009] In a possible implementation, performing spatial registration in combination with the assembly image dataset to construct three-dimensional twin structure parameters includes: mapping the assembly image dataset to the three-dimensional coordinate system according to the composite image sequence for correlation analysis to extract the correlation mapping relationship; setting reference marker points based on the topological structure data, and calculating the alignment parameters between the assembly image dataset and the multi-perspective point cloud data according to the reference marker points; performing spatial registration on the assembly image dataset and the multi-perspective point cloud data according to the alignment parameters and the correlation mapping relationship, and performing data twinning according to the registration result to construct the three-dimensional twin structure parameters.

[0010] In a possible implementation, traversing the three-dimensional twin structure parameters for assembly positioning includes: traversing the three-dimensional twin structure parameters to perform multi-scale spatial matching on the intelligent bed to determine the target assembly area of the intelligent bed; performing positioning according to the target assembly area to generate an assembly positioning coordinate set, and performing multi-dimensional extraction on the three-dimensional twin structure parameters according to the assembly positioning coordinate set to obtain multi-dimensional structure parameters, where the multi-dimensional structure parameters include geometric feature structure parameters, physical attribute structure parameters, and process structure parameters; correlating the geometric feature structure parameters, the physical attribute structure parameters, and the process structure parameters, establishing a correlation matrix, and constructing the assembly feature vector group according to the correlation matrix.

[0011] In a possible implementation, an assembly feature vector group is extracted for quality identification, and a quality defect map is drawn, including: retrieving a standard process parameter library, dynamically comparing the assembly feature vector group with the standard process parameter library to identify an assembly deviation pattern; performing defect probability analysis according to the assembly deviation pattern to construct a defect probability distribution map; performing defect scoring according to the assembly deviation pattern to generate a defect confidence score matrix; and performing quality identification by mapping the defect probability distribution map to the three-dimensional twin structure parameters according to the defect confidence score matrix to construct the quality defect map.

[0012] In a possible implementation, hierarchical early warning is performed based on the quality defect map, and component calibration operations are performed according to multi-level early warning information to construct an assembly quality assessment report for the intelligent bed, including: parsing the quality defect map to generate an RGB coding matrix, performing real-time grading according to the RGB coding matrix to construct multi-level early warning information, where the multi-level early warning information includes a multi-level early warning instruction set; performing component calibration operations according to the multi-level early warning instruction set in accordance with a gradient calibration operation sequence to generate real-time calibration data; and performing fusion evaluation of the real-time calibration data and the standard process parameter library to generate an assembly quality assessment report for the intelligent bed.

[0013] In a possible implementation, parsing the quality defect map to generate an RGB coding matrix, performing real-time grading according to the RGB coding matrix to construct multi-level early warning information, including: performing region segmentation on the quality defect map according to the RGB coding matrix to obtain a plurality of segmented regions; calculating the Euclidean distance between the plurality of segmented regions and the RGB coding matrix, performing warning allocation according to the plurality of region distances to generate a plurality of warning levels; making an online decision according to the plurality of warning levels to generate a multi-level early warning instruction set; and adding the multi-level early warning instruction set to the multi-level early warning information.

[0014] In a possible implementation, performing fusion evaluation of the real-time calibration data and the standard process parameter library to generate an assembly quality assessment report for the intelligent bed, including: performing spatio-temporal alignment fusion of the real-time calibration data and the standard process parameter library to generate a fusion data set; performing assembly quality interaction based on the fusion data set to construct multi-dimensional quality evaluation indicators; evaluating the fusion data set according to the multi-dimensional quality evaluation indicators to generate multi-dimensional quality scores; and adding the multi-dimensional quality scores to the assembly quality assessment report of the intelligent bed.

[0015] In a second aspect of the present application, a machine vision-driven intelligent bed assembly quality monitoring system is provided, and the system includes:

[0016] An image data acquisition module, configured to collect three-dimensional surface image data of the intelligent bed assembly through machine vision and generate an assembly image data set; a spatial registration module, configured to obtain topological structure data of the intelligent bed frame, perform spatial registration in combination with the assembly image data set, and construct three-dimensional twin structure parameters; a quality identification module, configured to traverse the three-dimensional twin structure parameters for assembly positioning, extract an assembly feature vector group for quality identification, and draw a quality defect map. Specifically, according to a defect confidence score matrix, map a defect probability distribution map to the three-dimensional twin structure parameters for quality identification and construct the quality defect map; a component calibration module, configured to perform hierarchical warning based on the quality defect map, execute component calibration operations according to multi-level warning information, and construct an assembly quality evaluation report for the intelligent bed. Specifically, based on the analysis of the quality defect map, generate an RGB coding matrix, perform real-time grading according to the RGB coding matrix, and construct multi-level warning information, where the multi-level warning information includes a multi-level warning instruction set.

[0017] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0018] This application collects three-dimensional surface image data of the intelligent bed assembly through machine vision to generate an assembly image data set; obtains topological structure data of the intelligent bed frame, performs spatial registration in combination with the assembly image data set to construct three-dimensional twin structure parameters; traverses the three-dimensional twin structure parameters for assembly positioning, extracts an assembly feature vector group for quality identification, and draws a quality defect map; performs hierarchical warning based on the quality defect map, executes component calibration operations according to multi-level warning information, and constructs an assembly quality evaluation report for the intelligent bed. This invention solves the technical problem of low accuracy in evaluating the assembly quality of intelligent beds in the prior art, and achieves the technical effect of improving the accuracy of evaluating the assembly quality of intelligent beds by combining machine vision, three-dimensional surface image data acquisition, and three-dimensional twin structure registration technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 It is a schematic flowchart of a method for monitoring the assembly quality of an intelligent bed driven by machine vision provided by an embodiment of this application;

[0021] Figure 2 It is a schematic structural diagram of a system for monitoring the assembly quality of an intelligent bed driven by machine vision provided by an embodiment of this application.

[0022] Description of the reference numerals: Image data acquisition module 11, spatial registration module 12, quality recognition module 13, component calibration module 14. Detailed implementation manners

[0023] By providing a machine vision-driven intelligent bed assembly quality monitoring method and system, the present application aims to solve the technical problem of low accuracy in the evaluation of the assembly quality of intelligent beds in the prior art. By combining machine vision, three-dimensional surface image data acquisition, and three-dimensional twin structure registration technologies, the technical effect of improving the accuracy of the assembly quality evaluation of intelligent beds is achieved.

