LED lamp bead production detection system based on machine vision
By adopting the multi-dimensional detection process and the intelligent optimization of the execution learning module in the LED lamp bead production and detection system, the problems of low detection accuracy and insufficient system adaptability are solved, and the accurate detection of LED lamp bead defects and efficient optimization of the production process are achieved.
Patent Information
- Application Number
- CN202510304981.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-14
- Publication Date
- 2025-06-27
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing LED lamp bead production and detection system has low detection accuracy in light changes, environmental interference and complex backgrounds, especially in the detection of tiny defects and uneven light sources; the system lacks adaptability and is difficult to deal with the shadow and reflection problems caused by uneven light; failures or abnormal situations during the production process cannot be reported in time, resulting in quality fluctuations and low production efficiency.
The multi-dimensional detection process is adopted, including position, shape, light source and function detection, combined with advanced visual acquisition technologies such as ASIC image processing and edge detection to capture and accurately locate the defects of LED lamp beads in real time. The execution learning module performs intelligent analysis and optimization based on real-time detection results and production data, automatically adjusts production parameters and processes, and optimizes equipment scheduling to avoid equipment overload.
Through multi-dimensional detection process and intelligent optimization, accurate detection and removal of LED lamp bead defects are achieved, product qualification rate and production efficiency are improved, and quality fluctuations and equipment failures are reduced.
Smart Images

Figure CN120212861A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of machine vision production inspection, and specifically to an LED lamp bead production inspection system based on machine vision. Background Art
[0002] The LED lamp bead production inspection system based on machine vision utilizes computer vision and image processing technologies. Through automated recognition and analysis, it can monitor the production process of LED lamp beads in real time. These systems adopt high-resolution cameras and advanced image processing algorithms, combined with deep learning and machine learning technologies, and can detect various quality indicators of LED lamp beads, including appearance defects, dimensional accuracy, and luminous effects. Compared with traditional manual inspection, the machine vision system has higher inspection speed and accuracy, and at the same time reduces the problems caused by human errors and fatigue.
[0003] However, despite the significant advantages of this technology, the existing LED lamp bead production inspection systems still have some deficiencies. First, limited by light changes, environmental interference, and complex backgrounds, the detection accuracy is relatively low, especially in the detection of minor defects (such as bubbles and cracks) and uneven light sources. Second, it is necessary to retrain the model or adjust the algorithm according to different batches of LED lamp beads, resulting in insufficient adaptability of the system. Third, it is difficult to handle problems such as shadows and reflections caused by uneven illumination. In addition, faults or abnormal situations that occur during the production process often cannot be timely feedback and adjusted, resulting in quality fluctuations or low production efficiency. Summary of the Invention
[0004] The purpose of the present invention is to solve the above problems and propose an LED lamp bead production inspection system based on machine vision.
[0005] The purpose of the present invention can be achieved through the following technical solutions: An LED lamp bead production inspection system based on machine vision, including: a production inspection module;
[0006] The production inspection module is responsible for receiving image data and performing inspections. The inspections include position inspection, shape inspection, light source inspection, and function inspection;
[0007] Performing position inspection: Using edge detection to obtain the feature points and corresponding coordinates of the LED lamp beads, and determining the position of the LED lamp beads in the image; replacing with a high-power lens, extracting the edge contours of the LED lamp beads of the target object and the lamp bead body; using the intersection point of two adjacent edge contours as the origin of the plane coordinate system, calculating vectors and matching the preset standard vector interval; if the vector is not within the standard interval, generating an abnormal rejection instruction; if it is within the interval, then performing subsequent processing; selecting the preset standard vector interval according to the LED lamp bead specifications, calculating through the image size, using the calibration formula to convert pixels into actual sizes, and matching the specifications in the database;
[0008] Perform shape detection: Install an industrial camera on a rotating slide rail, and cooperate with a fill light to rotate with the camera; the industrial camera captures images of the object surface to obtain information at different angles; the ASIC processes the images for denoising, enhancement, and standardization; use an image segmentation algorithm to extract the shadow part and analyze defects such as bubbles, cracks, and pores.
