A visual processing-based vehicle lamp gluing and pressing method and system

Abstract

The application discloses a vehicle lamp gluing and pressing method and system based on visual processing, and belongs to the technical field of image detection. The method comprises the following steps: a processing center receives visual image data, the visual image data comprising three-dimensional point cloud data and two-dimensional RGB data; the three-dimensional point cloud data and the two-dimensional RGB data are fused and processed, and a three-dimensional model library is established, and the data after the fusion processing is stored in the three-dimensional model library; a gluing path is generated, and glue amount distribution is calculated; an execution signal is generated and sent to an execution device, closed-loop verification is performed, and gluing track data based on coordinates is generated; based on the positioning holes of the vehicle lamp and the vehicle shell, the vehicle lamp and the vehicle shell are pressed by adopting a segmented linear pressing mode. In the implementation process of the technical scheme of the application, the visual image data is acquired, the gluing path is generated based on the visual image data and real-time detection data, the manual input and debugging links in the traditional method are avoided, and the gluing efficiency is improved.

Description

TECHNICAL FIELD

[0001] The application relates to the image detection technical field, in particular to a vehicle lamp gluing and pressing method and system based on visual processing. BACKGROUND

[0002] The vehicle lamp is an important component of a vehicle, a motorcycle and the like, and the quality and performance of the vehicle lamp directly affect driving safety.

[0003] In the production and manufacturing process of the vehicle lamp, the vehicle shell and the lampshade are usually processed separately, and after the electronic components in the vehicle shell are installed, the two are tightly combined through a gluing and pressing process, so as to form a complete vehicle lamp assembly. With the development of automatic equipment, the traditional manual gluing and pressing method cannot meet the production requirements of high precision and high efficiency. Therefore, in recent years, gluing and pressing equipment has gradually been popularized, and the off-line programming mode or the teaching programming mode is used to realize the automation of vehicle lamp gluing and pressing, so as to improve the production efficiency and precision.

[0004] However, whether it is the off-line programming mode or the teaching programming mode, the vehicle lamp contour data and the gluing path data need to be pre-recorded. When different models of vehicle lamps need to be glued and pressed, reprogramming is required, which not only consumes a lot of time and manpower, but also increases the error probability. For example, the off-line programming mode needs to be debugged many times according to the actual vehicle lamp model, and the teaching programming mode relies on manual guidance. These processes are difficult to avoid human errors and lack flexibility, and problems such as uneven glue amount and poor pressing are likely to occur, which affects the overall quality and service life of the vehicle lamp.

[0005] Therefore, it is necessary to provide a vehicle lamp gluing and pressing method and system based on visual processing to solve the above problems.

[0006] It should be noted that the above information disclosed in the background section is only used to understand the background of the application concept, and therefore, it can contain information that does not constitute prior art. SUMMARY

[0007] Based on the above problems existing in the prior art, the application aims to provide a vehicle lamp gluing and pressing method and system based on visual processing, which acquires an assembly surface image through visual processing and generates a gluing route, thereby reducing manual intervention.

[0008] The technical solution adopted by the application to solve the technical problems is: a vehicle lamp gluing and pressing method based on visual processing, comprising:

[0009] The processing center receives visual image data collected and transmitted by a collection device, the collection device comprising a structured light sensor and a binocular vision camera, the visual image data comprising three-dimensional point cloud data and two-dimensional RGB data;

[0010] The three-dimensional point cloud data and the two-dimensional RGB data are fused, a three-dimensional model library is established, and the fused data is stored in the three-dimensional model library;

[0011] A gluing path is generated according to known data in the three-dimensional model library and real-time detection data, and glue amount distribution is calculated based on the gluing path;

[0012] An execution signal is generated and sent to an execution device, and the gluing process is verified in a closed loop during the operation of the execution device, and a gluing trajectory data based on coordinates is generated;

[0013] After the gluing is completed, the positioning holes of the vehicle lamp and the vehicle shell are pressed together, and a segmented linear pressing method is used to press the vehicle lamp and the vehicle shell together.

[0014] In the implementation process of the technical scheme of the present application, the visual image data is obtained, and the gluing path is generated based on the visual image data and the real-time detection data, avoiding the manual input and debugging link in the traditional method, and improving the gluing efficiency.