[0024] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0025] It should be noted that any variations of the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those clearly listed steps or units, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0026] Embodiment 1, as Figure 1 shown, the present application provides a machine vision-driven intelligent bed assembly quality monitoring method, and the method includes:

[0027] Step S100: Collect three-dimensional surface image data of the intelligent bed assembly through machine vision to generate an assembly image data set.

[0028] In the embodiment of the present application, when collecting the three-dimensional surface image data of the intelligent bed assembly through machine vision, first, a plurality of trigger sensors are installed on the assembly line conveyor belt. When the intelligent bed assembly enters the shooting area, the trigger sensors will synchronously activate the multi-modal imaging module of the multi-spectral industrial camera. The multi-spectral industrial camera captures the detailed three-dimensional surface information of the intelligent bed assembly through imaging in different spectral bands. Then, continuous frame synchronization acquisition of the assembly is performed through the multi-modal imaging module to generate a composite image sequence. Finally, these image sequences are processed by dynamic enhancement to remove noise and enhance key features, generating an assembly image data set.

[0029] Furthermore, in the method provided by the embodiment of the application, collecting the three-dimensional surface image data of the intelligent bed assembly through machine vision to generate an assembly image data set further includes:

[0030] Based on the assembly line conveyor belt, a plurality of trigger sensors are set. When it is detected that the assembly of the smart bed enters the shooting area, the multimodal imaging module of the multispectral industrial camera is synchronously activated through the plurality of trigger sensors; the assembly is continuously frame-synchronously collected through the multimodal imaging module to obtain a composite image sequence; the composite image sequence is dynamically enhanced to generate the assembly image dataset.

[0031] In the embodiment of the present application, first, based on the existing assembly line conveyor belt, a plurality of trigger sensors are set. When the assembly of the smart bed passes through the conveyor belt and enters the designated shooting area, these trigger sensors will automatically sense the arrival of the assembly and synchronously activate the multimodal imaging module of the multispectral industrial camera. The multispectral industrial camera has multiple imaging modes, including visible light, near-infrared, and depth information acquisition. Through imaging in these different bands, comprehensive three-dimensional surface information of the smart bed assembly can be captured.

[0032] Then, the assembly is continuously frame-synchronously collected through the multimodal imaging module, that is, the image data of the assembly is continuously captured at a high frequency to generate a composite image sequence. The composite image sequence contains multispectral information, including not only visible light images but also near-infrared images and depth information.

[0033] Subsequently, the obtained composite image sequence is subjected to dynamic enhancement processing, and an adaptive exposure compensation technique is used to optimize the image quality. The adaptive exposure compensation technique adjusts the image exposure in real time according to the ambient light conditions, so that the images taken under different lighting environments reach the best balance in brightness and contrast, ensuring that the image details are clearly visible. Finally, the image sequence after adaptive exposure compensation is processed into a high-contrast assembly image dataset.

[0034] Step S200: Obtain the topological structure data of the smart bed frame, perform spatial registration in combination with the assembly image dataset, and construct three-dimensional twin structure parameters.

[0035] In the embodiment of the present application, to obtain the topological structure data of the smart bed frame, first, three-dimensional line laser scanning devices are deployed on both sides of the assembly station to synchronously scan the smart bed frame from multiple perspectives, thereby generating multi-perspective scan data. Through surface feature recognition based on these data, a three-dimensional coordinate system is constructed, and preliminary point cloud data is generated in this coordinate system. Then, rough registration is performed to identify and remove outlier point cloud data, and finally, multi-perspective point cloud data is generated. Map these point cloud data into the three-dimensional coordinate system, and construct the topological structure data of the smart bed frame by determining the coordinates of key connection nodes.

[0036] Next, the assembly image dataset is combined for spatial registration. Specifically, the assembly image dataset is mapped to a three-dimensional coordinate system through a composite image sequence, and an association analysis is performed to extract the association mapping relationship. The reference markers are set based on the topological structure data, the alignment parameters are calculated, and the assembly image dataset is accurately registered with the multi-view point cloud data based on these parameters. Finally, through spatial registration and data twin technology, the three-dimensional twin structure parameters are constructed to form a digital model of the smart bed assembly.

[0037] Furthermore, in the method provided in the embodiment of the application, the process of obtaining the topological structure data of the intelligent bed frame also includes:

[0038] A three-dimensional line laser scanning device is deployed on both sides of the assembly station to perform multi-perspective synchronous scanning of the smart bed frame to generate multi-perspective scanning data; surface feature recognition is performed on the smart bed frame based on the multi-perspective scanning data to construct a three-dimensional coordinate system, and the three-dimensional coordinate system includes multi-perspective initial point cloud data; coarse alignment is performed based on the multi-perspective initial point cloud data to determine multiple outlier point cloud data, and the multi-perspective initial point cloud data is eliminated according to the multiple outlier point cloud data to generate multi-perspective point cloud data; the multi-perspective point cloud data is mapped to the three-dimensional coordinate system to determine multiple key connection node coordinates, and the topological structure data of the smart bed frame is constructed according to the multiple key connection node coordinates.

[0039] In the embodiment of the present application, a three-dimensional line laser scanning device is first deployed on both sides of the assembly station to perform multi-view synchronous scanning. The three-dimensional line laser scanning device measures the spatial position of each point on the surface of the object by emitting a laser beam and receiving a reflected signal. By scanning from multiple angles at the same time, the laser scanning device captures the complete three-dimensional surface data of the smart bed frame and generates multi-view scanning data.

[0040] Next, based on the acquired multi-view scanning data, the smart bed frame is analyzed using surface feature recognition technology. The geometric features of the frame, such as edges, corners, and connection points, are identified from the scanning data through the surface feature recognition algorithm. These features are the iconic structures of the smart bed frame and are used to help build the shape and structure of the frame. Through feature recognition, a three-dimensional coordinate system is generated, which contains the multi-view initial point cloud data.

[0041] After obtaining the preliminary point cloud data, the coarse registration method is used to perform preliminary alignment of the multi-view point cloud data. The purpose of coarse registration is to ensure that data from different perspectives can be roughly aligned in space. To this end, the registration algorithm estimates the relative positions between data from different perspectives based on the geometric features of the point cloud data, thereby roughly mapping them to a unified three-dimensional coordinate system.