[0009] Perform light source detection: Detect the brightness of the LDE light source: Calculate the brightness by weighted calculation of the gray values of the image pixels, convert it into a grayscale image and perform weighted processing on different regions, compare the calculation result with the standard brightness value, and generate a brightness anomaly instruction if they do not match; otherwise, perform spectral measurement; Detect the color difference of the LDE light source: Measure and calculate the light output intensity using a spectral sensor, measure the color temperature of the light source and compare it with the CIE standard light source color temperature curve, and generate a color temperature anomaly instruction if it deviates from the standard; otherwise, perform further detection according to the color temperature range screening.
[0010] Perform function detection: Measure the initial parameters in a standard environment and place the lamp beads in a temperature and humidity control environmental chamber; regularly measure parameters such as brightness, color temperature, current, and voltage under different working conditions, and calculate the percentage change relative to the initial state; obtain the stability coefficient S by weighted calculation of each change rate and compare it with a preset stability threshold, and generate a function anomaly instruction if S is less than the threshold.
[0011] As a preferred embodiment of the present invention, it further includes a visual acquisition module and an execution learning module.
[0012] The visual acquisition module captures real-time images of LED lamp beads through an industrial camera or a camera, and transmits the data to the production detection module through the industrial Internet or a dedicated communication protocol.
[0013] The execution learning module records the abnormal events in each rejection operation, and synchronously uploads the detection results and production environment data; automatically adjusts the production parameters and optimizes the production process according to real-time analysis; adjusts the workload of the production equipment if equipment problems are found; optimizes the production line and automatically adjusts the process.
[0014] As a preferred embodiment of the present invention, the specific process of generating an abnormal rejection instruction is as follows:
[0015] Several LED bead feature points are obtained through edge detection, and each LED bead feature point corresponds to an LED bead; then, a plane coordinate system is established based on the picture size in the image to obtain the coordinate points corresponding to several LED bead feature points, that is, the positions of several LED beads in the image detection area are obtained; the industrial camera is replaced with a high-power lens, and the edge detection obtained contour is used to extract the feature information of the target object to obtain the edge contours of several LED beads and the LED bead body; a plane coordinate system is established based on two adjacent edge contours of the LED bead body edge contour, and the two adjacent edge contours are perpendicular to each other; the intersection point is the origin of the plane coordinate system, and the corresponding coordinate points of the edge contour of the LED bead are obtained. The coordinate points with the closest distance between the LED bead body edge contour and the LED bead edge contour are connected, and the vector between them is calculated; the preset position data in the database is retrieved, and the corresponding preset standard vector interval is selected according to the specifications of the detected LED beads; the calculated vector is matched with the corresponding preset standard vector interval; if the calculated vector does not belong to the corresponding preset standard vector interval, a position anomaly rejection instruction is generated.
[0016] As a preferred embodiment of the present invention, the specific process of matching the specifications in the database is as follows:
[0017] Obtain the geometric shapes of the LED bead and the lamp body; convert the pixel unit to the actual size through the calibration formula; if the diameter d of the LED bead and the diameter D of the lamp body are obtained, calculate through the formula: P d is the pixel length of the LED bead; P t is the pixel length corresponding to a known size object in the image; the actual size is the physical unit obtained during calibration; then, match the calculated sizes d and D with the specifications in the preset position data in the database.
[0018] As a preferred embodiment of the present invention, the specific process of performing shape detection is as follows:
[0019] Install an industrial camera on a rotating slide rail to perform a rotational movement around the target area; install a fill light at the relative opposite position of the industrial camera and make it perform a relative movement along with the rotating slide rail; comprehensively image the surface of the object from different angles and directions to capture the subtle changes on the surface; select a fixed brightness as the initial brightness of the fill light; during the rotation of the camera, the industrial camera continuously captures images of the target surface and obtains image information from different angles; the vision acquisition module uses ASIC to perform denoising, enhancement, and normalization processing on the acquired images to obtain analysis images; through analyzing the illumination direction of the images, combining the relative positions of the camera and the fill light, and extracting the shadow part through an image segmentation algorithm; the specific extraction of the shadow part is as follows: segment the image into multiple regions to highlight the defect regions such as bubbles, cracks, and pores; based on the pixel characteristics within the region and compare them with the preset pixel characteristics to determine whether the region is a bubble, crack, or pore; if the pixel characteristics of the region belong to the preset pixel characteristics, then determine that the region is a bubble, crack, or pore and generate a shape detection and rejection instruction; otherwise, if the pixel characteristics of the region do not belong to the preset pixel characteristics, increase the brightness of the fill light and re-perform the acquisition and detection until it passes.
