An optical sorting method and system for door and window hardware accessories

Abstract

This invention relates to the field of optical sorting technology, and discloses a method and system for optical sorting of door and window hardware accessories. The method includes: illuminating the door and window hardware accessories during transmission with a multispectral light source array, and simultaneously acquiring reflectance spectral images to obtain an optical dataset; separating and converting the original broadband image data into electrical signals, and analyzing the characteristic absorption peaks and reflectance distribution to obtain material characteristic information; reconstructing the three-dimensional morphology of the surface through stereoscopic vision triangulation to obtain geometric contour information; matching the material characteristic information with a standard spectral template to output a material identification conclusion; and performing morphological structure analysis on the geometric contour information to obtain a surface integrity judgment conclusion; applying multi-level decision rules to fuse the material identification conclusion and the surface integrity judgment conclusion to generate sorting control instructions; and encoding the sorting control instructions into pulse signals to complete the sorting of door and window hardware accessories. This invention can improve the efficiency of optical sorting of door and window hardware accessories.

Description

Technical Field

[0001] This invention relates to the field of optical sorting technology, and in particular to an optical sorting method and system for door and window hardware accessories. Background Technology

[0002] In existing door and window hardware sorting technologies, most rely on single-spectral detection methods or manual operation. Single-spectral methods are insufficient to comprehensively capture the characteristic absorption peaks and reflectance distributions of different materials, resulting in low material identification accuracy and a tendency for misjudgments and omissions. This makes them unsuitable for the precise sorting needs of multi-material components. Furthermore, the large number of manual interventions not only increases labor costs but also makes the results susceptible to human error, leading to poor consistency and difficulty in consistently meeting the quality standards of industrial production.

[0003] While some existing technologies attempt to introduce optical detection methods, they lack efficient stereoscopic vision reconstruction mechanisms, making it impossible to accurately acquire the three-dimensional morphology and geometric contour information of the accessory surface. The recognition accuracy for defects such as surface depressions, protrusions, and contour deformation is insufficient, resulting in low reliability of surface integrity judgment results. Furthermore, the fusion decision-making between material identification conclusions and surface integrity judgment conclusions lacks scientific multi-level decision-making rules, and the sorting logic is relatively simple, leading to poor overall sorting process connections, long processing times, and difficulty in meeting the high-efficiency requirements of industrial production. Therefore, how to improve the efficiency of optical sorting of door and window hardware accessories has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides an optical sorting method and system for door and window hardware accessories to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an optical sorting method for door and window hardware accessories, comprising:

[0006] S1. Illuminate the door and window hardware accessories during transmission using a multispectral light source array, and simultaneously acquire reflectance spectrum images to obtain the optical dataset of the door and window hardware accessories;

[0007] S2. Perform band separation on the original broadband image data in the optical dataset, convert the optical signals of each band into electrical signals, and then analyze the characteristic absorption peaks and reflectivity distribution in the electrical signals to obtain the material characteristic information of the door and window hardware accessories.

[0008] S3. Based on the multi-view image data in the optical dataset, the surface three-dimensional shape of the door and window hardware is reconstructed by stereo vision triangulation to obtain the geometric contour information of the door and window hardware.

[0009] S4. Match the material feature information with the standard spectral sample of the door and window hardware, output the material identification conclusion of the door and window hardware, and perform morphological structure analysis on the geometric contour information to obtain the surface integrity judgment conclusion of the door and window hardware.

[0010] S5. Apply multi-level decision rules to integrate the material identification conclusion and the surface integrity judgment conclusion to generate the sorting control instruction for the door and window hardware accessories;

[0011] S6. Encode the sorting control command into a pulse signal to complete the sorting of the door and window hardware accessories.

[0012] In a preferred embodiment, the step of illuminating the door and window hardware during transmission with a multispectral light source array and simultaneously acquiring reflectance spectral images to obtain an optical dataset of the door and window hardware includes:

[0013] Activate the multi-band light source in the multispectral light source array to illuminate the surface of the door and window hardware accessories, forming a composite lighting field for the door and window hardware accessories;

[0014] In the composite lighting field, the position status of the door and window hardware is monitored. When the door and window hardware is detected to have reached the preset acquisition position, the position trigger signal of the door and window hardware is obtained.

[0015] By responding to the position trigger signal, the band switching operation of the multispectral light source array and the acquisition operation of the reflectance spectrum image in the door and window hardware are synchronously controlled to obtain the multi-band image of the door and window hardware.

[0016] Based on the correlation between feature points of adjacent images in the multi-band images, spatial transformation parameters between images in the door and window hardware accessories are established.

[0017] Based on the spatial transformation parameters, the spatial dimension positions in the multi-band images are registered to obtain the optical dataset of the door and window hardware accessories.

[0018] In a preferred embodiment, the step of performing band separation on the original broadband image data in the optical dataset, converting the optical signals of each band into electrical signals, and then analyzing the characteristic absorption peaks and reflectance distribution in the electrical signals to obtain the material characteristic information of the door and window hardware accessories includes:

[0019] Based on the typical spectral characteristics of the target material of the door and window hardware, the target band for separating the original broadband image data in the optical dataset is determined;

[0020] Based on the target band, digital spectrum separation is performed on the original broadband image data to obtain narrow band spectral data of the optical dataset;

[0021] The narrow-band spectral data is mapped into corresponding multi-channel digital signals;

[0022] The signal characteristics between channels in the multi-channel digital signal are deconstructed to identify the peak position of the characteristic absorption peak and the reflectance distribution pattern in the signal characteristics;

[0023] The peak positions of the characteristic absorption peaks and the reflectance distribution patterns are mapped to the material characteristic reference range of the door and window hardware to obtain the material characteristic information of the door and window hardware.

[0024] In a preferred embodiment, the step of reconstructing the surface three-dimensional shape of the door and window hardware fittings based on the multi-view image data in the optical dataset and obtaining the geometric contour information of the door and window hardware fittings includes:

[0025] Extract feature point pairs from the multi-view image data in the optical dataset;

[0026] The three-dimensional spatial coordinates of the feature point pairs are calculated using the triangulation principle to generate the three-dimensional point cloud data of the door and window hardware.

[0027] Noise points in the three-dimensional point cloud data are filtered out, and the filtered three-dimensional point cloud data is reconstructed into a surface to obtain the continuous surface of the door and window hardware.

[0028] The curvature distribution characteristics of the continuous surface are analyzed to obtain the curvature changes of the curvature distribution characteristics;

[0029] Based on the curvature change, the edge contour area of ​​the door and window hardware is identified, and the geometric contour information of the door and window hardware is obtained.

[0030] In a preferred embodiment, the formula for calculating the three-dimensional spatial coordinates is as follows:

[0031] ;

[0032] In the formula, Let x be the horizontal coordinate of the feature point in the world coordinate system. Let be the vertical coordinate of the feature point in the world coordinate system. The depth distance of the feature point relative to the camera's optical axis. The three-dimensional spatial coordinates are... The preset baseline distance, The preset camera focal length, The disparity value of the feature point pair. The horizontal pixel coordinates of the feature point pair in the reference view. The horizontal coordinates of the principal point in the image obtained through camera calibration. The vertical pixel coordinates of the feature point pair in the reference view. The vertical coordinates of the principal point in the image obtained through camera calibration. It is a matrix.