[0015] Further, the fusion processing of the three-dimensional point cloud data and the two-dimensional RGB data includes:

[0016] The three-dimensional point cloud data is subjected to bilateral filtering, and the ICP algorithm is used to preliminarily register the three-dimensional point cloud data, and a three-dimensional coordinate system based on the product is established;

[0017] The two-dimensional RGB image is subjected to parameter calibration, the left and right images obtained by the binocular camera are projected into the three-dimensional coordinate system, and abnormal points are filtered out;

[0018] The point cloud data and the two-dimensional RGB data in the three-dimensional coordinate system are subjected to symmetric feature extraction, and feature matching is performed through auxiliary constraints, to generate a feature point set of the left and right views, and to obtain the three-dimensional coordinates of the vehicle lamp assembly surface.

[0019] Further, the method for symmetric feature extraction of the point cloud data and the two-dimensional RGB data in the three-dimensional coordinate system includes: first, setting the exposure time of the binocular camera based on the ambient light of the area where the product is located, and converting the left and right view data collected by the binocular camera into grayscale images; then, performing local brightness difference correction on the grayscale images of the left and right views, adjusting the brightness of each pixel point by using the mean filtering method, and balancing the grayscale distribution of the left and right views; after the local brightness difference correction, the conjugate feature points of the left and right views are extracted by using the SIFT algorithm, and the conjugate feature points are matched to construct a disparity map, and the three-dimensional coordinates of the vehicle lamp assembly surface are calculated by using the disparity map.

[0020] Further, the exposure time of the binocular camera is inversely proportional to the ambient light intensity, a reference ambient light intensity is set first, and then the exposure time is adjusted according to the ratio of the actual ambient light intensity to the reference ambient light intensity.

[0021] Further, the conjugate feature points refer to feature points at the same physical position in left and right views, and the center coordinates of the lamp assembly surface are obtained using any set of conjugate feature points.

[0022] Further, the glue application path includes glue application parameters, including glue application starting point, glue application ending point, glue application width, glue application speed and glue application amount, wherein the glue application starting point and the glue application ending point are directly determined according to the assembly surface boundary points in the three-dimensional model library.

[0023] Further, the glue application width is determined by the intersection of the basic width and the equipment constraint width, and the basic width is the design effective width of the assembly surface minus twice the single-sided process allowance.

[0024] Further, the equipment constraint width is determined by the glue discharge width range of the execution equipment, the width maximum and minimum values of the assembly surface, and the spreading coefficient, and the equipment constraint width is between the minimum and maximum values of the width of the assembly surface.

[0025] Further, the glue amount distribution of the execution equipment is determined based on the glue discharge efficiency, and the glue discharge efficiency is equal to the target thickness multiplied by the glue application speed and then multiplied by the glue application width.

[0026] A vehicle lamp glue application and pressing system based on visual processing, the system comprising:

[0027] A multi-dimensional acquisition module for processing central visual image data, the visual image data being acquired and transmitted by an acquisition device, the acquisition device including a structured light sensor and a binocular vision camera, the visual image data including three-dimensional point cloud data and two-dimensional RGB data;

[0028] A fusion processing module for fusion processing of the three-dimensional point cloud data and the two-dimensional RGB data, and establishing a three-dimensional model library, and storing the fusion processed data into the three-dimensional model library;

[0029] A glue application path generation module for generating a glue application path based on known data and real-time detection data in the three-dimensional model library, and calculating glue amount distribution based on the glue application path;

[0030] A closed-loop verification module for generating an execution signal and sending it to an execution equipment, and performing closed-loop verification of the glue application process during the operation of the execution equipment, and generating glue application trajectory data based on coordinates;

[0031] A pressing control module for pressing the vehicle lamp and the vehicle shell based on the positioning holes of the vehicle lamp and the vehicle shell after the glue application is completed, and using a segmented linear pressing method to press the vehicle lamp and the vehicle shell.

[0032] The beneficial effects of this application are as follows: This application provides a method and system for applying adhesive to vehicle lights based on vision processing. By acquiring visual image data and generating an adhesive application path based on the visual image data and real-time detection data, it avoids the manual input and debugging steps in traditional methods, improves adhesive application efficiency, and adopts segmented linear pressing during the pressing process to improve pressing efficiency.