[0020] As a preferred embodiment of the present invention, the specific process of performing the light source detection is as follows:
[0021] Measure the light output intensity of the LED light source through a spectral sensor; through P(λ i ) is the light intensity of the i-th wavelength, V(λ i ) is the photometric function at this wavelength, Δλ is the change in wavelength, and K1 is a constant coefficient; the spectrometer is also used to measure the color temperature of the light source, and the color temperature refers to the color shift degree of the light source, specifically: use the color temperature approximation method, by measuring a specific wavelength and using a predetermined formula to quickly calculate the color temperature: R is the light intensity of the red channel, g is the light intensity of the green channel, and b is the light intensity of the blue channel; the calculated color temperature value needs to be compared with the light source color temperature curve of the CIE standard to determine whether the LED light source meets the predetermined color temperature range; if the color temperature result deviates from the set standard value, then there may be a color quality problem with the LED light source and a color temperature anomaly removal instruction is generated; otherwise, perform screening according to the predetermined color temperature range to obtain LED lamp beads corresponding to the color temperature.
[0022] As a preferred embodiment of the present invention, the specific process of performing the function detection is as follows:
[0023] Analyze the stability of the LED lamp beads. Set a standard environment to measure the initial state of the LED lamp beads, and record the initial brightness, color temperature, current, and voltage parameters. Place the LED lamp beads in a temperature and humidity control environmental chamber and operate them for a certain period of time under different working conditions. During the test, regularly measure the brightness, color temperature, and electrical characteristic parameters of the LED lamp beads, and record the performance data at each time point. Then calculate the stability index. At each time point, calculate the percentage change of each performance parameter relative to the initial state: Let Lt be the brightness after t hours, Tt be the color temperature after t hours, It and Vt be the current and voltage after t hours. Then, perform a weighted calculation on the brightness change rate, color temperature change rate, current change rate, and voltage change rate to output the LED stability coefficient S. Compare the LED stability coefficient S with the preset stability threshold. If the LED stability coefficient S is less than the preset stability threshold, generate a function abnormality removal instruction.
[0024] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0025] 1. Through a multi-dimensional detection process (including detections of position, shape, light source, function, etc.) and by combining advanced vision acquisition technologies (such as ASIC image processing and edge detection, etc.), the present invention can capture and accurately locate the defects of LED lamp beads in real time, such as bubbles, cracks, color differences, etc. The system performs rejection through precise robotic arm operations, which can not only reduce errors during the process but also ensure the comprehensiveness and accuracy of detection in multiple dimensions, effectively improving the product qualification rate.
[0026] 2. The execution learning module of the present invention can perform intelligent analysis and optimization based on real-time detection results and production data, automatically adjust production parameters, optimize the production process, and avoid overloading of equipment through equipment scheduling strategies. By combining deep learning and path planning algorithms, it can achieve efficient and flexible production scheduling during the detection and rejection process, improve the overall efficiency of the production line, and reduce quality fluctuations caused by equipment failures or improper settings. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] For the convenience of those skilled in the art to understand, the present invention will be further described below with reference to the accompanying drawings.
[0028] Figure 1 It is the principle flow chart of the present invention. DETAILED DESCRIPTION OF THE INVENTION
[0029] The technical solution of the present invention will be clearly and completely described below in conjunction with embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
[0030] It should be understood that the terms "including" and "comprising" used in the specification and claims of this disclosure indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.
[0031] It should also be understood that the terms used in this disclosure specification are only for the purpose of describing specific embodiments and are not intended to limit this disclosure. As used in this disclosure specification and claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms. It should also be further understood that the term "and / or" used in this disclosure specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.