[0033] In a preferred embodiment, the step of filtering out noise points from the 3D point cloud data and reconstructing a surface from the filtered 3D point cloud data to obtain the continuous surface of the door and window hardware includes:

[0034] Remove isolated data points from the 3D point cloud data to obtain clean point cloud data.

[0035] Connect adjacent data points in the clean point cloud data to construct a triangular patch network of the clean point cloud data;

[0036] The triangular mesh is topologically optimized to obtain the initial curved surface mesh of the door and window hardware.

[0037] Adjusting the patch connections of the initial curved mesh yields a continuous surface for the door and window hardware.

[0038] In a preferred embodiment, the step of matching the material characteristic information with a standard spectral sample of the door and window hardware to output a material identification conclusion for the door and window hardware, and performing morphological structure analysis on the geometric contour information to obtain a surface integrity determination conclusion for the door and window hardware, includes:

[0039] The position of the characteristic absorption peak and the reflectance distribution in the material characteristic information are compared with the standard spectral sample of the door and window hardware in a hierarchical manner to obtain the material characteristic comparison result of the door and window hardware.

[0040] Based on the material characteristic comparison results, the material type with the highest matching degree in the standard spectral sample is selected as the material identification conclusion for the door and window hardware accessories;

[0041] The surface curvature distribution and contour size parameters in the geometric contour information are extracted and compared with the preset benchmark morphological features to obtain the morphological difference data of the geometric contour information.

[0042] Based on the morphological difference data, the distribution characteristics of surface depressions, protrusions and contour deformation areas in the door and window hardware are statistically analyzed to obtain the morphological abnormality characteristics of the door and window hardware.

[0043] Based on the abnormal morphological characteristics, the surface integrity level of the door and window hardware is evaluated, and a conclusion on the surface integrity of the door and window hardware is obtained.

[0044] In a preferred embodiment, the application of multi-level decision rules integrates the material identification conclusion and the surface integrity determination conclusion to generate sorting control instructions for the door and window hardware, including:

[0045] The material type identifier in the material identification conclusion and the integrity level in the surface integrity judgment conclusion are reconstructed into the decision feature set of the door and window hardware accessories;

[0046] Key decision parameters are selected from the set of features to be decided, and path trajectory evolution is performed on the key decision parameters to determine the sorting decision path for the door and window hardware accessories;

[0047] Based on the weight allocation rules in the sorting decision path, the sorting priority of the material type identifier and the integrity level is determined;

[0048] The sorting priority is encoded into instructions to obtain the sorting control instructions for the door and window hardware.

[0049] In a preferred embodiment, encoding the sorting control command into a pulse signal to complete the sorting of the door and window hardware accessories includes:

[0050] Using the pulse width in the sorting control command as the first level and the pulse interval parameter as the second level, a pulse modulation spectrum of the sorting control command is constructed.

[0051] The pulse width value in the pulse modulation spectrum is mapped to the high-level duration, and the pulse interval value in the pulse modulation spectrum is set to the low-level duration;

[0052] The high-level duration and the low-level duration are arranged sequentially according to time order to synthesize the timing control waveform of the door and window hardware.

[0053] The high-pressure air valve is switched on and off by controlling the timing control waveform, thereby completing the sorting of the door and window hardware accessories.

[0054] To address the above problems, the present invention also provides an optical sorting system for door and window hardware accessories, the system comprising:

[0055] A multispectral imaging module is used to illuminate the door and window hardware accessories in transit with a multispectral light source array, and simultaneously acquire reflectance spectrum images to obtain the optical dataset of the door and window hardware accessories.

[0056] The material feature analysis module is used to perform band separation on the original broadband image data in the optical dataset, convert the optical signals of each band into electrical signals, and then analyze the characteristic absorption peaks and reflectivity distribution in the electrical signals to obtain the material feature information of the door and window hardware accessories.

[0057] The three-dimensional shape reconstruction module is used to reconstruct the surface three-dimensional shape of the door and window hardware accessories based on the multi-view image data in the optical dataset and through stereo vision triangulation, so as to obtain the geometric contour information of the door and window hardware accessories.

[0058] The quality assessment module is used to match the material feature information with the standard spectral sample of the door and window hardware, output the material identification conclusion of the door and window hardware, and perform morphological structure analysis on the geometric contour information to obtain the surface integrity assessment conclusion of the door and window hardware.

[0059] The intelligent decision-making module is used to apply multi-level decision-making rules to fuse the material identification conclusion and the surface integrity judgment conclusion, and generate sorting control instructions for the door and window hardware accessories;

[0060] The sorting execution module is used to encode the sorting control instructions into pulse signals to complete the sorting of the door and window hardware accessories.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] 1. This invention illuminates door and window hardware accessories during transmission using a multispectral light source array and simultaneously acquires reflectance spectrum images. By combining band separation and electrical signal analysis, it can accurately capture the characteristic absorption peaks and reflectance distribution of the accessories, efficiently obtaining accurate material characteristic information. At the same time, relying on stereoscopic vision triangulation technology, it can effectively reconstruct the three-dimensional morphology of the accessory surface, clearly analyze geometric contour information, and accurately identify surface depressions, protrusions, and contour deformations. This ensures the comprehensiveness and accuracy of the data required for material identification and surface integrity determination, providing a reliable data foundation for subsequent sorting and improving the scientific nature of sorting decisions.

[0063] 2. This invention, by applying multi-level decision-making rules, can scientifically integrate material identification conclusions and surface integrity judgment conclusions, rationally determine sorting priorities, and generate precise sorting control instructions. These instructions are then encoded into pulse signals to control sorting execution, achieving precise triggering and efficient connection of sorting operations. This reduces redundant steps in the sorting process and significantly improves the overall efficiency of door and window hardware component sorting. Simultaneously, it ensures the consistency and stability of sorting results, meeting the demands of industrial production for high-quality, high-efficiency sorting, and providing strong support for quality control in the component production process. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating an optical sorting method for door and window hardware accessories according to an embodiment of the present invention.

[0065] Figure 2 A functional module diagram of an optical sorting system for door and window hardware accessories provided in an embodiment of the present invention;

[0066] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0067] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0068] This application provides an optical sorting method for door and window hardware accessories. The executing entity of this optical sorting method for door and window hardware accessories includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application embodiment: a server, a terminal, etc. In other words, the optical sorting method for door and window hardware accessories can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0069] Reference Figure 1 The diagram shown is a flowchart illustrating an optical sorting method for door and window hardware accessories according to an embodiment of the present invention. In this embodiment, the optical sorting method for door and window hardware accessories includes:

[0070] S1. Illuminate the door and window hardware accessories during transmission using a multispectral light source array, and simultaneously acquire reflectance spectrum images to obtain the optical dataset of the door and window hardware accessories;

[0071] In this embodiment of the invention, the step of illuminating the door and window hardware accessories during transmission with a multispectral light source array and simultaneously acquiring reflectance spectral images to obtain an optical dataset of the door and window hardware accessories includes:

[0072] Activate the multi-band light source in the multispectral light source array to illuminate the surface of the door and window hardware accessories, forming a composite lighting field for the door and window hardware accessories;

[0073] In the composite lighting field, the position status of the door and window hardware is monitored. When the door and window hardware is detected to have reached the preset acquisition position, the position trigger signal of the door and window hardware is obtained.