[0033] In addition to the purposes, features, and advantages described above, this application has other purposes, features, and advantages. A further detailed description of this application will be provided below with reference to the figures. Attached Figure Description

[0034] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0035] Figure 1 This is a schematic diagram of the overall process of a vision-processing-based automotive lamp adhesive bonding method in this application;

[0036] Figure 2 This is a schematic diagram of the module structure of a vision-processing-based automotive headlight adhesive bonding system according to this application. Detailed Implementation

[0037] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0038] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0039] Example 1: As Figure 1 As shown, this application provides a vision-based method for applying adhesive and pressing automotive lamps. This method is applied to the assembly process of automotive lamps, automating the application and pressing processes through vision processing. This improves the quality of adhesive application and the precision of pressing, effectively ensuring the stability of the automotive lamp production process and reducing human error. The method includes the following steps:

[0040] Step 01: The processing center receives visual image data collected and transmitted by a collection device, which includes a structured light sensor and a binocular vision camera, the visual image data includes three-dimensional point cloud data and two-dimensional RGB data;

[0041] In a conventional automatic compression coating device, offline programming method and teaching programming method are needed to realize device operation, wherein the offline programming method refers to pre-designing operation path and parameters before device operation, so that the coating device performs coating operation according to the predetermined trajectory during operation, and the teaching programming method is to guide the mechanical arm to move along the lamp contour by manual teaching, and import the trajectory data into the device control system, and reproduce the teaching trajectory during device operation, so as to realize automatic coating, but both of them have certain limitations, such as they can only operate on known lamp contour, and cannot adapt to the coating and compression needs of lamps of different models and shapes, and the teaching programming method depends on manual experience, is time-consuming and prone to error, and it is difficult to ensure coating precision and consistency, therefore, in the embodiment, by introducing visual processing technology, the lamp contour image to be coated and compressed is collected, and the lamp contour characteristics are analyzed in real time, and the coating path and compression parameters are automatically generated, so as to avoid the defects in the prior art.

[0042] Among them, the visual image data includes three-dimensional point cloud data and two-dimensional RGB data, the three-dimensional point cloud data is obtained by a structured light sensor, which is arranged on the top of the device, and can obtain three-dimensional point cloud data of the lamp surface, the scanning range covers the lamp shell, the lampshade and the assembly surface, the two-dimensional RGB data is obtained by a binocular vision camera, which is located on both sides of the device, and synchronously collects two-dimensional RGB images containing texture information of the lamp surface;

[0043] In the embodiment, the model of the structured light sensor and the binocular vision camera is not limited, but the resolution and accuracy thereof should meet the needs of lamp coating and compression, and the specific parameter requirements are not described in detail in the embodiment;

[0044] Step 02: The three-dimensional point cloud data and the two-dimensional RGB data are fused and processed, and a three-dimensional model library is established, and the fused and processed data is stored in the three-dimensional model library;

[0045] After obtaining the three-dimensional point cloud data and the two-dimensional RGB data respectively, since they belong to different data types, fusion processing is needed, the three-dimensional point cloud data and the two-dimensional RGB data are registered, error correction and optimization are performed in the registration process, and coordinate system conversion is performed, the fused data is uniformly converted to the device coordinate system, which is convenient for subsequent device execution process, specifically, the fusion processing of the three-dimensional point cloud data and the two-dimensional RGB data includes the following steps:

[0046] Step 201: bilateral filtering is performed on the three-dimensional point cloud data, and ICP algorithm is used for preliminary registration of the three-dimensional point cloud data, and a three-dimensional coordinate system based on the product is established;

[0047] After collecting the three-dimensional point cloud data, due to the noise and redundant information in the collection process, it is necessary to first perform denoising processing, in this embodiment, bilateral filtering algorithm is used for denoising of the three-dimensional point cloud data, bilateral filtering algorithm is a filtering method considering both geometric information and texture characteristics, which can reduce the loss of details while denoising the three-dimensional point cloud data;

[0048] After bilateral filtering, preliminary registration is also needed, in the three-dimensional point cloud data, registration refers to aligning the point cloud data collected under different angles, so that they coincide in the spatial coordinate system, thereby realizing the overall consistency of the point cloud data, in this embodiment, the registration process uses ICP algorithm, ICP algorithm is a registration method based on nearest point iteration, which gradually reduces the error by continuously iterating and optimizing the corresponding relationship between point clouds, until the preset accuracy is reached, thereby completing the preliminary registration of the three-dimensional point cloud data;