[0032] Please refer to Figure 1 As shown, the LED lamp bead production detection system based on machine vision includes: a vision acquisition module, a production detection module, and an execution learning module;
[0033] The vision acquisition module is responsible for capturing images of LED lamp beads in real time through an industrial camera or a camera; through the industrial Internet or a dedicated communication protocol (such as Ethernet, GigE Vision, etc.), the image data is transmitted to the production detection module; it should be noted that: the vision acquisition module also uses an ASIC (application-specific integrated circuit) to denoise, enhance, and standardize the captured images for subsequent analysis; the ASIC is usually optimized for specific denoising or enhancement algorithms and can perform outstandingly in terms of low power consumption and high efficiency.
[0034] The production detection module is responsible for receiving the image data sent by the vision acquisition module and performing detections; including position detection, shape detection, light source detection, and function detection;
[0035] Perform position detection to obtain several LED bead feature points through edge detection, and each LED bead feature point corresponds to an LED bead; then, based on the frame in the image, establish a plane coordinate system to obtain the coordinate points corresponding to several LED bead feature points, that is, obtain the positions of several LED beads in the image detection area; when the industrial camera replaces the high-power lens and uses the contour obtained by edge detection, further extract the feature information of the target object to obtain the edge contours of several LED beads and the LED bead body; similarly, based on two adjacent edge contours of the LED bead body, establish a plane coordinate system, and the two adjacent edge contours are perpendicular to each other; the intersection point is the origin of the plane coordinate system, obtain the corresponding coordinate points of the edge contour of the LED bead, connect the coordinate points with the closest distance between the edge contour of the LED bead body and the edge contour of the LED bead, and calculate the vector between them; retrieve the preset position data in the database, and select the corresponding preset standard vector interval according to the specifications of the detected LED beads; match the calculated vector with the corresponding preset standard vector interval; if the calculated vector does not belong to the corresponding preset standard vector interval, generate a position anomaly rejection instruction; if the calculated vector belongs to the corresponding preset standard vector interval, then perform further processing, such as encapsulation, etc.;
[0036] It should be noted that: the corresponding preset standard vector interval is selected according to the specifications of the detected LED beads, and is obtained by image acquisition and recognition to calculate the size of the edge contours of the LED beads and the LED bead body. Specifically: obtain the geometric shapes of the LED bead and the lamp body; convert the pixel unit to the actual size through the calibration formula; if the diameter d of the LED bead and the diameter D of the lamp body are obtained, calculate through the formula: Where: P d is the pixel length of the LED bead; P t is the pixel length corresponding to a known-size object (such as a calibration board) in the image; "actual size" is the physical unit obtained during calibration; then match the calculated sizes d and D with the specifications in the preset position data in the database;
[0037] Perform shape detection; install an industrial camera on a rotating slide rail and perform a rotational motion around the target area; install a fill light at the relative reverse position of the industrial camera and make it perform relative motion along with the rotating slide rail; through such settings, it is possible to comprehensively image the surface of an object from different angles and directions and capture the subtle changes on the surface; select a fixed brightness as the initial brightness of the fill light to ensure stable light sources and avoid affecting the image quality due to light changes; the role of the fill light is to eliminate shadows, enhance the details of the object surface, and reduce the impact caused by uneven reflection or light transmission; during the rotation of the camera, the industrial camera continuously captures images of the target surface and obtains image information from different angles; the visual acquisition module also uses an ASIC (Application Specific Integrated Circuit) to perform denoising, enhancement, and normalization processing on the captured images to obtain analysis images; in the images, defects such as bubbles, cracks, and pores usually appear as local dark areas (shadow areas); by analyzing the light direction of the images, combining the relative positions of the camera and the fill light, and using an image segmentation algorithm to extract the shadow part;
[0038] The specific process of using the image segmentation algorithm to extract the shadow part is as follows: segment the image into multiple regions to highlight defect regions such as bubbles, cracks, and pores; based on the pixel features within the region (such as color, brightness changes, etc.) and compare them with the preset pixel features to determine whether the region is a bubble, crack, or pore; if the pixel features of the region belong to the preset pixel features, then determine that the region is a bubble, crack, or pore and generate a shape detection rejection instruction; conversely, if the pixel features of the region do not belong to the preset pixel features, then increase the brightness of the fill light and re-perform the acquisition and detection until it passes; and perform further detection;