[0074] By responding to the position trigger signal, the band switching operation of the multispectral light source array and the acquisition operation of the reflectance spectrum image in the door and window hardware are synchronously controlled to obtain the multi-band image of the door and window hardware.

[0075] Based on the correlation between feature points of adjacent images in the multi-band images, spatial transformation parameters between images in the door and window hardware accessories are established.

[0076] Based on the spatial transformation parameters, the spatial dimension positions in the multi-band images are registered to obtain the optical dataset of the door and window hardware accessories.

[0077] The multispectral light source array contains ultraviolet, visible, and near-infrared light sources. The light source array is arranged in a 3×3 matrix, with each band corresponding to an independent light source module. The current and voltage of each module are adjusted by the light source driving circuit to stabilize the power at a preset value. At the same time, all light source modules are controlled to light up simultaneously according to a time sequence that matches the speed of accessory delivery. After the light shines on the surface of the door and window hardware accessories, the light of different bands forms a uniformly superimposed composite lighting field on the surface of the accessories, ensuring that there are no shadow areas on the surface of the accessories and that each position is fully covered by multi-band light.

[0078] A laser displacement sensor is installed above the entrance of the acquisition area in the composite lighting field. It emits a 650nm wavelength laser beam that vertically illuminates the surface of the component. The sensor receives the reflected laser signal and converts it into an electrical signal, monitoring the distance change between the component and the sensor in real time. A vision sensor is installed directly above the acquisition area to continuously capture real-time images of the component. After grayscale processing, the contour features of the component are extracted by setting a grayscale threshold. When the laser displacement sensor detects that the distance between the component and the sensor reaches a preset 5cm threshold, and the vision sensor recognizes that the component's contour has completely entered the acquisition area (i.e., the proportion of the contour edge extending beyond the acquisition area boundary is less than 20%), the microcontroller in the system receives confirmation signals from both sensors and outputs a high-level position trigger signal for 10ms via the I / O port.

[0079] After receiving the position trigger signal, the PLC controller in the control system sends a band switching command to the multispectral light source array controller via the Modbus communication protocol. The working bands of the light source are switched sequentially in the order of ultraviolet 200-400nm, visible 400-760nm, and near-infrared 760-1100nm. The duration of each band is set to 50ms, which is synchronized with the 20fps frame rate of the industrial camera. At the same time, the PLC controller sends a synchronous acquisition command to the PCIe interface image acquisition card, which controls the industrial camera to capture the reflection spectrum image of the part with an exposure time of 30ms during the illumination of each band light source. The image acquisition card converts the analog signal output by the camera into digital image data and stores it in the buffer. After continuous acquisition, a multi-band image containing the spectral information of the part in three bands is obtained.

[0080] For each multi-band image, grayscale conversion is performed first, followed by smoothing using a 5×5 Gaussian filter to remove noise interference. Next, the grayscale gradient of each pixel in the image is calculated in the x and y directions. Edge candidate points are determined based on the gradient magnitude and direction. Finally, by setting high and low thresholds, strong edges with gradient values ​​higher than the high threshold and weak edges connected to strong edges with gradient values ​​higher than the low threshold are retained, thus extracting the contour feature points of the component. Simultaneously, pixels with drastic grayscale value changes are identified in the image. By analyzing the grayscale distribution within a 3×3 region surrounding these pixels, it is determined whether they are corner points, i.e., points with significant horizontal, vertical, and diagonal variations. The gray values ​​along the line direction are significantly different from those of the surrounding pixels, thus extracting the corner features of the parts. Then, descriptive information is constructed for each feature point. A 16×16 area is selected around the feature point, and this area is divided into 4×4 sub-regions. Gradient histograms in 8 directions are calculated in each sub-region. The gradient histograms of all sub-regions are combined to form a 128-dimensional feature description vector. The description vectors of feature points in different images are then compared pairwise to calculate the Euclidean distance. If the distance is less than a set threshold, it is determined to be a matching feature point pair. Based on the displacement changes and geometric distribution of these matching point pairs, the translation amount, rotation angle and scaling ratio between images are determined, and spatial transformation parameters are established.

[0081] The intermediate frame image of the visible band is selected as the reference image. For each pixel in the other band images, the corresponding coordinates in the reference image are calculated according to the established spatial transformation parameters. Bilinear interpolation is used to fill the pixel gaps caused by the coordinate transformation, ensuring that the resolution of all band images is consistent after registration, and that the same physical position of the accessory is at the same pixel coordinate in different band images. This eliminates the spatial position differences caused by shooting angle deviation and accessory position offset. Finally, an optical dataset containing registered multi-band image data, spectral information labels for each band, image acquisition timestamps, and accessory position information is obtained. The data is stored in TIFF format for subsequent processing.

[0082] The beneficial effects are as follows: by coordinating and uniformly illuminating different bands of light sources in a multispectral light source array, combined with precise position monitoring by laser displacement sensors and vision sensors, it ensures that when door and window hardware accessories reach the acquisition position, multi-band switching and image acquisition can be triggered simultaneously, comprehensively capturing the reflection characteristics of different materials on the surface of the accessories for each band of light; then, through grayscale conversion, Gaussian filtering, gradient calculation and other steps, the contour and corner features of the accessories are extracted, feature description vectors are constructed and matched, spatial transformation parameters between images are determined, and pixel-level precise registration of multi-band images is performed to eliminate spatial position deviations, forming an optical dataset containing complete spectral and spatial information, providing high-fidelity and comprehensive data support for subsequent material identification and surface defect detection. At the same time, the entire process timing control from light source activation to dataset generation is optimized, redundant operations are reduced, and the detection accuracy and operating efficiency of the sorting system are significantly improved.

[0083] S2. Perform band separation on the original broadband image data in the optical dataset, convert the optical signals of each band into electrical signals, and then analyze the characteristic absorption peaks and reflectivity distribution in the electrical signals to obtain the material characteristic information of the door and window hardware accessories.

[0084] In this embodiment of the invention, the step of performing band separation on the original broadband image data in the optical dataset, converting the optical signals of each band into electrical signals, and then analyzing the characteristic absorption peaks and reflectance distribution in the electrical signals to obtain the material characteristic information of the door and window hardware accessories includes:

[0085] Based on the typical spectral characteristics of the target material of the door and window hardware, the target band for separating the original broadband image data in the optical dataset is determined;

[0086] Based on the target band, digital spectrum separation is performed on the original broadband image data to obtain narrow band spectral data of the optical dataset;

[0087] The narrow-band spectral data is mapped into corresponding multi-channel digital signals;

[0088] The signal characteristics between channels in the multi-channel digital signal are deconstructed to identify the peak position of the characteristic absorption peak and the reflectance distribution pattern in the signal characteristics;

[0089] The peak positions of the characteristic absorption peaks and the reflectance distribution patterns are mapped to the material characteristic reference range of the door and window hardware to obtain the material characteristic information of the door and window hardware.