[0049] Bilateral filtering algorithm and ICP algorithm are relatively mature technical solutions in the prior art, and their specific implementation processes have been described in detail in related documents, therefore, no detailed description is given;

[0050] Step 202: parameter calibration is performed on the two-dimensional RGB image, the left and right images obtained by the binocular camera are projected into the three-dimensional coordinate system, and the abnormal points are filtered out;

[0051] The binocular camera can obtain left and right views of the surface of the vehicle lamp, through parameter calibration, the two-dimensional image is mapped to the three-dimensional coordinate system, ensuring accurate alignment of the image and the point cloud data, and due to errors, angle differences and other factors in the collection process, some data points are abnormal, therefore, these abnormal points need to be filtered out, when filtering out the abnormal points, the method includes: identifying outliers by statistical threshold method, combining RANSAC algorithm to remove mis-matched points, and then registering the remaining data to fuse the two-dimensional RGB data and the three-dimensional point cloud data;

[0052] RANSAC algorithm is a random sample consensus algorithm, which selects an inlier set in the data by iteration, estimates the model parameters, and thus obtains effective sample data, although it does not have direct abnormal point detection and removal function, but its iteration process can effectively exclude mis-matched points, thereby fusing the two-dimensional RGB data and the three-dimensional point cloud data;

[0053] Step 203: symmetric feature extraction is performed on the point cloud data and two-dimensional RGB data in the three-dimensional coordinate system, and feature matching is performed through auxiliary constraints to generate feature point sets of the left and right views, and three-dimensional coordinates of the vehicle lamp assembly surface are obtained;

[0054] In the process of collecting the vehicle lamp surface image, the light and dark differences caused by different reflection angles and different materials will not only affect the image analysis and processing process, but also affect the subsequent generation of the gluing route. For example, the reflection characteristics of polished plastic and ordinary plastic are different under light, resulting in uneven brightness and darkness of the image, and the two-dimensional RGB image collected by the binocular camera does not have depth information, which leads to the inability to match the three-dimensional point cloud data;

[0055] At the same time, since the data collected by the structured light sensor is discrete point cloud, there is no light and dark difference problem, therefore, in order to prevent the inconsistency between the three-dimensional data and the two-dimensional image features, a three-dimensional coordinate system based on the product is first established through the three-dimensional point cloud data, and then the feature points in the two-dimensional RGB image are mapped into the coordinate system, so that they have spatial correspondence. When performing symmetric feature extraction, it is not necessary to process them separately, but only in the unified three-dimensional coordinate system;

[0056] Specifically, the method for performing symmetric feature extraction on the point cloud data and two-dimensional RGB data in the three-dimensional coordinate system comprises the following steps:

[0057] Firstly, the exposure time of the binocular camera is set based on the ambient light in the area where the product is located, and the left and right view data collected by the binocular camera is converted into a gray image;

[0058] In actual application scenarios, the gluing and pressing scheme based on visual processing needs to consider the influence of ambient light changes on image quality, therefore, the exposure time of the binocular camera is set based on the ambient light in the area where the product is located. In this embodiment, the selection of the exposure amount is determined based on the exposure amount formula. In the prior art, the exposure amount is equal to the product of the ambient light intensity and the exposure time multiplied by the aperture size. In the case where the exposure amount and the aperture size are constant, the ambient light intensity is inversely proportional to the exposure time, therefore, a reference ambient light intensity needs to be set first, and then the exposure time is adjusted according to the ratio of the actual ambient light intensity to the reference ambient light intensity, so as to keep the exposure amount unchanged, and ensure the stability of the image quality. For example, the reference ambient light intensity is set to 1000 lux, and the actual ambient light intensity is 800 lux, then the exposure time needs to be adjusted to 1.25 times of the original time, to ensure that the image remains consistent under different ambient light conditions;

[0059] In addition, the left and right view data collected by the binocular camera needs to be converted into a gray image to reduce the interference of color information on feature extraction. After being converted into a gray image, the images of the left and right views will have two different gray distributions;

[0060] Then the local brightness difference correction is performed on the gray images of the left and right views, and the mean filtering method is used to adjust the brightness of each pixel point to balance the gray distribution of the left and right views.