[0042] Next, outlier removal algorithms are used to identify and remove abnormal data points that do not conform to the actual geometric structure of the smart bed frame. By adopting distance-based removal algorithms, such as the RANSAC (Random Sample Consensus) algorithm, data points in the point cloud are randomly selected to construct a model and verify its fitness. Then, those data points that deviate from the model (i.e., outliers) are identified. These outliers are usually caused by measurement errors, environmental interference, or reflection problems and do not conform to the true geometric structure of the frame. Through continuous iterative optimization, the RANSAC algorithm finally identifies and removes these invalid points, thereby retaining the valid data that conforms to the actual shape of the smart bed frame and generating more accurate multi-viewpoint cloud data.

[0043] Then, point cloud registration methods are used to map the processed multi-viewpoint cloud data into the previously constructed three-dimensional coordinate system. Point cloud registration calculates the alignment parameters between each point cloud data set, enabling the data from each perspective to be accurately docked into a unified coordinate system, ensuring that all point cloud data is correctly aligned in space and providing accurate spatial positioning for subsequent topological structure construction.

[0044] Finally, based on the point cloud data mapped into the three-dimensional coordinate system, the coordinates of multiple key connection nodes in the smart bed frame are determined. These key connection nodes are located at the intersection points of various structural components in the smart bed frame, such as support points and connection points. According to these node coordinates, topological structure construction methods are used to generate the topological structure data of the smart bed frame. The topological structure data not only includes the spatial positions of each component but also describes the connection relationships between them.

[0045] Furthermore, in the method provided by the application embodiment, when performing spatial registration in combination with the assembly image data set to construct three-dimensional twin structure parameters, it further includes:

[0046] Mapping the assembly image data set to the three-dimensional coordinate system according to the composite image sequence for correlation analysis, and extracting the correlation mapping relationship; setting reference marker points based on the topological structure data, and calculating the alignment parameters between the assembly image data set and the multi-viewpoint cloud data according to the reference marker points; according to the alignment parameters, performing spatial registration on the assembly image data set and the multi-viewpoint cloud data according to the correlation mapping relationship, and performing data twinning according to the registration result to construct the three-dimensional twin structure parameters.

[0047] In the embodiment of the present application, first, according to the composite image sequence, the assembly image dataset is mapped to the constructed three-dimensional coordinate system. This step realizes the docking of the image data with the point cloud data in the three-dimensional coordinate system through the feature matching between the image and the point cloud data. The texture features in the image (such as color, brightness, pattern, etc.) and the geometric features in the point cloud (i.e., the three-dimensional surface structure of the object) are matched through a specific feature matching algorithm (such as SIFT, SURF) to extract the correlation mapping relationship between the two. Through this process, it is ensured that the image data and the point cloud data can be accurately corresponding in the three-dimensional space.

[0048] Then, based on the topological structure data, the alignment parameters between the image data and the point cloud data are calculated by setting reference marker points. These reference marker points are usually located at the key connection points in the intelligent bed frame, such as the support points and the connection parts of the structural components, serving as references for spatial alignment. Using geometric calculation methods, through the least squares method or the ICP (Iterative Closest Point) algorithm, the spatial transformation parameters such as translation and rotation between the image data and the point cloud data are calculated based on the reference marker points, so as to obtain the alignment parameters and ensure the accurate docking of the two in the three-dimensional coordinate system.

[0049] Once the alignment parameters are obtained, according to these alignment parameters and the correlation mapping relationship, the assembly image dataset and the multi-viewpoint cloud data are spatially registered. The process of spatial registration is to accurately align the image data and the point cloud data to the same three-dimensional coordinate system, so that they can perform data fusion in the same coordinate system. This process uses a registration algorithm, and through an optimization algorithm (such as the method based on the iterative closest point), the image data is docked with the point cloud data to ensure that their positions, scales, and directions in space are exactly the same. Finally, according to the registered result, data twin is performed to generate the three-dimensional twin structure parameters of the intelligent bed assembly.

[0050] Step S300: Traverse the three-dimensional twin structure parameters for assembly positioning, extract the assembly feature vector group for quality identification, and draw a quality defect map, wherein, according to the defect confidence score matrix, the defect probability distribution map is mapped to the three-dimensional twin structure parameters for quality identification, and the quality defect map is constructed.

[0051] In the embodiments of the present application, by traversing the three-dimensional twin structure parameters, multi-scale spatial matching is performed on the intelligent bed to determine the target assembly area. By locating these areas, an assembly positioning coordinate set is generated, and based on these coordinates, multi-dimensional extraction of the three-dimensional twin structure parameters is performed to obtain multi-dimensional structure parameters including geometric feature structure parameters, physical property structure parameters, and process structure parameters. Next, an assembly feature vector group is extracted. These feature vectors synthesize the aforementioned multi-dimensional structure parameters and represent the key information in the assembly process of the intelligent bed. By performing quality identification on these feature vectors and dynamically comparing them with the standard process parameter library, potential assembly deviation patterns such as size non-conformance and position deviation are identified.

[0052] Then, according to the assembly deviation pattern, defect probability analysis is performed, and a defect probability distribution map is generated to show the probability of possible defects in each assembly part. Next, defect scoring is performed based on the defect pattern to generate a defect confidence scoring matrix to quantify the severity of each defect. Finally, in combination with the defect confidence scoring matrix, the defect probability distribution map is mapped into the three-dimensional twin structure parameters to draw a quality defect map.

[0053] Furthermore, in the method provided by the application embodiments, when traversing the three-dimensional twin structure parameters for assembly positioning, it further includes:

[0054] Traverse the three-dimensional twin structure parameters to perform multi-scale spatial matching on the intelligent bed to determine the target assembly area of the intelligent bed; perform positioning according to the target assembly area to generate an assembly positioning coordinate set, and perform multi-dimensional extraction of the three-dimensional twin structure parameters according to the assembly positioning coordinate set to obtain multi-dimensional structure parameters, where the multi-dimensional structure parameters include geometric feature structure parameters, physical property structure parameters, and process structure parameters; correlate the geometric feature structure parameters, the physical property structure parameters, and the process structure parameters to establish a correlation matrix, and construct the assembly feature vector group according to the correlation matrix.

[0055] In the embodiments of the present application, first, traverse the three-dimensional twin structure parameters to perform multi-scale spatial matching on the intelligent bed. In this step, a multi-scale matching algorithm is used to gradually analyze the three-dimensional model of the intelligent bed. This process starts from the overall structure at a larger scale and gradually transitions to the detail matching at a smaller scale to ensure that each part of the intelligent bed can be accurately identified at different spatial scales. In this way, the target assembly area of the intelligent bed is determined, that is, the area that needs special attention and precise adjustment during the assembly process. For example, identify the assembly areas of the bed frame, bed board, and connectors to ensure that these key parts can be accurately docked during assembly.