[0039] Perform light source detection, and the light source detection includes LDE light source brightness detection and light source color difference detection;
[0040] For the LDE light source brightness detection: quantify the brightness of the light source by analyzing the pixel gray values of the images; calculate for each image pixel, and calculate the gray value of each pixel through weighted calculation: L i = 0.299×R + 0.587×G + 0.587×B to obtain the gray value L i , where R, G, and B are the pixel values of the red, green, and blue channels respectively; convert the color image into a gray image; where L iis the gray value of the i-th pixel, and N is the total number of pixels in the image; further, according to the illumination distribution in different regions of the image, different regions in the image are weighted; for example, the central region may require a higher weight, while the edge region may receive less attention; this weighting process can ensure that the brightness value more conforms to the actual light source distribution; finally, the calculated brightness value will be compared with a predetermined standard brightness value. If the calculated brightness value does not belong to the predetermined standard brightness value, a brightness anomaly removal instruction will be generated; otherwise, spectral measurement will be performed;
[0041] Color difference detection of LDE light source: The spectral sensor can accurately measure the light output intensity (lumen value) of the LED light source; it is calculated through a formula (based on international standards such as the CIE photometric function): where P(λ i ) is the light intensity of the i-th wavelength, V(λ i ) is the photometric function at this wavelength, Δλ is the change in wavelength, and K1 is a constant coefficient; the spectrometer is also used to measure the color temperature of the light source. Color temperature refers to the color deviation degree of the light source, and the unit is usually K (Kelvin). Specifically: Using the color temperature approximation method (a calculation method based on the spectral line characteristics of the light source and the approximate relationship between color temperature and spectral form), by measuring a specific wavelength (such as the characteristic wavelength of the CIE standard), a predetermined formula is used to quickly calculate the color temperature; through the formula: where R is the light intensity of the red channel, g is the light intensity of the green channel, and b is the light intensity of the blue channel; the calculated color temperature value needs to be compared with the CIE standard light source color temperature curve to determine whether the LED light source meets the predetermined color temperature range. The predetermined color temperature range includes warm white light (2700K–3200K), natural white light (4000K–5000K), and cold white light (6000K–6500K); if the color temperature result deviates from the set standard value, the LED light source may have color unevenness or quality problems, and a color temperature anomaly removal instruction will be generated; otherwise, screening will be performed according to the predetermined color temperature range to obtain LED lamp beads corresponding to the color temperature; and further detection will be carried out;
[0042] Perform functional detection, mainly for the stability of the LED lamp beads, specifically:
[0043] Analyze the stability of the LED lamp beads. Set a standard environment (e.g., 25°C, rated voltage and current conditions) to measure the initial state of the LED lamp beads, and record the initial parameters such as brightness, color temperature, current, and voltage; Initial parameters: L0 (initial brightness), T0 (initial color temperature), V0 (initial voltage), I0 (initial current); Place the LED lamp beads in a temperature and humidity control environmental chamber and operate them for a certain period of time under different working conditions (e.g., high temperature, high humidity, high current, etc.); During the test, regularly measure the parameters such as the brightness, color temperature, and electrical characteristics of the LED lamp beads, and record the performance data at each time point; Then calculate the stability index; At each time point, calculate the percentage change of each performance parameter relative to the initial state: Among them, Lt is the brightness after t hours, Tt is the color temperature after t hours, It and Vt are the current and voltage after t hours; Then perform a weighted calculation on the brightness change rate, color temperature change rate, current change rate, and voltage change rate, and output the LED stability coefficient S through the formula: S = brightness change rate × ω1 + color temperature change rate × ω2 + current change rate × ω3 + voltage change rate × ω4, where ω1, ω2, ω3, and ω4 are the preset weights for each parameter, and the value range is (0, 1), and the sum of the preset weights is 1 (i.e., ω1 + ω2 + ω3 + ω4 = 1), which is adjusted according to the specific situation; The setting of the preset weights depends on the sensitivity of each parameter to different applications; Compare the LED stability coefficient S with the preset stability threshold. If the LED stability coefficient S is less than the preset stability threshold, generate a function abnormality removal instruction;
[0044] The production detection module sends control instructions (the control instructions are collectively referred to as function abnormality removal instructions, color temperature abnormality removal instructions, brightness abnormality removal instructions, shape detection rejection instructions, and position abnormality rejection instructions) to the execution learning module according to the detection results, and the execution learning module takes corresponding operations based on these instructions;