[0061] Due to the existence of the perspective difference between the left and right views collected by the binocular camera and the lens distortion, there is a difference in the gray distribution between the left and right views, for example, the edge part of the image of the left view is dark, while the right view is relatively bright, which will cause errors when analyzing and directly affect the subsequent generation of the gluing track, therefore, the mean filtering method is used to adjust the brightness of each pixel point to balance the gray distribution of the left and right views, so as to eliminate the influence of the perspective difference and the lens distortion;

[0062] In the prior art, image gray processing is a common image processing method, through the image processed by gray processing, not only the subsequent feature extraction process can be simplified, but also the efficiency and accuracy of image processing can be improved, and the gray processing can preserve the edge information of the image, so as to accurately obtain the edge features of the lamp assembly surface;

[0063] After the local brightness difference correction, the conjugate feature points of the left and right views are extracted by the SIFT algorithm, and the conjugate feature points are matched to construct a disparity map, and the three-dimensional coordinates of the lamp assembly surface are calculated by using the disparity map;

[0064] The conjugate feature points refer to the feature points at the same physical position in the left and right views, by matching these feature points, the spatial position of the lamp assembly surface can be accurately determined, and by using any one set of conjugate feature points, the center coordinates of the lamp assembly surface can be obtained, thereby providing accurate positioning basis for the subsequent assembly process;

[0065] The SIFT algorithm is an image matching algorithm based on scale-invariant features, which can stably extract feature points under different scales and different lighting conditions, based on the algorithm, the conjugate feature points in the left and right views can be extracted and matched to generate a disparity map, the disparity map reflects the horizontal position difference of the same object between the left and right views, then the disparity map is converted into depth information by using the principle of triangulation, through the calculation of the disparity value, the three-dimensional model of the lamp assembly surface can be accurately reconstructed, for example, when the disparity value of a certain conjugate feature point is 5 pixels, it means that the point is 5 centimeters away from the camera in the actual space, by point-by-point calculation of all the conjugate feature points, the complete three-dimensional coordinate contour of the assembly surface can be constructed to determine its general trend;

[0066] It should be noted that the accuracy of the parallax value directly affects the accuracy of the three-dimensional model, so multiple checks need to be performed during the calculation process to ensure that the parallax value of each conjugate feature point is correct, and the final generated three-dimensional coordinate contour can provide reliable positioning reference for lamp assembly. Therefore, in this embodiment, the calculation process of the parallax value is matched and checked. Specifically, the matching and checking process uses the parallax threshold of the conjugate feature points in the left and right views to determine whether the parallax value is within the threshold range. If the parallax value is within the threshold range, it is considered to be valid matching, otherwise it is rejected. For example, if the parallax threshold is set to 2 pixels, the coordinates of a pair of conjugate feature points are (100, 200) and (105, 200), and the parallax value is 5 pixels, then the pair of feature points is determined to be invalid matching and needs to be re-extracted or corrected, thereby optimizing the parallax map and improving the accuracy of the three-dimensional model.

[0067] Step 03: According to the known data in the three-dimensional model library and the real-time detection data, the glue application path is generated, and the glue amount distribution is calculated based on the glue application path;

[0068] After the above steps are completed, the specific glue application path can be generated according to the known data in the three-dimensional model library and the real-time detection data. Since the geometric parameters of the current lamp assembly surface are already stored in the three-dimensional model library, only the determination of the glue application parameters is required to generate an accurate glue application path. The glue application parameters include the glue application starting point, the glue application ending point, the glue application width, the glue application speed and the glue application amount. After determining these parameters, an execution signal can be sent to the execution device to perform glue application according to the glue application path.

[0069] The glue application starting point and the glue application ending point are directly determined according to the boundary points of the assembly surface in the three-dimensional model library. The glue application width needs to be determined by considering the assembly surface width and the glue output of the execution device. Due to process design reasons, a certain process allowance needs to be reserved, so the glue application width needs to meet multiple constraints, including determining the basic width, the device constraint width, and then taking the intersection of the two as the final glue application width to meet the process requirements.

[0070] The basic width is the design effective width of the assembly surface minus twice the single-sided process allowance. The design effective width of the assembly surface is provided by the process file, and the single-sided process allowance is set according to the actual assembly requirements. For example, the design effective width of a batch of lamp assembly surfaces is 20 mm, and the single-sided process allowance is 1 mm, so the basic width is 18 mm.