[0056] Next, according to the identified target assembly area, assembly positioning is carried out by means of geometric positioning methods. In this step, geometric calculation and spatial transformation techniques are used to determine the exact position of the area in the three-dimensional coordinate system based on the dimensions and spatial coordinates of the target assembly area. Assuming that the connection part between the bed frame and the bed board has specific assembly tolerances (for example, the gap between the bed frame and the bed board is 5 millimeters), these tolerances are calculated to accurately position the assembly and generate an assembly positioning coordinate set, which represents the exact positions of the various assembly components of the smart bed in three-dimensional space.

[0057] After completing the assembly positioning, a multi-dimensional feature extraction method is used to extract the three-dimensional twin structure parameters. In this step, through feature extraction algorithms (such as based on three-dimensional modeling technology), key feature information of the smart bed assembly is extracted from the three-dimensional model, including geometric feature structure parameters (such as the dimensions, shapes, and angles of the bed frame, etc.), physical property structure parameters (such as the hardness, density, etc. of the bed frame material), and process structure parameters (such as assembly tolerances, manufacturing errors, etc.). For example, the thickness of the bed board and the length of the bed frame can be used as geometric features, while the hardness and elastic modulus of the bed frame belong to physical properties. By extracting these multi-dimensional structure parameters, detailed data information is provided for subsequent quality monitoring and optimization design.

[0058] Then, through the correlation analysis method, the geometric feature structure parameters, physical property structure parameters, and process structure parameters are correlated. In this process, statistical methods such as regression analysis or principal component analysis (PCA) are applied to identify the mutual relationships between different types of structure parameters, and an association matrix is constructed through these relationships. This matrix quantifies the mutual dependence and influence degree between each parameter. For example, it may be found that there is a strong correlation between the dimensions of the bed frame and the thickness of the bed board, with a correlation coefficient of 0.9, indicating that these two parameters have a greater mutual influence during the assembly process.

[0059] Finally, based on the association matrix, a feature vector construction method is used to generate an assembly feature vector group. In this process, first, the association matrix quantifies the mutual relationships between the various structure parameters in the smart bed assembly, and these structure parameters include the geometric features (such as the dimensions of the bed frame) of the smart bed, physical properties (such as the hardness and density of the material), and process features (such as assembly tolerances). Then, through the feature vector construction method, these structure parameters are transformed into assembly feature vectors, that is, the numerical representation of each structure parameter is expressed as a component in the vector. Through this process, an assembly feature vector group is generated.

[0060] Furthermore, in the method provided by the application embodiment, when extracting the assembly feature vector group for quality identification and drawing a quality defect map, it further includes:

[0061] Retrieve a standard process parameter library, dynamically compare the assembly feature vector group with the standard process parameter library, and identify the assembly deviation pattern; perform defect probability analysis based on the assembly deviation pattern, and construct a defect probability distribution map; perform defect scoring based on the assembly deviation pattern, and generate a defect confidence scoring matrix; map the defect probability distribution map to the three-dimensional twin structure parameters based on the defect confidence scoring matrix for quality identification, and construct the quality defect map.

[0062] In the embodiment of the present application, the standard process parameter library is first retrieved, which contains process parameters such as standard size, material properties, assembly tolerance, etc. of the smart bed assembly. The parameter library is accessed through the standard process data query method to obtain standard data related to the smart bed assembly.

[0063] Then, the assembly feature vector group is dynamically compared with the data in the standard process parameter library. The dynamic comparison process uses a hierarchical analysis method to perform multi-level analysis on the geometric features, physical properties and process parameters of the assembly feature vector group. For example, assume that the size of the bed frame is 2000mm×1500mm, and the bed frame size tolerance given in the standard process parameter library is ±1mm. Compare the actual measured bed frame size with the standard tolerance and calculate the difference between the two. To quantify these differences, the cosine similarity method is used to calculate the similarity between the assembly feature vector and the standard parameters. If the similarity is lower than the pre-set threshold, it is marked as an assembly deviation pattern. For example, if the length of the bed frame is different from the standard size, this deviation pattern is identified and its location is indicated.

[0064] After identifying the assembly deviation pattern, a defect probability analysis is performed. This process combines historical data with current assembly data and uses probability statistics to calculate the probability of defects in each component or area. Assume that historical data shows that when the size deviation of the bed frame is ±1mm, the probability of assembly failure is 30%. Compare the current assembly data with the historical data. If the current size deviation of the bed frame is 1mm, it is considered that the probability of current assembly failure is 30%. Through this process, the defect probability analysis is completed and a defect probability distribution map is produced. This map intuitively shows the defect risk areas of each component of the smart bed.

[0065] Then, according to the assembly deviation pattern, the defect score is performed, and a defect confidence score matrix is ​​generated. In the scoring process, multiple factors such as geometric deviation (such as excessive assembly tolerance, unqualified size), texture abnormality (such as uneven surface or missing coating) and stress distribution (such as uneven local stress caused by assembly) are considered. Assume that the size deviation of the bed frame is 1mm and there are texture abnormalities (such as scratches) on the surface. Then, according to the assembly deviation pattern, the defect score is performed, and a defect confidence score matrix is ​​generated. In this process, various factors affecting the assembly quality are considered, including geometric deviation (such as unqualified size, excessive assembly tolerance), texture abnormality (such as uneven surface, missing coating) and stress distribution (such as local stress concentration caused by improper assembly). By quantifying and scoring these factors, the defect risk of each component is evaluated. For example, when analyzing the assembly quality of the bed frame, its geometric deviation is calculated first. Assume that the size deviation of the bed frame is 1mm, while the tolerance required by the standard process is ±1mm. According to the preset scoring criteria, although the deviation exceeds 50% of the tolerance, it is still within the allowable range, so the geometric deviation score is 0.6, indicating that the impact of the deviation is small. Then consider the texture anomaly, assuming that there are slight scratches on the surface of the bed frame. According to the preset rule table, it is considered that the scratches have little impact on the appearance and function of the bed frame, so the texture anomaly score is 0.4. Finally, the stress distribution of the bed frame is analyzed. It is assumed that there is a high stress concentration at the connection part of the bed frame, which may affect the long-term stability of the assembly. According to the degree of stress concentration, the stress distribution score is given as 0.7, indicating that the defect is more serious. According to these scores, the impact of each defect is comprehensively calculated, and finally the total defect score of the bed frame is obtained by simple averaging. The defect confidence score matrix is ​​obtained by performing similar defect scoring calculations on all assembly parts.