[0045] The execution learning module receives the detection results fed back by the production detection module, mainly including position detection, shape detection, light source detection, and function detection information, and determines which LED lamp beads are defective; through the position detection process, the coordinate information of each LED lamp bead within the detection area is obtained; based on the current position of the robotic arm and the position coordinates of the LED lamp beads, the execution learning module calculates the optimal movement path of the robotic arm through path planning algorithms (such as the A* algorithm, Dijkstra algorithm, or path planning methods based on deep learning); according to the results of path planning, the execution learning module generates a series of robotic arm movement instructions; the instructions include the joint angles, grasping positions, grasping forces, etc. of the robotic arm to ensure that the robotic arm can accurately move to the target position and perform the rejection operation; after the robotic arm reaches the position of the target LED lamp bead, it performs a grasping action; through precise grasping force control, the robotic arm removes the LED lamp bead from its original position; the removed LED lamp bead is safely placed in the designated rejection area or recycling bin; throughout the process, the execution learning module continuously monitors the movement state of the robotic arm, receives feedback information in real time to ensure the smooth progress of the rejection operation; if any abnormality occurs (such as unstable grasping or incorrect path), the path will be immediately adjusted and re-planned; within one cycle, the execution learning module may receive the position coordinates of multiple LED lamp beads that need to be rejected; based on factors such as priority and distance, the optimal path for each target will be calculated one by one, and all the lamp beads that need to be rejected will be processed through multiple iterations; after each rejection operation is completed, the robotic arm returns to the starting position or the next target position to continue the next round of operations;
[0046] The execution learning module also conducts abnormal traceability analysis, specifically:
[0047] The execution learning module records the abnormal events that occur during each rejection operation, including the defect types of the LED lamp beads (such as bubbles, cracks, color differences, etc.), the time and location of the abnormality, and other data; all the detection results are synchronously uploaded to the execution learning module, and each rejected lamp bead is marked with the specific defect type, such as "crack", "unqualified color temperature", "bubble"; during the rejection process, the execution learning module also collects production environment data, such as temperature, humidity, illumination intensity, current, etc.;
[0048] Based on the results of real-time analysis, the execution learning module automatically adjusts the production parameters and optimizes the production process; for example, if it is found that too high packaging temperature will cause the brightness of the LED lamp beads to decay rapidly, the system will automatically adjust the temperature control parameters of the packaging equipment to reduce the occurrence of this problem;
[0049] In terms of equipment scheduling and optimization, the execution learning module adjusts the workload of production equipment through equipment scheduling strategies. When detecting performance fluctuations or signs of failure in a certain equipment, the system can automatically balance the load to avoid individual equipment running overloaded for a long time, resulting in quality fluctuations. For example, in the PCB board soldering process, if the peak temperature of an automatic soldering machine is abnormal or the solder supply is unstable, other soldering equipment will be scheduled to share the task. In the LED packaging process, if the encapsulation glue curing of a certain packaging equipment is abnormal, it can be temporarily switched to a standby equipment to ensure stable production. At the same time, the execution learning module combines predictive maintenance strategies. Based on the historical failure modes of equipment, it warns of possible equipment failures in advance and arranges maintenance during off-peak periods, such as lubrication, component replacement or software update, to reduce unplanned downtime. For example, when the nozzle of an SMT mounter detects abnormal pressure, the system will automatically adjust the air supply pressure and issue cleaning or replacement instructions when necessary;
[0050] In terms of production line optimization, the execution learning module can automatically adjust the production process and optimize each link; analyze production bottlenecks. If it is found that the processing time of a certain process is too long or there is serious material backlog, it will suggest or automatically execute adjustments, such as increasing the processing speed of the process or adding additional equipment. For example, when the die bonding process of LED chips is slower than the wire bonding process, it is recommended to increase die bonding equipment or optimize the dispensing accuracy to improve efficiency. At the same time, the system will dynamically adjust the working rhythm of the equipment to ensure the coordination of the beats of each link and avoid blockage of the front-end assembly process caused by too slow detection process beats. In addition, the execution learning module can dynamically optimize production process parameters. For example, when the packaging temperature is too high and affects the brightness attenuation of LEDs, the temperature control strategy will be automatically adjusted to ensure stable process conditions. For example, monitor the dispensing viscosity and automatically adjust the dispensing pressure and flow rate to avoid too much or too little glue affecting