[0071] The device constraint width is determined by the glue dispensing width range of the execution device, the width maximum and minimum of the assembly surface, and the spreading coefficient. The execution device has an adjustable opening size that determines the glue dispensing amount, and the spreading coefficient is considered, which refers to the natural spreading rate of the glue during the coating process. The glue strip will naturally spread on the assembly surface after being extruded from the nozzle. The device constraint width should be greater than or equal to the product of the spreading coefficient and the glue strip diameter. The final device constraint width should be between the minimum and maximum width of the assembly surface. For example, if the maximum glue dispensing width of the execution device is 15 mm and the spreading coefficient is 1.2, i.e., the glue strip naturally spreads to 18 mm after coating. In order to prevent glue overflow, the device constraint width also needs to be less than or equal to the minimum width of the assembly surface. For example, the minimum width of the assembly surface is 17 mm and the maximum width is 20 mm, so the device constraint width should be set to between 17 mm and 20 mm.

[0072] In addition to ensuring uniformity and preventing glue overflow, the glue dispensing efficiency of the execution device needs to be calculated based on the glue dispensing speed and target thickness to prevent uneven coating or insufficient thickness caused by excessive speed, and glue accumulation caused by slow speed. Specifically, the target glue layer thickness is generally a fixed value. According to industry standards, after calculating the device constraint width, the target thickness and glue dispensing speed can be combined to calculate the glue dispensing efficiency under different glue dispensing conditions. That is, the glue dispensing efficiency is equal to the target thickness multiplied by the glue dispensing speed and then multiplied by the glue dispensing width. Finally, the glue dispensing amount per second is obtained. For example, if the target thickness is 0.5 mm, the glue dispensing speed is 10 mm / s, and the glue dispensing width is 18 mm, the glue dispensing efficiency is 90 mm³ / s. When the device constraint width changes (such as changes in the width of the assembly surface or different car light models), the glue dispensing speed needs to be adjusted without changing the glue dispensing device to match the new device constraint width and ensure the quality of glue dispensing.

[0073] The intersection of the basic width and the device constraint width is taken as the final glue dispensing width. The glue dispensing speed is adjusted based on this final glue dispensing width value during the glue dispensing process to control the glue dispensing amount under different sizes to achieve precise coating. For example, if the basic width is 18 mm and the device constraint width is 17-20 mm, the final glue dispensing width should be set to 17-18 mm to meet the assembly precision requirements and simultaneously control the glue dispensing speed to ensure the same glue dispensing efficiency under different widths to meet the consistency of the glue dispensing process.

[0074] Step 04: Generate an execution signal and send it to the execution device. Perform closed-loop verification on the glue dispensing process during the execution of the execution device and generate glue dispensing trajectory data based on coordinates.

[0075] After the calculation of the foregoing steps, the calculation result also needs to be converted into an execution signal, for example, the parameters such as the glue coating width and speed are encoded into machine language by the PLC control system, and are transmitted to the execution device. After receiving the signal, the device performs glue coating according to the preset track. Meanwhile, the glue coating process is monitored in real time by the acquisition device. When the glue coating track deviates from the preset path or the glue layer thickness is abnormal, an adjustment signal is immediately fed back, and the glue coating track data is generated according to the coordinates. The glue coating track data is composed of coordinates and time stamps, which record the glue coating state of each coordinate at a specific time point, facilitating subsequent analysis and optimization of the glue coating process and ensuring the glue coating precision.

[0076] Step 05: After the glue coating is completed, the positioning holes of the vehicle lamp and the vehicle shell are used for pressing, and a segmented linear pressing method is adopted to press the vehicle lamp and the vehicle shell.

[0077] After the glue coating process is completed, the vehicle lamp and the vehicle shell need to be pressed. In this embodiment, the pressing device uses an existing device, such as an automatic pressing machine, and the vehicle lamp and the vehicle shell are provided with positioning holes in advance. After obtaining the coordinate information of the positioning holes, the pressing device gradually presses the two by adopting a segmented linear pressing method. The segmented linear pressing method outputs different pressing speeds based on the pressing period. In the initial stage, low-speed contact is adopted to avoid collision between the workpiece and the device. In the middle stage, uniform pressure is applied. In the end stage, continuous pressure is maintained, so as to ensure uniform and firm pressing.

[0078] For example, in the initial stage, low-speed contact is performed at a speed less than or equal to 0.1 mm / s. After the vehicle lamp and the vehicle shell are in contact, uniform pressure is applied at a speed of 0.5 mm / s to 1 mm / s. In the end stage, continuous pressure is maintained at a speed of 0.1 mm / s for 3 to 5 s, so as to complete the pressing process.