[0066] Finally, according to the defect confidence score matrix, the defect probability distribution map is mapped to the three-dimensional twin structure parameters for quality identification. Through the three-dimensional data mapping method, the defect information is connected to the three-dimensional digital model of the smart bed, so that the defect area can be visualized in the three-dimensional model. For example, the defect area of ​​the bed frame is marked in red in the three-dimensional model, indicating that there is a high risk of defects in this area. Through this mapping, quality problems in the assembly are displayed in real time, and quality control personnel are helped to accurately locate the areas that need to be corrected. Finally, a quality defect map is generated.

[0067] Step S400: Perform graded warning based on the quality defect map, execute component calibration operation according to the multi-level warning information, and construct an assembly quality assessment report for the smart bed, wherein analysis is performed based on the quality defect map, an RGB coding matrix is ​​generated, real-time grading is performed according to the RGB coding matrix, and multi-level warning information is constructed, and the multi-level warning information includes a multi-level warning instruction set.

[0068] In the embodiment of the present application, firstly, the quality defect map is analyzed to generate an RGB coding matrix. The matrix encodes the defective areas of each assembly component with RGB colors according to different defect risks, and each color represents a different defect level. According to the RGB coding matrix, real-time grading is performed, and each component in the assembly is divided into multiple levels according to the severity of the defect, and corresponding multi-level warning information is generated. The warning information includes multiple warning instruction sets, each instruction set corresponds to a different defect risk level.

[0069] Then, according to the warning instruction set in the multi-level warning information, the component calibration operation is performed. In this process, each component in the assembly is adjusted according to the gradient calibration operation sequence to ensure that it meets the standard process requirements. The calibration operation is performed step by step according to the warning level, and the higher priority defective areas are calibrated first, thereby improving the overall quality of the assembly. During the calibration process, real-time calibration data is recorded and generated in real time.

[0070] Finally, the generated real-time calibration data is integrated and evaluated with the data in the standard process parameter library to ensure that the assembly meets all standard process requirements. Through this step, the assembly quality evaluation report of the smart bed is generated.

[0071] Furthermore, in the method provided in the embodiment of the application, a graded warning is performed based on the quality defect map, a component calibration operation is performed according to the multi-level warning information, and an assembly quality assessment report of the smart bed is constructed, which also includes:

[0072] Based on the quality defect map, analysis is performed to generate an RGB coding matrix, and real-time grading is performed according to the RGB coding matrix to construct multi-level warning information, wherein the multi-level warning information includes a multi-level warning instruction set; based on the multi-level warning instruction set, component calibration operations are performed according to a gradient calibration operation sequence to generate real-time calibration data; the real-time calibration data is fused and evaluated with the standard process parameter library to generate an assembly quality assessment report for the smart bed.

[0073] In the embodiment of the present application, firstly, the quality defect map is analyzed to generate an RGB coding matrix. The quality defect map graphically represents the defect areas of each component of the smart bed assembly, and intuitively identifies the quality problems in the assembly. The RGB coding matrix maps these defect areas to different RGB color values ​​according to their severity, and the depth of the color represents the severity of the defect. For example, red represents high-risk areas and green represents low-risk areas.

[0074] Then, real-time grading is performed based on the RGB coding matrix. This process uses a grading algorithm to divide the parts of the assembly into different defect levels according to the color of each area in the matrix (i.e., the severity of the defect). Based on this grading information, multi-level warning information is constructed, including multi-level warning instruction sets. These instruction sets will indicate how to take different treatment measures according to the severity of the defect. High-priority warning instructions are for more serious defects (such as large deviations in the size of the bed frame), while low-priority instructions are for minor defects (such as minor scratches on the surface), guiding quality control personnel to handle them according to priority.

[0075] After the multi-level warning information is generated, the multi-level warning instruction set performs component calibration operations according to the gradient calibration operation sequence. The gradient calibration operation sequence is implemented through a phased calibration process, giving priority to high-priority defect areas and then gradually resolving low-priority defects. The calibration process includes the use of automated calibration tools (such as robotic arms, laser measuring equipment, etc.) for precise size adjustment, position correction or surface repair. Through these operations, real-time calibration data is generated to record the specific parameter changes during the adjustment process of each component, such as bed frame size correction, surface coating repair, and assembly tolerance adjustment.

[0076] After the calibration is completed, the real-time calibration data is fused and evaluated with the data in the standard process parameter library. The standard process parameter library contains parameters such as standard dimensions, material properties, and tolerances for smart bed assembly. The spatiotemporal alignment fusion method is used to ensure the consistency of the real-time calibration data and the standard process data. By comparing these data, it is confirmed whether the calibrated parts meet the design requirements, and an assembly quality assessment report for the smart bed is generated. The report includes all assembly quality data, including defect risks, calibration results, and component quality status.

[0077] Furthermore, in the method provided in the embodiment of the application, parsing is performed based on the quality defect map to generate an RGB coding matrix, and real-time grading is performed according to the RGB coding matrix to construct multi-level warning information, which also includes:

[0078] The quality defect map is segmented into regions according to the RGB coding matrix to obtain multiple segmented regions; the Euclidean distances between the multiple segmented regions and the RGB coding matrix are calculated, warnings are allocated according to the multiple region distances, and multiple warning levels are generated; online decisions are made according to the multiple warning levels to generate a multi-level warning instruction set; and the multi-level warning instruction set is added to the multi-level warning information.

[0079] In an embodiment of the present application, the quality defect map is first segmented based on the RGB coding matrix. The quality defect map is generated by combining the three-dimensional model data, sensor data and image data of the assembly, and the map presents the defect risk of each component in the assembly. For example, the red area indicates a serious defect, the green area indicates no defect, and other colors represent different degrees of risk. The RGB coding matrix represents these defect risk areas with color coding, and performs regional segmentation according to the color difference of each area, thereby dividing the entire assembly into multiple segmented areas.

[0080] Then calculate the Euclidean distance between each segmented area and the RGB encoding matrix. In this scenario, each pixel in the RGB encoding matrix represents a color, and the color of each segmented area represents the severity of the defect. The severity of the defect is quantified by calculating the Euclidean distance between the color of the area and the color value of the matrix. If the color of the area deviates greatly from the color required by the standard (for example, the area is red, and the standard is green), then the Euclidean distance of the area is large, indicating that the defect in the area is more serious. Conversely, a smaller color difference indicates that the defect in the area is less serious.