[0090] The multispectral light source array contains ultraviolet, visible, and near-infrared light sources. The light source array is arranged in a 3×3 matrix, with each band corresponding to an independent light source module. The current and voltage of each module are adjusted by the light source driving circuit to stabilize the power at a preset value. At the same time, all light source modules are controlled to light up simultaneously according to a time sequence that matches the speed of accessory delivery. After the light shines on the surface of the door and window hardware accessories, the light of different bands forms a uniformly superimposed composite lighting field on the surface of the accessories, ensuring that there are no shadow areas on the surface of the accessories and that each position is fully covered by multi-band light.

[0091] A laser displacement sensor is installed above the entrance of the acquisition area in the composite lighting field. It emits a 650nm wavelength laser beam that vertically illuminates the surface of the component. The sensor receives the reflected laser signal and converts it into an electrical signal, monitoring the distance change between the component and the sensor in real time. A vision sensor is installed directly above the acquisition area to continuously capture real-time images of the component. After grayscale processing, the contour features of the component are extracted by setting a grayscale threshold. When the laser displacement sensor detects that the distance between the component and the sensor reaches a preset 5cm threshold, and the vision sensor recognizes that the component's contour has completely entered the acquisition area (i.e., the proportion of the contour edge extending beyond the acquisition area boundary is less than 20%), the microcontroller in the system receives confirmation signals from both sensors and outputs a high-level position trigger signal for 10ms via the I / O port.

[0092] After receiving the position trigger signal, the PLC controller in the control system sends a band switching command to the multispectral light source array controller via the Modbus communication protocol. The working bands of the light source are switched sequentially in the order of ultraviolet 200-400nm, visible 400-760nm, and near-infrared 760-1100nm. The duration of each band is set to 50ms, which is synchronized with the 20fps frame rate of the industrial camera. At the same time, the PLC controller sends a synchronous acquisition command to the PCIe interface image acquisition card, which controls the industrial camera to capture the reflection spectrum image of the part with an exposure time of 30ms during the illumination of each band light source. The image acquisition card converts the analog signal output by the camera into digital image data and stores it in the buffer. After continuous acquisition, a multi-band image containing the spectral information of the part in three bands is obtained.

[0093] For each multi-band image, grayscale conversion is performed first, followed by smoothing using a 5×5 Gaussian filter to remove noise interference. Next, the grayscale gradient of each pixel in the image is calculated in the x and y directions. Edge candidate points are determined based on the gradient magnitude and direction. Finally, by setting high and low thresholds, strong edges with gradient values ​​higher than the high threshold and weak edges connected to strong edges with gradient values ​​higher than the low threshold are retained, thus extracting the contour feature points of the component. Simultaneously, pixels with drastic grayscale value changes are identified in the image. By analyzing the grayscale distribution within a 3×3 region surrounding these pixels, it is determined whether they are corner points, i.e., points with significant horizontal, vertical, and diagonal variations. The gray values ​​along the line direction are significantly different from those of the surrounding pixels, thus extracting the corner features of the parts. Then, descriptive information is constructed for each feature point. A 16×16 area is selected around the feature point, and this area is divided into 4×4 sub-regions. Gradient histograms in 8 directions are calculated in each sub-region. The gradient histograms of all sub-regions are combined to form a 128-dimensional feature description vector. The description vectors of feature points in different images are then compared pairwise to calculate the Euclidean distance. If the distance is less than a set threshold, it is determined to be a matching feature point pair. Based on the displacement changes and geometric distribution of these matching point pairs, the translation amount, rotation angle and scaling ratio between images are determined, and spatial transformation parameters are established.

[0094] The intermediate frame image of the visible band is selected as the reference image. For each pixel in the other band images, the corresponding coordinates in the reference image are calculated according to the established spatial transformation parameters. Bilinear interpolation is used to fill the pixel gaps caused by the coordinate transformation, ensuring that the resolution of all band images is consistent after registration, and that the same physical position of the accessory is at the same pixel coordinate in different band images. This eliminates the spatial position differences caused by shooting angle deviation and accessory position offset. Finally, an optical dataset containing registered multi-band image data, spectral information labels for each band, image acquisition timestamps, and accessory position information is obtained. The data is stored in TIFF format for subsequent processing.

[0095] The beneficial effects are as follows: by coordinating and uniformly illuminating different bands of light sources in a multispectral light source array, combined with precise position monitoring by laser displacement sensors and vision sensors, it ensures that when door and window hardware accessories reach the acquisition position, multi-band switching and image acquisition can be triggered simultaneously, comprehensively capturing the reflection characteristics of different materials on the surface of the accessories for each band of light; then, through grayscale conversion, Gaussian filtering, gradient calculation and other steps, the contour and corner features of the accessories are extracted, feature description vectors are constructed and matched, spatial transformation parameters between images are determined, and pixel-level precise registration of multi-band images is performed to eliminate spatial position deviations, forming an optical dataset containing complete spectral and spatial information, providing high-fidelity and comprehensive data support for subsequent material identification and surface defect detection. At the same time, the entire process timing control from light source activation to dataset generation is optimized, redundant operations are reduced, and the detection accuracy and operating efficiency of the sorting system are significantly improved.

[0096] S3. Based on the multi-view image data in the optical dataset, the surface three-dimensional shape of the door and window hardware is reconstructed by stereo vision triangulation to obtain the geometric contour information of the door and window hardware.

[0097] In this embodiment of the invention, the step of reconstructing the surface three-dimensional shape of the door and window hardware accessories based on the multi-view image data in the optical dataset and obtaining the geometric contour information of the door and window hardware accessories includes:

[0098] Extract feature point pairs from the multi-view image data in the optical dataset;

[0099] The three-dimensional spatial coordinates of the feature point pairs are calculated using the triangulation principle to generate the three-dimensional point cloud data of the door and window hardware.

[0100] Noise points in the three-dimensional point cloud data are filtered out, and the filtered three-dimensional point cloud data is reconstructed into a surface to obtain the continuous surface of the door and window hardware.

[0101] The curvature distribution characteristics of the continuous surface are analyzed to obtain the curvature changes of the curvature distribution characteristics;

[0102] Based on the curvature change, the edge contour area of ​​the door and window hardware is identified, and the geometric contour information of the door and window hardware is obtained.

[0103] The formula for calculating the three-dimensional spatial coordinates is as follows:

[0104] ;

[0105] In the formula, Let x be the horizontal coordinate of the feature point in the world coordinate system. Let be the vertical coordinate of the feature point in the world coordinate system. The depth distance of the feature point relative to the camera's optical axis. The three-dimensional spatial coordinates are... The preset baseline distance, The preset camera focal length, The disparity value of the feature point pair. The horizontal pixel coordinates of the feature point pair in the reference view. The horizontal coordinates of the principal point in the image obtained through camera calibration. The vertical pixel coordinates of the feature point pair in the reference view. The vertical coordinates of the principal point in the image obtained through camera calibration. It is a matrix.

[0106] The process of filtering out noise points from the 3D point cloud data and reconstructing the filtered 3D point cloud data to obtain the continuous surface of the door and window hardware includes:

[0107] Remove isolated data points from the 3D point cloud data to obtain clean point cloud data.