the packaging quality. During the reflow soldering process, fine-tune the heating curve according to real-time temperature monitoring data to ensure stable welding quality. In practical applications, the deployment of the execution learning module includes the following steps: First, install intelligent sensors, such as temperature sensors, vibration sensors, current monitors, etc., on the corresponding key production equipment and connect them to the cloud data analysis platform to achieve data collection and monitoring; collect equipment operation data, train the AI model, establish an equipment health status model, and predict potential production anomalies; monitor the equipment status in real time during the production process, give early warnings in time when anomalies are found, and execute adjustments when necessary, such as adjusting the welding temperature or switching to standby equipment; if it is detected that a certain equipment is operating overloaded, automatically schedule tasks to the standby equipment and optimize the production line rhythm for the production bottleneck link to ensure smooth production; the execution learning module continuously adjusts and optimizes the optimization strategy through continuous optimization and feedback learning to improve production scheduling and optimization efficiency. For example, if a certain optimization strategy has limited effect in practice, re-evaluate and improve the adjustment method; through this series of intelligent optimizations, the execution learning module can greatly improve the utilization rate of production equipment, optimize the production rhythm, reduce quality fluctuations, and achieve more stable and efficient production operations.
[0051] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only the specific implementation manners. Obviously, according to the content of this specification, many modifications and changes can be made. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the present invention, so that those skilled in the art in the relevant technical field can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.
Claims
1. LED lamp bead production inspection system based on machine vision, including: Production detection module; its characteristics are: The production inspection module is responsible for receiving image data and performing inspection; the inspection includes position inspection, shape inspection, light source inspection and function inspection; Perform position detection: Use edge detection to obtain the characteristic points and corresponding coordinates of LED lamp beads, and determine the position of LED lamp beads in the image; replace the high-power lens to extract the edge contours of the LED lamp beads and the lamp bead body of the target object; use the intersection of two adjacent edge contours as the origin of the plane coordinate system, calculate the vector and match the preset standard vector interval; if the vector is not within the standard interval, generate an abnormal rejection instruction; if it is within the interval, perform subsequent processing; select the preset standard vector interval according to the specifications of the LED lamp beads, calculate the image size, use the calibration formula to convert pixels to actual size, and match the specifications in the database; Perform shape detection: install the industrial camera on a rotating slide rail and rotate it with the camera in conjunction with a fill light; the industrial camera takes images of the surface of the object and captures information at different angles; denoises, enhances and standardizes the image through ASIC; uses image segmentation algorithms to extract shadow parts and analyze air bubbles, cracks and pore defects; Light source detection: For LDE light source brightness detection: calculate the brightness by weighting the grayscale value of image pixels, convert it into a grayscale image and weight the different areas, compare the calculation result with the standard brightness value, if it does not match, generate a brightness abnormality instruction; otherwise, perform spectral measurement; For LDE light source color difference detection: use a spectral sensor to measure and calculate the light output intensity, measure the light source color temperature and compare it with the CIE standard light source color temperature curve, if it deviates from the standard, generate a color temperature abnormality instruction; otherwise, screen according to the color temperature range and perform further detection; Perform functional testing: measure the initial parameters under a standard environment and place the lamp beads in a temperature and humidity controlled environment box; regularly measure the brightness, color temperature, current, and voltage parameters under different working conditions, and calculate the percentage of change relative to the initial state; obtain the stability coefficient S by weighted calculation of each change rate, and compare it with the preset stability threshold. If S is less than the threshold, a functional abnormality instruction is generated.
2. The LED lamp bead production detection system based on machine vision according to claim 1 is characterized in that: It also includes a visual acquisition module and an executive learning module; The visual acquisition module captures the LED lamp bead image in real time through an industrial camera or a camera, and transmits the data to the production inspection module through the industrial Internet or a dedicated communication protocol; The execution learning module records abnormal events in each culling operation and uploads the test results and production environment data synchronously; Automatically adjust production parameters and optimize production processes based on real-time analysis; If equipment problems are found, adjust the workload of production equipment; optimize the production line and automatically adjust the process.