[0081] Alerts are assigned based on the calculated Euclidean distance. The purpose of alert assignment is to assign a corresponding alert level to each segmented area based on the severity of the defect (i.e., the Euclidean distance). Specifically, the alert level is assigned according to a preset standard. For example, a threshold is set. When the Euclidean distance is greater than a certain value, it indicates that the defect is serious and a first-level alert (high priority) is assigned. When the Euclidean distance is at a certain intermediate value, a second-level alert is assigned, and areas with a smaller Euclidean distance are assigned a third-level alert.

[0082] Next, the generation of multi-level warning instruction sets is based on the warning level of each area. Each warning level corresponds to different treatment measures to ensure that quality control work responds according to the severity of the defect. For example, the first-level warning (high priority) will automatically trigger the assembly deviation correction equipment, such as the six-axis linkage calibration module of the robot arm, for automatic calibration and correction; the second-level warning (medium priority) involves some process defects and requires manual intervention, such as manually adjusting the size of the bed frame or repairing surface defects; the third-level warning (low priority) may only trigger a warning message without immediate repair. The operator may be prompted through an audible and visual alarm device, and a work order containing a defect snapshot will be generated and pushed to the maintenance terminal.

[0083] After the multi-level warning instruction sets are generated, they are added to the multi-level warning information to form a warning response plan. The warning information not only contains the defect level of each component, but also clarifies the processing instructions at each level. Quality control personnel can prioritize high-risk components based on this information. For example, for the first-level warning, assembly deviations are automatically corrected to avoid further quality problems, while for the second-level warning, maintenance personnel will be reminded to take manual repair operations to ensure the stability of assembly quality.

[0084] Furthermore, in the method provided in the embodiment of the application, the real-time calibration data is integrated and evaluated with the standard process parameter library to generate an assembly quality evaluation report of the smart bed, and the method further includes:

[0085] The real-time calibration data is fused with the standard process parameter library in time and space to generate a fused data set; assembly quality interaction is performed based on the fused data set to construct a multidimensional quality evaluation index; the fused data set is evaluated according to the multidimensional quality evaluation index to generate a multidimensional quality score; and the multidimensional quality score is added to the assembly quality evaluation report of the smart bed.

[0086] In an embodiment of the present application, the real-time calibration data is first docked with the data in the standard process parameter library through a spatiotemporal alignment fusion method. Real-time calibration data refers to the adjustment data of each component recorded during the assembly process, including size correction, position correction, etc., while the standard process parameter library contains process data such as standard size, material properties, assembly tolerances, etc. for the assembly of smart beds. Through spatiotemporal alignment, it is ensured that the real-time data and the standard process data are accurately docked in time and space. A fused data set is generated through this docking, so that the calibration data and the standard process data can be effectively compared.

[0087] After data fusion is completed, the fused dataset is used for assembly quality interaction. The goal of this process is to identify potential quality issues in the assembly process by comparing the actual data of each component with the standard process parameters. Assembly quality interaction is based on the comparison of standard process parameters and real-time calibration data to automatically identify potential issues such as geometric deviations, material property problems, and process tolerances that may occur during the assembly process. For example, assume that the dimensional error of the bed frame is 0.8 mm, while the standard tolerance range is ±1 mm. The dimensional error is evaluated based on preset rules. When evaluating the dimensional deviation of the bed frame, it is first quantified according to the preset rules. Assume that the standard tolerance is ±1 mm, and the preset scoring rule is that if the deviation is less than 50% of the tolerance range (i.e., less than 0.5 mm), the score is 0.2, indicating a minor defect; if the deviation is within the tolerance range (i.e., greater than 0.5 mm and less than 1 mm), the score is 0.6, indicating a medium defect; if the deviation exceeds the tolerance range (i.e., greater than 1 mm), the score is 0.8, indicating a major defect. In this example, the dimensional deviation of the bed frame is 0.8 mm, which exceeds half of the standard tolerance range. Therefore, according to the preset rules, a relatively high quality score (such as 0.6) is assigned to this component. This score reflects the degree of influence of the assembly deviation on the overall assembly quality, and a higher score indicates that this issue needs to be prioritized for handling.

[0088] Next, a comprehensive evaluation of the fused dataset is performed based on multi-dimensional quality evaluation indicators to generate multi-dimensional quality scores. This step comprehensively considers geometric deviations, physical properties (such as material hardness, density, etc.), and process tolerances (such as assembly tolerances, errors, etc.). Each evaluation indicator is quantified according to preset rules. Based on the actual data of each dimension and the preset rules, scores are assigned to each indicator respectively. Assume that the geometric deviation score of the bed frame is 0.6 (because the dimensional deviation of the bed frame is 0.8 mm), the texture abnormality score is 0.4 (due to slight scratches on the surface), and the stress distribution score is 0.5 (because there is stress concentration at the connection part of the bed frame). Finally, the multi-dimensional quality score of the bed frame is calculated using the simple average method based on these scores of each dimension, that is, (0.6 + 0.4 + 0.5) / 3 = 0.5.

[0089] Finally, the multi-dimensional quality score is added to the assembly quality evaluation report of the smart bed, and the report details the quality status and scores of each component.

[0090] In the embodiments of this application, in summary, the embodiments of this application have at least the following technical effects:

[0091] This application collects three-dimensional surface image data of the smart bed assembly through machine vision to generate an assembly image data set; obtains the topological structure data of the smart bed frame, combines the assembly image data set for spatial alignment, and constructs three-dimensional twin structure parameters; traverses the three-dimensional twin structure parameters for assembly positioning, extracts the assembly feature vector group for quality identification, and draws a quality defect map; performs graded warnings based on the quality defect map, performs component calibration operations according to multi-level warning information, and constructs an assembly quality assessment report for the smart bed. The present invention solves the technical problem of low accuracy in smart bed assembly quality assessment in the prior art, and achieves the technical effect of improving the accuracy of smart bed assembly quality assessment by combining machine vision, three-dimensional surface image data acquisition, and three-dimensional twin structure alignment technology.

[0092] Embodiment 2, based on the same inventive concept as the machine vision driven smart bed assembly quality monitoring method in the aforementioned embodiment, Figure 2 As shown, the present application provides a machine vision driven smart bed assembly quality monitoring system, and the system and method embodiments in the present application embodiments are based on the same inventive concept. The system includes:

[0093] An image data acquisition module 11 is used to acquire three-dimensional surface image data of the smart bed assembly through machine vision to generate an assembly image data set; a spatial registration module 12 is used to obtain the topological structure data of the smart bed frame, perform spatial registration in combination with the assembly image data set, and construct three-dimensional twin structure parameters; a quality identification module 13 is used to traverse the three-dimensional twin structure parameters for assembly positioning, extract the assembly feature vector group for quality identification, and draw a quality defect map, wherein the defect probability distribution map is mapped to the three-dimensional twin structure parameters for quality identification according to the defect confidence scoring matrix to construct the quality defect map; a component calibration module 14 is used to perform graded warning based on the quality defect map, perform component calibration operations according to the multi-level warning information, and construct an assembly quality assessment report for the smart bed, wherein analysis is performed based on the quality defect map to generate an RGB coding matrix, real-time grading is performed according to the RGB coding matrix, and multi-level warning information is constructed, and the multi-level warning information includes a multi-level warning instruction set.