[0108] Connect adjacent data points in the clean point cloud data to construct a triangular patch network of the clean point cloud data;

[0109] The triangular mesh is topologically optimized to obtain the initial curved surface mesh of the door and window hardware.

[0110] Adjusting the patch connections of the initial curved mesh yields a continuous surface for the door and window hardware.

[0111] Two cameras for acquiring multi-view images are calibrated in advance. A calibration board with known size and feature point distribution is placed at different positions and angles. Multiple images of the calibration board are taken by the two cameras respectively. The correspondence between the pixel coordinates of the feature points of the calibration board in the images and the actual three-dimensional coordinates is analyzed. The focal length of the camera and the horizontal and vertical coordinates of the principal point of the image are calculated. The baseline distance is the straight-line distance between the optical centers of the two cameras when they are installed. This distance is measured using a high-precision ranging tool and recorded as a preset fixed value. These parameters provide the basic support for subsequent three-dimensional spatial coordinate calculations.

[0112] For each multi-view image in the optical dataset, grayscale processing is first performed, converting the three-channel color image into a single-channel grayscale image. Then, a 5×5 Gaussian filter is used to smooth the image and remove random noise interference. Next, a corner detection method is employed to calculate the grayscale gradient of each pixel in the image in both the horizontal and vertical directions. If the gradient magnitude of a point is greater than the gradient thresholds in both the horizontal and vertical directions, it is identified as a feature point. The pixel coordinates of all feature points are extracted. Then, a description vector is constructed for each feature point. A 16×16 region is selected around the feature point and divided into 4×4 sub-regions. Within each sub-region, a calculation is performed... The gradient histograms in eight directions are combined with the histograms of all sub-regions to form a 128-dimensional description vector. Finally, the description vectors of feature points in images from different viewpoints are compared, and the Euclidean distance between the vectors is calculated. Feature points with a distance less than a set threshold are considered matching pairs, thus obtaining feature point pairs in multi-view image data. The disparity value is obtained by reading the horizontal pixel coordinates of the feature point pair in the two images and subtracting the horizontal pixel coordinates in the other image from the horizontal pixel coordinates in the reference view. The horizontal and vertical pixel coordinates in the reference view are obtained by reading the corresponding position values ​​of the feature points in the image pixel coordinate system through the image acquisition system.

[0113] The core function of the above formula is to convert the pixel coordinates of feature points in a two-dimensional image into three-dimensional spatial coordinates in the world coordinate system. By combining the inherent parameters of the camera and the relative position parameters during the shooting process, a correspondence between two-dimensional pixel information and three-dimensional spatial position is established. The horizontal position, vertical position, and depth distance of the feature points relative to the camera optical axis in the world coordinate system are accurately calculated. For each pair of feature points, the pixel coordinates in the images of the two cameras are read. According to the camera imaging model, the pixel coordinates are converted into light direction vectors in the camera coordinate system. A triangle is constructed with the optical centers of the two cameras as vertices and the two light direction vectors as two sides. By solving for the intersection of the two light rays, the three-dimensional spatial coordinates of the feature point pair are obtained. By traversing all feature point pairs, the calculated set of three-dimensional coordinates is the three-dimensional point cloud data of the door and window hardware accessories.

[0114] Traverse each data point in the 3D point cloud data. With that point as the center, set a spherical neighborhood with a radius of 5mm. Count the number of other data points contained in the neighborhood. If the number of data points in the neighborhood is less than 3, the point is determined to be an isolated data point and removed from the 3D point cloud data. Repeat this process until all isolated data points are removed to obtain clean point cloud data.

[0115] For clean point cloud data, the Delaunay triangulation method is used to construct a triangular patch network. First, the convex hull in the point cloud is determined, and the points on the convex hull are used to form an initial triangle. Then, the points inside the convex hull are added to the triangulation structure one by one. For each newly added point, the triangle it belongs to is found and decomposed into three new triangles. At the same time, it is ensured that the circumcircle of each new triangle does not contain other data points to avoid the formation of long and narrow triangles. All internal points are added in sequence to finally form a triangular patch network that connects adjacent data points.

[0116] To optimize the topology of the triangular mesh network, first check and remove duplicate data points, merge points with the same coordinates and their associated triangles, then calculate the normal vector of each triangle. If the angle between the normal vectors of two adjacent triangles is less than 5 degrees, they are considered coplanar triangles and are merged into one large triangle. Then, check the non-manifold edges in the triangular mesh network, delete one of the associated triangles, and reconnect the surrounding points to ensure that each edge is shared by only two triangles, thus obtaining the initial surface mesh of the door and window hardware fittings.

[0117] Examine the triangles in the initial surface mesh. If the aspect ratio of a triangle is greater than 5, it is identified as a narrow triangle. Add a new vertex on its longest side and divide the narrow triangle into two nearly equilateral triangles. For cracks in the surface mesh, add new vertices and triangles at the cracks to fill the gaps. Finally, adjust the three-dimensional coordinates of the vertex according to the normal vector of the triangle around each vertex to make the vertex closer to the actual surface of the fitting, thus obtaining a continuous surface for the door and window hardware fitting.

[0118] For each vertex on a continuous surface, select three adjacent vertices around it to form a local triangle, calculate the normal vector of the plane containing the triangle, then select another three adjacent vertices of the same vertex to form another local triangle, calculate its normal vector, and calculate the curvature value of the vertex by the angle between the two normal vectors. The larger the curvature value, the more severe the curvature of the surface. Traverse all vertices on the continuous surface and calculate the curvature value of each vertex to obtain the curvature distribution characteristics. Then analyze the change of curvature from low to high to determine the range and trend of curvature change.

[0119] Based on the curvature distribution characteristics, a curvature threshold is set. If the curvature values ​​of all vertices in a certain region on a continuous surface are greater than the threshold, the region is determined to be an edge contour region. All vertices within the edge contour region are extracted, and these vertices are connected sequentially according to their position order in three-dimensional space to form a closed contour line. The contour line includes the outer contour line of the door and window hardware and the contour line of the internal holes. These contour lines together constitute the geometric contour information of the door and window hardware.

[0120] The beneficial effects are that by extracting feature point pairs from multi-view images and performing triangulation, the three-dimensional spatial coordinates of the parts can be accurately obtained, generating high-quality three-dimensional point cloud data. After removing isolated points, triangulation, and topology optimization, a complete and continuous surface model is constructed. Combined with curvature analysis, edge contour regions are accurately identified, and comprehensive and accurate geometric contour information is obtained, providing reliable data support for subsequent determination of the surface integrity of the parts and effectively improving the accuracy and reliability of the determination results.

[0121] S4. Match the material feature information with the standard spectral sample of the door and window hardware, output the material identification conclusion of the door and window hardware, and perform morphological structure analysis on the geometric contour information to obtain the surface integrity judgment conclusion of the door and window hardware.

[0122] In this embodiment of the invention, the step of matching the material feature information with the standard spectral sample of the door and window hardware to output the material identification conclusion of the door and window hardware, and performing morphological structure analysis on the geometric contour information to obtain the surface integrity judgment conclusion of the door and window hardware, includes:

[0123] The position of the characteristic absorption peak and the reflectance distribution in the material characteristic information are compared with the standard spectral sample of the door and window hardware in a hierarchical manner to obtain the material characteristic comparison result of the door and window hardware.