[0079] Embodiment Two: As shown in Figure 2 The present application also proposes a vehicle lamp glue coating and pressing system based on visual processing. The system runs the glue coating and pressing method in Embodiment One, realizes the automation of glue coating and pressing through visual processing, improves the glue coating quality and pressing precision, effectively guarantees the stability in the vehicle lamp production process, and reduces the manual error. The system comprises:

[0080] A multi-dimensional acquisition module is configured to receive visual image data from the center. The visual image data is acquired and transmitted by an acquisition device, which comprises a structured light sensor and a binocular vision camera. The visual image data comprises three-dimensional point cloud data and two-dimensional RGB data.

[0081] A fusion processing module is configured to perform fusion processing on the three-dimensional point cloud data and the two-dimensional RGB data, and establish a three-dimensional model library. The data after fusion processing is stored in the three-dimensional model library.

[0082] The glue applying path generating module is configured to generate a glue applying path according to known data in the three-dimensional model library and real-time detection data, and calculate glue amount distribution based on the glue applying path;

[0083] The closed loop verification module is configured to generate an execution signal and send the execution signal to an execution device, to perform closed loop verification on the glue applying process during operation of the execution device, and to generate glue applying track data based on coordinates.

[0084] The pressing control module is configured to press the vehicle lamp and the vehicle shell after the glue applying is completed, and to press the vehicle lamp and the vehicle shell in a segmented linear pressing manner based on positioning holes of the vehicle lamp and the vehicle shell.

[0085] The above merely describes preferred embodiments of the present application and is not intended to limit the present application. Various modifications and changes can be made by those skilled in the art based on the spirit and principles of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for applying adhesive and bonding automotive lights based on vision processing, characterized in that: include: The processing center receives visual image data, which is acquired and transmitted by acquisition devices, including structured light sensors and binocular vision cameras. The visual image data includes three-dimensional point cloud data and two-dimensional RGB data. The 3D point cloud data and 2D RGB data are fused together, and a 3D model library is established. The fused data is then stored in the 3D model library. Based on known data in the 3D model library and real-time detection data, a glue application path is generated, and the glue amount distribution is calculated based on the glue application path. An execution signal is generated and sent to the execution device. During the operation of the execution device, the glue application process is verified in a closed loop, and glue application trajectory data based on coordinates is generated. After the adhesive is applied, the headlights and the body shell are pressed together using the positioning holes, and a segmented linear pressing method is used to press the headlights and the body shell together. The fusion processing of 3D point cloud data and 2D RGB data includes: Bilateral filtering is performed on the 3D point cloud data, and the ICP algorithm is used to perform preliminary registration of the 3D point cloud data to establish a 3D coordinate system based on the product. The parameters of the 2D RGB image are calibrated, the left and right images acquired by the binocular camera are projected into the 3D coordinate system, and outliers are filtered out. Symmetrical features are extracted from point cloud data in a 3D coordinate system and 2D RGB data. Feature matching is performed through auxiliary constraints to generate feature point sets for the left and right views. The 3D coordinates of the headlight assembly surface are obtained. A 3D coordinate system based on the product is established through the 3D point cloud data. Then, the feature points in the 2D RGB image are mapped to this coordinate system to make the two have spatial correspondence. The method for extracting symmetrical features from point cloud data and 2D RGB data in a 3D coordinate system includes: first, setting the exposure time of the binocular camera based on the ambient light of the product's location, and converting the left and right view data acquired by the binocular camera into grayscale images; then, correcting local brightness differences in the grayscale images of the left and right views, using mean filtering to adjust the brightness of each pixel to balance the grayscale distribution of the left and right views; after correcting local brightness differences, extracting conjugate feature points of the left and right views using the SIFT algorithm, matching the conjugate feature points to construct a disparity map, and using the disparity map to calculate the 3D coordinates of the headlight assembly surface; The process of calculating disparity values ​​is used to perform matching verification. The matching verification process includes judging the disparity threshold of the conjugate feature points of the left and right views. If the disparity value is within the threshold range, it is considered a valid match; otherwise, it is rejected.