[0094] Furthermore, the system is also used to implement the following functions:

[0095] A plurality of trigger sensors are arranged based on the conveyor belt of the assembly line. When it is detected that the assembly of the smart bed enters the shooting area, the multimodal imaging module of the multispectral industrial camera is synchronously activated by the plurality of trigger sensors. The assembly is continuously frame-synchronously acquired by the multimodal imaging module to obtain a composite image sequence. The composite image sequence is dynamically enhanced to generate the assembly image data set.

[0096] Furthermore, the system is also used to implement the following functions:

[0097] A three-dimensional line laser scanning device is deployed on both sides of the assembly station to perform multi-perspective synchronous scanning of the smart bed frame to generate multi-perspective scanning data; surface feature recognition is performed on the smart bed frame based on the multi-perspective scanning data to construct a three-dimensional coordinate system, and the three-dimensional coordinate system includes multi-perspective initial point cloud data; coarse alignment is performed based on the multi-perspective initial point cloud data to determine multiple outlier point cloud data, and the multi-perspective initial point cloud data is eliminated according to the multiple outlier point cloud data to generate multi-perspective point cloud data; the multi-perspective point cloud data is mapped to the three-dimensional coordinate system to determine multiple key connection node coordinates, and the topological structure data of the smart bed frame is constructed according to the multiple key connection node coordinates.

[0098] Furthermore, the system is also used to implement the following functions:

[0099] According to the composite image sequence, the assembly image dataset is mapped to the three-dimensional coordinate system for association analysis, and the association mapping relationship is extracted; based on the topological structure data, reference mark points are set, and the alignment parameters of the assembly image dataset and the multi-view point cloud data are calculated according to the reference mark points; according to the alignment parameters and the association mapping relationship, the assembly image dataset and the multi-view point cloud data are spatially registered, and data twinning is performed according to the registration result to construct the three-dimensional twin structure parameters.

[0100] Furthermore, the system is also used to implement the following functions:

[0101] The three-dimensional twin structure parameters are traversed to perform multi-scale spatial matching on the smart bed to determine the target assembly area of ​​the smart bed; positioning is performed according to the target assembly area to generate an assembly positioning coordinate set, and multi-dimensional extraction is performed on the three-dimensional twin structure parameters according to the assembly positioning coordinate set to obtain multi-dimensional structure parameters, which include geometric feature structure parameters, physical property structure parameters, and process structure parameters; the geometric feature structure parameters, the physical property structure parameters, and the process structure parameters are associated to establish an association matrix, and the assembly feature vector group is constructed according to the association matrix.

[0102] Furthermore, the system is also used to implement the following functions:

[0103] Retrieve the standard process parameter library, dynamically compare the assembly feature vector group with the standard process parameter library to identify the assembly deviation pattern; perform defect probability analysis according to the assembly deviation pattern to construct a defect probability distribution map; perform defect scoring according to the assembly deviation pattern to generate a defect confidence score matrix; map the defect probability distribution map to the three-dimensional twin structure parameters according to the defect confidence score matrix for quality identification to construct the quality defect map.

[0104] Furthermore, the system is also used to implement the following functions:

[0105] Based on the analysis of the quality defect map, generate an RGB coding matrix, perform real-time grading according to the RGB coding matrix to construct multi-level warning information, and the multi-level warning information includes a multi-level warning instruction set; perform component calibration operations based on the multi-level warning instruction set according to the gradient calibration operation sequence to generate real-time calibration data; fuse and evaluate the real-time calibration data with the standard process parameter library to generate an assembly quality evaluation report for the intelligent bed.

[0106] Furthermore, the system is also used to implement the following functions:

[0107] Perform region segmentation on the quality defect map according to the RGB coding matrix to obtain multiple segmented regions; calculate the Euclidean distance between the multiple segmented regions and the RGB coding matrix, perform warning allocation according to the multiple regional distances to generate multiple warning levels; make an online decision according to the multiple warning levels to generate a multi-level warning instruction set; add the multi-level warning instruction set to the multi-level warning information.

[0108] Furthermore, the system is also used to implement the following functions:

[0109] Perform spatio-temporal alignment and fusion of the real-time calibration data with the standard process parameter library to generate a fusion data set; perform assembly quality interaction based on the fusion data set to construct multi-dimensional quality evaluation indicators; evaluate the fusion data set according to the multi-dimensional quality evaluation indicators to generate multi-dimensional quality scores; add the multi-dimensional quality scores to the assembly quality evaluation report of the intelligent bed.

[0110] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above description of specific embodiments of this specification is made. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multi-tasking and parallel processing are also possible or may be advantageous.

[0111] The above are only the preferred embodiments of the present application, and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0112] This specification and the drawings are merely exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications therein.

Claims

1. A machine vision-driven smart bed assembly quality monitoring method, characterized in that: The method comprises: Collect the three-dimensional surface image data of the intelligent bed assembly through machine vision to generate an assembly image data set; Acquire the topological structure data of the intelligent bed frame, perform spatial registration in combination with the assembly image data set, and construct three-dimensional twin structure parameters; Traversing the three-dimensional twin structure parameters for assembly positioning, extracting the assembly feature vector group for quality identification, and drawing a quality defect map, wherein the defect probability distribution map is mapped to the three-dimensional twin structure parameters for quality identification according to the defect confidence score matrix to construct the quality defect map; Based on the quality defect map, graded warnings are performed, component calibration operations are performed according to the multi-level warning information, and an assembly quality assessment report for the smart bed is constructed, wherein analysis is performed based on the quality defect map, an RGB coding matrix is ​​generated, real-time grading is performed according to the RGB coding matrix, and multi-level warning information is constructed, and the multi-level warning information includes a multi-level warning instruction set.