[0124] Based on the material characteristic comparison results, the material type with the highest matching degree in the standard spectral sample is selected as the material identification conclusion for the door and window hardware accessories;

[0125] The surface curvature distribution and contour size parameters in the geometric contour information are extracted and compared with the preset benchmark morphological features to obtain the morphological difference data of the geometric contour information.

[0126] Based on the morphological difference data, the distribution characteristics of surface depressions, protrusions and contour deformation areas in the door and window hardware are statistically analyzed to obtain the morphological abnormality characteristics of the door and window hardware.

[0127] Based on the abnormal morphological characteristics, the surface integrity level of the door and window hardware is evaluated, and a conclusion on the surface integrity of the door and window hardware is obtained.

[0128] The system retrieves pre-stored standard spectral samples of various standard materials for door and window hardware. Each sample contains clearly defined characteristic absorption peak positions and corresponding reflectance distribution data. The characteristic absorption peak positions in the material feature information of the door and window hardware are compared one by one with the corresponding wavelength positions of the characteristic absorption peaks in the standard spectral samples. Then, from low wavelength to high wavelength, the numerical trends and specific values ​​of reflectance are compared segment by segment. The comparison results for each segment are recorded in detail. Finally, the comparison results of all bands are summarized to obtain the material feature comparison results of the door and window hardware.

[0129] The material characteristic comparison results are quantitatively evaluated by counting the number of characteristic absorption peaks that match the material characteristics of each standard spectral sample and the door and window hardware, as well as the degree of match in reflectance distribution across different wavelengths. A higher number of matches and smaller reflectance differences indicate a higher degree of matching. All standard spectral samples are then ranked by their matching degree, and the material type corresponding to the standard spectral sample with the highest matching degree is directly determined as the material identification conclusion for the door and window hardware.

[0130] From the geometric contour information of door and window hardware, the curvature value of each surface vertex and the overall curvature change trend are extracted to form complete surface curvature distribution data. Simultaneously, key contour dimension parameters such as the overall length, width, thickness, and internal hole diameter and depth of the hardware are measured and extracted. A preset benchmark morphological feature is retrieved, which includes the standard distribution range of surface curvature and standard values ​​of various contour dimensions for standard qualified door and window hardware. The extracted surface curvature distribution data is compared point-by-point with the benchmark curvature distribution range, and the contour dimension parameters are compared one by one with the benchmark dimension values. The magnitude and trend of the numerical difference for each comparison item are recorded. After comprehensive summarization, the morphological difference data of the geometric contour information is obtained.

[0131] By deeply analyzing the morphological difference data, when the surface curvature value is higher than the upper limit of the reference curvature distribution and shows a localized concentrated distribution, the area is determined to be a raised area; when the curvature value is lower than the lower limit of the reference curvature distribution and shows a localized concentrated distribution, it is determined to be a recessed area; when the deviation of the contour dimension parameters from the reference value exceeds the preset allowable range, the corresponding part is determined to be a contour deformation area. The entire geometric contour information is traversed to accurately mark the specific locations of all raised, recessed, and deformed contour areas. Key indicators such as the area, length, or deformation of each defect area are measured. The number, distribution range, and positional relationships of various defect areas are statistically analyzed to obtain the morphological abnormalities of door and window hardware.

[0132] A pre-defined standard for surface integrity levels is established: Level 1 is the absence of any abnormal morphological features; Level 2 is the presence of minor abnormal features, with the total area of ​​all abnormal areas not exceeding 5% of the total surface area of ​​the accessory; Level 3 is the total area of ​​abnormal areas between 5% and 15% with no severe contour deformation; and Level 4 is the total area of ​​abnormal areas exceeding 15% or with severe contour deformation affecting the normal use of the accessory. Based on the type, number, size, and distribution of defective areas within the abnormal morphological features, each item is evaluated against the pre-defined standard to determine the corresponding surface integrity level of the door and window hardware, thus obtaining a conclusion on the surface integrity of the door and window hardware.

[0133] The beneficial effects are as follows: by comparing material characteristics with standard spectral samples at each level, the material type with the highest matching degree can be accurately screened, ensuring the accuracy of the material identification conclusion. By extracting key parameters of geometric contour and analyzing the depth difference with the benchmark morphological features, the system identifies defect areas such as surface depressions, protrusions, and contour deformation, and statistically analyzes abnormal features to scientifically evaluate the surface integrity level, making the surface integrity judgment conclusion more reliable. This provides a comprehensive and accurate quality basis for subsequent sorting decisions, ensuring the rationality and effectiveness of the sorting results.

[0134] S5. Apply multi-level decision rules to integrate the material identification conclusion and the surface integrity judgment conclusion to generate the sorting control instruction for the door and window hardware accessories;

[0135] In this embodiment of the invention, the application of multi-level decision rules integrates the material identification conclusion and the surface integrity judgment conclusion to generate sorting control instructions for the door and window hardware accessories, including:

[0136] The material type identifier in the material identification conclusion and the integrity level in the surface integrity judgment conclusion are reconstructed into the decision feature set of the door and window hardware accessories;

[0137] Key decision parameters are selected from the set of features to be decided, and path trajectory evolution is performed on the key decision parameters to determine the sorting decision path for the door and window hardware accessories;

[0138] Based on the weight allocation rules in the sorting decision path, the sorting priority of the material type identifier and the integrity level is determined;

[0139] The sorting priority is encoded into instructions to obtain the sorting control instructions for the door and window hardware.

[0140] The material type identifier in the material identification conclusion is retrieved. This identifier is a unique character code for the corresponding material type. At the same time, the integrity level in the surface integrity judgment conclusion is extracted. This level is a pre-defined explicit level code. These two pieces of information are integrated according to the fixed structure of "material type identifier - integrity level" to form a structured data set containing the core quality characteristics of door and window hardware. This set is the decision feature set of door and window hardware.

[0141] An importance analysis was conducted on two pieces of information with concentrated decision-making characteristics. Material type identification directly determines whether accessories meet basic material requirements, while integrity level affects the performance and lifespan of accessories. Both are core bases for sorting decisions; therefore, these two pieces of information were selected as key decision parameters. For these key decision parameters, all possible parameter combinations were analyzed. Each combination corresponds to a clear sorting direction. Starting from the combination logic of material qualification and integrity level, the sorting process that accessories should enter under each combination was gradually deduced, ultimately determining a sorting decision path for door and window hardware accessories covering all possible situations.

[0142] The pre-defined weighting rules for sorting decision paths prioritize material type (whether it falls within the pre-defined acceptable material range) over completeness level. This means that if a material type does not fall within the acceptable range, it is sorted according to the path for unacceptable materials. If the material type falls within the acceptable range, the completeness level determines the subsequent sorting direction. Based on these weighting rules, the priority of material type and completeness level in the sorting decision is determined: the acceptance of the material is prioritized over the completeness level. The completeness level, assuming the material is acceptable, determines the specific sorting category, thus clarifying the sorting priority of both.