2. The method for applying adhesive to and bonding automotive lights based on vision processing according to claim 1, characterized in that: The exposure time of a binocular camera is inversely proportional to the ambient light intensity. First, set a reference ambient light intensity, and then adjust the exposure time according to the ratio of the actual ambient light intensity to the reference ambient light intensity.

3. The method for applying adhesive to and bonding automotive lights based on vision processing according to claim 1, characterized in that: The conjugate feature points refer to feature points that have the same physical position in the left and right views, and the center coordinates of the headlight assembly surface can be obtained by using any set of conjugate feature points.

4. The method for applying adhesive to and bonding automotive lights based on vision processing according to claim 1, characterized in that: The adhesive application path includes adhesive application parameters, which include the adhesive application start point, adhesive application end point, adhesive application width, adhesive application speed, and adhesive application amount. The adhesive application start point and adhesive application end point are directly determined based on the assembly surface boundary points in the 3D model library.

5. The method for applying adhesive to and bonding automotive lights based on vision processing according to claim 4, characterized in that: The adhesive application width is determined by the intersection of the base width and the equipment constraint width. The base width is the effective design width of the assembly surface minus twice the single-sided process allowance.

6. The method for applying adhesive to and bonding automotive lights based on vision processing according to claim 5, characterized in that: The equipment constraint width is determined by the dispensing width range of the executing equipment, the maximum and minimum width of the assembly surface, and the spreading coefficient. The equipment constraint width is between the minimum and maximum width of the assembly surface.

7. The method for applying adhesive to and bonding automotive lights based on vision processing according to claim 6, characterized in that: The amount of adhesive dispensed by the equipment is determined based on the dispensing efficiency, which is equal to the target thickness multiplied by the dispensing speed and then by the dispensing width.

8. A vision-processing-based automotive lamp adhesive bonding system, used to implement the vision-processing-based automotive lamp adhesive bonding method as described in any one of claims 1 to 7, characterized in that: The system includes: The multidimensional acquisition module is used to process the visual image data received by the center. The visual image data is acquired and transmitted by the acquisition device, which includes a structured light sensor and a binocular vision camera. The visual image data includes three-dimensional point cloud data and two-dimensional RGB data. The fusion processing module is used to fuse 3D point cloud data and 2D RGB data, and to build a 3D model library, storing the fused data into the 3D model library. The glue application path generation module is used to generate a glue application path based on known data in the 3D model library and real-time detection data, and to calculate the glue amount distribution based on the glue application path. The closed-loop verification module is used to generate execution signals and send them to the execution device. During the operation of the execution device, it performs closed-loop verification of the glue application process and generates glue application trajectory data based on coordinates. The pressing control module is used to press the headlights and body shell together based on the positioning holes after the adhesive is applied, and adopts a segmented linear pressing method to press the headlights and body shell together; The fusion processing of 3D point cloud data and 2D RGB data includes: Bilateral filtering is performed on the 3D point cloud data, and the ICP algorithm is used to perform preliminary registration of the 3D point cloud data to establish a 3D coordinate system based on the product. The parameters of the 2D RGB image are calibrated, the left and right images acquired by the binocular camera are projected into the 3D coordinate system, and outliers are filtered out. Symmetrical features are extracted from point cloud data in a 3D coordinate system and 2D RGB data. Feature matching is performed through auxiliary constraints to generate feature point sets for the left and right views. The 3D coordinates of the headlight assembly surface are obtained. A 3D coordinate system based on the product is established through the 3D point cloud data. Then, the feature points in the 2D RGB image are mapped to this coordinate system to make the two have spatial correspondence. The method for extracting symmetrical features from point cloud data and 2D RGB data in a 3D coordinate system includes: first, setting the exposure time of the binocular camera based on the ambient light of the product's location, and converting the left and right view data acquired by the binocular camera into grayscale images; then, correcting local brightness differences in the grayscale images of the left and right views, using mean filtering to adjust the brightness of each pixel to balance the grayscale distribution of the left and right views; after correcting local brightness differences, extracting conjugate feature points of the left and right views using the SIFT algorithm, matching the conjugate feature points to construct a disparity map, and using the disparity map to calculate the 3D coordinates of the headlight assembly surface; The process of calculating disparity values ​​is used to perform matching verification. The matching verification process includes judging the disparity threshold of the conjugate feature points of the left and right views. If the disparity value is within the threshold range, it is considered a valid match; otherwise, it is rejected.

Images