2. The machine vision driven smart bed assembly quality monitoring method according to claim 1, characterized in that: The three-dimensional surface image data of the intelligent bed assembly is collected by machine vision to generate an assembly image data set, and the method includes: A plurality of trigger sensors are arranged based on the assembly line conveyor belt, and when it is detected that the assembly of the intelligent bed enters the shooting area, the multimodal imaging module of the multispectral industrial camera is synchronously activated by the plurality of trigger sensors; Performing continuous frame synchronization acquisition on the assembly through the multimodal imaging module to obtain a composite image sequence; The composite image sequence is dynamically enhanced to generate the assembly image data set.

3. The machine vision driven smart bed assembly quality monitoring method according to claim 2, characterized in that: The process of obtaining the topological structure data of the intelligent bed frame includes: Deploy 3D line laser scanning devices on both sides of the assembly station to perform multi-view synchronous scanning of the smart bed frame to generate multi-view scanning data; Based on the multi-view scanning data, surface features of the smart bed frame are identified to construct a three-dimensional coordinate system, wherein the three-dimensional coordinate system includes multi-view initial point cloud data; Performing rough registration according to the multi-view initial point cloud data, determining a plurality of outlier point cloud data, removing the multi-view initial point cloud data according to the plurality of outlier point cloud data, and generating multi-view point cloud data; The multi-view point cloud data is mapped to the three-dimensional coordinate system, multiple key connection node coordinates are determined, and the topological structure data of the smart bed frame is constructed according to the multiple key connection node coordinates.

4. The machine vision driven smart bed assembly quality monitoring method according to claim 3, characterized in that: The assembly image data set is combined for spatial registration to construct three-dimensional twin structure parameters, the method comprising: According to the composite image sequence, the assembly image data set is mapped to the three-dimensional coordinate system for association analysis, and an association mapping relationship is extracted; Setting reference marking points based on the topological structure data, and calculating alignment parameters of the assembly image data set and the multi-view point cloud data according to the reference marking points; According to the alignment parameters and the associative mapping relationship, the assembly image data set and the multi-view point cloud data are spatially registered, and data twinning is performed according to the registration result to construct the three-dimensional twin structure parameters.

5. The machine vision driven smart bed assembly quality monitoring method according to claim 1, characterized in that: Traversing the three-dimensional twin structure parameters for assembly positioning, the method includes: Traversing the three-dimensional twin structure parameters to perform multi-scale spatial matching on the smart bed to determine a target assembly area of ​​the smart bed; Positioning is performed according to the target assembly area to generate an assembly positioning coordinate set, and multi-dimensional extraction is performed on the three-dimensional twin structure parameters according to the assembly positioning coordinate set to obtain multi-dimensional structure parameters, wherein the multi-dimensional structure parameters include geometric feature structure parameters, physical property structure parameters, and process structure parameters; The geometric feature structure parameters, the physical property structure parameters, and the process structure parameters are associated to establish an association matrix, and the assembly feature vector group is constructed according to the association matrix.

6. The machine vision driven smart bed assembly quality monitoring method according to claim 1, characterized in that: Extracting assembly feature vector groups for quality identification and drawing quality defect maps, the methods include: Retrieving a standard process parameter library, dynamically comparing the assembly feature vector group with the standard process parameter library, and identifying an assembly deviation pattern; Perform defect probability analysis according to the assembly deviation pattern and construct a defect probability distribution diagram; Perform defect scoring according to the assembly deviation mode to generate a defect confidence scoring matrix; According to the defect confidence score matrix, the defect probability distribution diagram is mapped to the three-dimensional twin structure parameters for quality identification to construct the quality defect map.

7. The machine vision driven smart bed assembly quality monitoring method according to claim 6, characterized in that: Based on the quality defect map, a graded warning is performed, a component calibration operation is performed according to the multi-level warning information, and an assembly quality assessment report of the smart bed is constructed. The method includes: Analyze based on the quality defect map, generate an RGB coding matrix, perform real-time classification according to the RGB coding matrix, and construct multi-level warning information, wherein the multi-level warning information includes a multi-level warning instruction set; Based on the multi-level warning instruction set, perform component calibration operations according to a gradient calibration operation sequence to generate real-time calibration data; The real-time calibration data is integrated and evaluated with the standard process parameter library to generate an assembly quality evaluation report for the smart bed.

8. The machine vision driven smart bed assembly quality monitoring method according to claim 7, characterized in that: Based on the quality defect map, the method is analyzed to generate an RGB coding matrix, and real-time classification is performed according to the RGB coding matrix to construct multi-level warning information, including: Performing regional segmentation on the quality defect map according to the RGB coding matrix to obtain a plurality of segmented regions; Calculating the Euclidean distances between the multiple segmented regions and the RGB coding matrix, performing warning allocation according to the multiple region distances, and generating multiple warning levels; Make online decisions based on the multiple warning levels and generate a multi-level warning instruction set; The multi-level warning instruction set is added to the multi-level warning information.

9. The machine vision driven smart bed assembly quality monitoring method according to claim 7, characterized in that: The real-time calibration data is integrated and evaluated with the standard process parameter library to generate an assembly quality evaluation report of the smart bed, the method comprising: Performing spatiotemporal alignment fusion of the real-time calibration data and the standard process parameter library to generate a fused data set; Perform assembly quality interaction based on the fused data set to construct a multi-dimensional quality assessment index; Evaluate the fused data set according to the multidimensional quality evaluation index to generate a multidimensional quality score; The multidimensional quality score is added to the assembly quality assessment report of the smart bed.

10. The machine vision driven smart bed assembly quality monitoring system is characterized by: The system is used to implement the machine vision driven smart bed assembly quality monitoring method according to any one of claims 1 to 9, and the system comprises: An image data acquisition module is used to acquire three-dimensional surface image data of the intelligent bed assembly through machine vision to generate an assembly image data set; A spatial registration module is used to obtain the topological structure data of the intelligent bed frame, perform spatial registration in combination with the assembly image data set, and construct three-dimensional twin structure parameters; A quality identification module, used to traverse the three-dimensional twin structure parameters for assembly positioning, extract the assembly feature vector group for quality identification, and draw a quality defect map, wherein the defect probability distribution map is mapped to the three-dimensional twin structure parameters for quality identification according to the defect confidence score matrix to construct the quality defect map; A component calibration module is used to perform graded warning based on the quality defect map, execute component calibration operations according to multi-level warning information, and construct an assembly quality assessment report for the smart bed, wherein analysis is performed based on the quality defect map, an RGB coding matrix is ​​generated, real-time grading is performed according to the RGB coding matrix, and multi-level warning information is constructed, and the multi-level warning information includes a multi-level warning instruction set.

Images