[0143] Based on the determined sorting priorities, fixed instruction coding rules are established. Different sorting priorities correspond to unique binary number sequences. The priority code for unqualified materials is a fixed four-bit binary number, while qualified materials with different integrity levels correspond to different four-bit binary numbers. The final determined sorting priorities are matched with their corresponding binary codes, converting the sorting priority information into machine-recognizable binary number instructions. These instructions contain key control information such as sorting direction and target area, and are the sorting control instructions for door and window hardware accessories.

[0144] The beneficial effects are that by reconstructing the core quality characteristics into a set of features to be decided, the key parameters for sorting can be accurately identified. Combined with logical path deduction and clear weight allocation rules, the sorting priority can be clearly determined and transformed into standardized sorting control instructions, ensuring the scientific nature and consistency of sorting decisions, effectively avoiding sorting chaos, and improving the accuracy and efficiency of sorting door and window hardware accessories.

[0145] S6. Encode the sorting control command into a pulse signal to complete the sorting of the door and window hardware accessories.

[0146] In this embodiment of the invention, encoding the sorting control command into a pulse signal to complete the sorting of the door and window hardware accessories includes:

[0147] Using the pulse width in the sorting control command as the first level and the pulse interval parameter as the second level, a pulse modulation spectrum of the sorting control command is constructed.

[0148] The pulse width value in the pulse modulation spectrum is mapped to the high-level duration, and the pulse interval value in the pulse modulation spectrum is set to the low-level duration;

[0149] The high-level duration and the low-level duration are arranged sequentially according to time order to synthesize the timing control waveform of the door and window hardware.

[0150] The high-pressure air valve is switched on and off by controlling the timing control waveform, thereby completing the sorting of the door and window hardware accessories.

[0151] Extract the pulse width and pulse interval parameters from the sorting control commands and determine their specific values. Using pulse width as the first horizontal level, arrange different pulse width values ​​in ascending order to form a horizontal dimension; using pulse interval as the second vertical level, arrange different pulse interval values ​​in ascending order to form a vertical dimension. Mark the corresponding control combinations at the intersection of the horizontal and vertical dimensions to form a two-dimensional data table containing all pulse width and pulse interval combinations. This table is the pulse modulation spectrum of the sorting control commands.

[0152] Consulting the pulse modulation spectrum, the pulse width value corresponding to each intersection point is directly mapped to the duration of the high-level signal in the circuit. The value and duration are completely consistent, requiring no additional conversion. Simultaneously, the corresponding pulse interval value in the pulse modulation spectrum is directly set to the duration of the low-level signal in the circuit, ensuring a strict match between the pulse interval value and the low-level duration value, forming a clear correspondence between high and low-level durations.

[0153] According to the execution sequence set by the sorting control command, the electrical signal corresponding to the high-level duration is output first. After the high-level duration ends, the signal is immediately switched to the electrical signal corresponding to the low-level duration. If the low-level duration needs to be repeated, the signal is switched back to high level. This cycle is repeated once or multiple times until the signal output required for a single sorting is completed. These high and low level signals, arranged in chronological order, are connected in series to form a continuous timing control waveform for door and window hardware accessories that meets the sorting requirements.

[0154] The timing control waveform is transmitted to the drive circuit of the high-pressure air valve. When the drive circuit receives a high-level signal, the circuit is turned on to provide operating voltage to the high-pressure air valve. The valve core opens rapidly under air pressure, applying a directional thrust to the door and window hardware. When a low-level signal is received, the circuit is turned off and power is stopped. The valve core closes under the action of the return spring, and the thrust disappears. According to the sorting requirements, the single or multiple opening and closing actions of the air valve push the accessories to the corresponding sorting area, ultimately completing the sorting of the door and window hardware.

[0155] The beneficial effects are that by constructing a pulse modulation spectrum in layers, the duration of high and low levels can be accurately mapped, and a timing control waveform that meets the sorting requirements can be synthesized. This enables precise timing control of the high-pressure air valve's opening and closing actions, ensuring that door and window hardware can be accurately pushed to the target sorting area, improving the stability and accuracy of the sorting action, while optimizing the sorting execution process and improving the overall sorting efficiency.

[0156] like Figure 2 The diagram shown is a functional block diagram of an optical sorting system for door and window hardware accessories provided in an embodiment of the present invention.

[0157] The optical sorting system 100 for door and window hardware accessories described in this invention can be installed in an electronic device. Depending on the functions implemented, the optical sorting system 100 may include a multispectral imaging module 101, a material feature analysis module 102, a three-dimensional shape reconstruction module 103, a quality judgment module 104, an intelligent decision-making module 105, and a sorting execution module 106. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0158] In this embodiment, the functions of each module / unit are as follows:

[0159] The multispectral imaging module 101 is used to illuminate the door and window hardware accessories in transit with a multispectral light source array, and simultaneously acquire reflectance spectrum images to obtain the optical dataset of the door and window hardware accessories.

[0160] The material feature analysis module 102 is used to perform band separation on the original broadband image data in the optical dataset, convert the optical signals of each band into electrical signals, and then analyze the characteristic absorption peaks and reflectivity distribution in the electrical signals to obtain the material feature information of the door and window hardware accessories.

[0161] The three-dimensional shape reconstruction module 103 is used to reconstruct the surface three-dimensional shape of the door and window hardware accessories based on the multi-view image data in the optical dataset and through stereo vision triangulation, so as to obtain the geometric contour information of the door and window hardware accessories.

[0162] The quality judgment module 104 is used to match the material feature information with the standard spectral sample of the door and window hardware, output the material identification conclusion of the door and window hardware, and perform morphological structure analysis on the geometric contour information to obtain the surface integrity judgment conclusion of the door and window hardware.

[0163] The intelligent decision-making module 105 is used to apply multi-level decision rules to fuse the material identification conclusion and the surface integrity judgment conclusion to generate the sorting control instruction for the door and window hardware accessories.

[0164] The sorting execution module 106 is used to encode the sorting control command into a pulse signal to complete the sorting of the door and window hardware accessories.

[0165] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0166] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0167] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0168] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0169] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method of optically sorting door and window hardware, characterized in that, The method comprises: S1, using a multi-spectral light source array to irradiate a door and window hardware fitting in conveying, synchronously collecting a reflected spectrum image, and obtaining optical data set of the door and window hardware fitting; S2, carrying out wave band separation on original wide spectrum image data in the optical data set, converting each wave band light signal into an electric signal, analyzing characteristic absorption peak and reflectivity distribution in the electric signal, and obtaining material characteristic information of the door and window hardware fitting; S3, based on multi-view image data in the optical data set, reconstructing surface three-dimensional topography of the door and window hardware fitting through stereo vision triangulation, and obtaining geometric contour information of the door and window hardware fitting, comprising: extracting feature point pairs of the multi-view image data in the optical data set; calculating three-dimensional space coordinates of the feature point pairs through triangulation principle, and generating three-dimensional point cloud data of the door and window hardware fitting, wherein the calculation formula of the three-dimensional space coordinates is as follows: ； In the formula, is a horizontal coordinate of the feature point in a world coordinate system, is a vertical coordinate of the feature point in the world coordinate system, is a depth distance of the feature point relative to a camera optical axis, is the three-dimensional space coordinate, is a preset baseline distance, is a preset camera focal length, is a parallax value of the feature point pair, is a horizontal pixel coordinate of the feature point pair in a reference view, is a horizontal coordinate of an image principal point obtained through camera calibration, is a vertical pixel coordinate of the feature point pair in the reference view, is a vertical coordinate of the image principal point obtained through camera calibration, is a matrix; removing isolated data points in the three-dimensional point cloud data, and obtaining clean point cloud data of the three-dimensional point cloud data; connecting adjacent data points in the clean point cloud data to construct a triangular facet network of the clean point cloud data; topologically optimizing the triangular facet network to obtain an initial curved surface mesh of the door and window hardware fitting; adjusting facet connection relationship of the initial curved surface mesh to obtain a continuous surface of the door and window hardware fitting; analyzing curvature distribution characteristics of the continuous surface to obtain curvature change of the curvature distribution characteristics; according to the curvature change, identifying an edge contour region of the door and window hardware fitting to obtain the geometric contour information of the door and window hardware fitting; S4, matching the material characteristic information with a standard spectrum sample plate of the door and window hardware fitting, outputting material identification conclusion of the door and window hardware fitting, and analyzing topography structure of the geometric contour information to obtain surface integrity determination conclusion of the door and window hardware fitting; S5, applying multi-level decision rule to fuse the material identification conclusion and the surface integrity determination conclusion to generate sorting control instruction of the door and window hardware fitting, comprising: reconstructing material type identification in the material identification conclusion and integrity level in the surface integrity determination conclusion into a to-be-decided feature set of the door and window hardware fitting; screening out key decision parameters in the to-be-decided feature set, and performing path trajectory evolution on the key decision parameters to determine a sorting decision path of the door and window hardware fitting; determining sorting priority of the material type identification and the integrity level according to weight distribution rule in the sorting decision path; performing instruction coding on the sorting priority to obtain the sorting control instruction of the door and window hardware fitting; S6, coding the sorting control instruction into a pulse signal to complete sorting of the door and window hardware fitting.

2. A method of optically sorting door and window hardware according to claim 1, wherein, The method comprises: activating multi-wave band light sources in the multi-spectral light source array to irradiate the surface of the door and window hardware fitting to form a composite illumination field of the door and window hardware fitting; In the composite illumination field, the position state of the door and window hardware is monitored, and when it is identified that the door and window hardware reaches a preset collection position, a position trigger signal of the door and window hardware is obtained; By responding to the position trigger signal, the wavelength switching operation of the multi-spectral light source array and the collection operation of the reflected spectral image in the door and window hardware are synchronously controlled, and a multi-wavelength image of the door and window hardware is obtained; According to the correlation between the adjacent image feature points in the multi-wavelength image, the spatial transformation parameters between the images in the door and window hardware are established; Based on the spatial transformation parameters, the spatial dimension positions in the multi-wavelength image are registered, and an optical data set of the door and window hardware is obtained.

3. A method of optically sorting door and window hardware according to claim 1, wherein, The original wide-spectrum image data in the optical data set is separated by wavelength, and each wavelength light signal is converted into an electrical signal, and then the characteristic absorption peak and reflectivity distribution in the electrical signal are analyzed to obtain the material characteristic information of the door and window hardware, including: According to the typical spectral characteristics of the target material of the door and window hardware, the target wavelength for separating the original wide-spectrum image data in the optical data set is determined; Based on the target wavelength, the original wide-spectrum image data is digitally spectrally separated to obtain narrow-wavelength spectral data of the optical data set; The narrow-wavelength spectral data is mapped into the corresponding multi-channel digital signal; The signal characteristics between the channels in the multi-channel digital signal are deconstructed to identify the peak position of the characteristic absorption peak and the reflectivity distribution rule in the signal characteristics; The peak position of the characteristic absorption peak and the reflectivity distribution rule are mapped to the material characteristic reference range of the door and window hardware to obtain the material characteristic information of the door and window hardware.

4. A method of optically sorting fenestration hardware according to claim 1, wherein, The material characteristic information is matched with the standard spectral sample of the door and window hardware, and the material identification conclusion of the door and window hardware is output, and the topographic structure analysis of the geometric contour information is performed to obtain the surface integrity determination conclusion of the door and window hardware, including: The characteristic absorption peak position and reflectivity distribution in the material characteristic information are compared with the standard spectral sample of the door and window hardware layer by layer to obtain the material characteristic comparison result of the door and window hardware; According to the material characteristic comparison result, the material type with the highest matching degree in the standard spectral sample is selected as the material identification conclusion of the door and window hardware; The surface curvature distribution and contour size parameters in the geometric contour information are extracted and difference analyzed with the preset reference topographic features to obtain topographic difference data of the geometric contour information; According to the topographic difference data, the distribution characteristics of the surface concave, convex and contour deformation regions in the door and window hardware are counted to obtain the topographic abnormal features of the door and window hardware; According to the topographic abnormal features, the surface integrity level of the door and window hardware is evaluated to obtain the surface integrity determination conclusion of the door and window hardware.

5. A method of optically sorting fenestration hardware according to claim 1, wherein, The sorting control instruction is encoded into a pulse signal to complete the sorting of the door and window hardware, including: The pulse width in the sorting control instruction is a first level, and the pulse interval parameter is a second level, to construct a pulse modulation map of the sorting control instruction; The pulse width value in the pulse modulation map is mapped as a high level duration, and the pulse interval value in the pulse modulation map is set as a low level duration; The high level duration and the low level duration are sequentially arranged in time sequence, to synthesize a timing control waveform of the door and window hardware fitting; The timing control waveform is used to control the opening and closing of the high pressure air valve, to complete the sorting of the door and window hardware fitting.

6. An optical sorting system of a door and window hardware fitting, used to implement the optical sorting method of the door and window hardware fitting in claim 1, the system comprising: a multi-spectral imaging module, used to irradiate the door and window hardware fitting in conveying with a multi-spectral light source array, to synchronously collect a reflected spectral image, to obtain an optical data set of the door and window hardware fitting; a material feature analysis module, used to perform waveband separation on original wide spectral image data in the optical data set, to convert each waveband light signal into an electric signal, to analyze a characteristic absorption peak and reflectivity distribution in the electric signal, to obtain material feature information of the door and window hardware fitting; a three-dimensional topography reconstruction module, used to reconstruct a surface three-dimensional topography of the door and window hardware fitting through stereo vision triangulation based on multi-view image data in the optical data set, to obtain geometric contour information of the door and window hardware fitting; a quality determination module, used to match the material feature information with a standard spectral template of the door and window hardware fitting, to output a material identification conclusion of the door and window hardware fitting, and to perform topography structure analysis on the geometric contour information, to obtain a surface integrity determination conclusion of the door and window hardware fitting; an intelligent decision module, used to apply multi-level decision rules to fuse the material identification conclusion and the surface integrity determination conclusion, to generate a sorting control instruction of the door and window hardware fitting; a sorting execution module, used to encode the sorting control instruction into a pulse signal, to complete the sorting of the door and window hardware fitting.

Images