3. The LED lamp bead production detection system based on machine vision according to claim 1 is characterized in that: The specific process of generating the exception elimination instruction is as follows: Through edge detection, several LED lamp bead feature points are obtained, and each LED lamp bead feature point corresponds to an LED lamp bead; then, a plane coordinate system is made according to the frame in the image to obtain the coordinate points corresponding to several LED lamp bead feature points, that is, the positions of several LED lamp beads in the image detection area are obtained; the high-power lens of the industrial camera is replaced, and the contour obtained by edge detection is used to extract the characteristic information of the target object to obtain the edge contours of several LED lamp beads and LED lamp bead bodies; based on the edge contour of the LED lamp bead body, two adjacent edge contours are made into a plane coordinate system, and the two adjacent edge contours are vertically intersecting; the intersection point is the origin of the plane coordinate system, and the corresponding coordinate points of the edge contour of the corresponding LED lamp bead are obtained, and the coordinate points with the closest distance between the edge contour of the LED lamp bead body and the edge contour of the LED lamp bead are connected, and the vector between them is calculated; the preset position data in the database is retrieved, and the corresponding preset standard vector interval is selected according to the specifications of the detected LED lamp bead; The calculated vector is matched with the corresponding preset standard vector interval; if the calculated vector does not belong to the corresponding preset standard vector interval, a position anomaly elimination instruction is generated.
4. The LED lamp bead production detection system based on machine vision according to claim 3 is characterized in that: The specific process of matching the specifications in the database is as follows: Get the geometric shape of the LED lamp bead and the lamp body; convert the pixel unit into the actual size through the calibration formula; if the diameter d of the LED lamp bead and the diameter D of the lamp body are obtained, calculate through the formula: P d is the pixel length of the LED lamp bead; P t It is the pixel length corresponding to an object of known size in the image; The actual size is the physical unit obtained during calibration; the calculated dimensions d and D are then matched with the specifications in the preset position data in the database.
5. The LED lamp bead production detection system based on machine vision according to claim 4 is characterized in that: The specific process of shape detection is as follows: The industrial camera is installed on a rotating slide rail and rotates around the target area; a fill light is installed at a relatively reverse position of the industrial camera and is made to move relative to the rotating slide rail; the surface of the object is fully imaged from different angles and directions to capture subtle changes on the surface; a fixed brightness is selected as the initial brightness of the fill light; during the rotation of the camera, the industrial camera continuously captures images of the target surface and captures image information from different angles; the visual acquisition module uses ASIC to denoise, enhance and standardize the acquired images to obtain an analysis image; the image is analyzed in terms of the illumination direction, combined with the relative position of the camera and the fill light, and the shadow part is extracted through an image segmentation algorithm; the shadow part is extracted specifically by: dividing the image into multiple areas, highlighting the bubble, crack, and pore defect areas; based on the pixel features in the area, and comparing with the preset pixel features, it is determined whether the area is a bubble, crack or pore; if the pixel features of the area belong to the preset pixel features, the area is determined to be a bubble, crack or pore, and a shape detection removal instruction is generated; otherwise, if the pixel features of the area do not belong to the preset pixel features, the brightness of the fill light is increased, and the acquisition and detection are performed again until it passes.
6. The LED lamp bead production detection system based on machine vision according to claim 5 is characterized in that: The specific process of performing light source detection is as follows: The light output intensity of the LED light source is measured by a spectrum sensor; Output lumen value, P(λ i ) is the light intensity of the i-th wavelength, V(λ i ) is the luminosity function at this wavelength, Δλ is the change in wavelength, and K1 is the constant coefficient; the spectrometer is also used to measure the color temperature of the light source. Color temperature refers to the color deviation of the light source. Specifically, the color temperature approximation method is used to quickly calculate the color temperature by measuring a specific wavelength using a predetermined formula: R is the light intensity of the red channel, g is the light intensity of the green channel, and b is the light intensity of the blue channel; the calculated color temperature value needs to be compared with the light source color temperature curve of the CIE standard to determine whether the LED light source meets the predetermined color temperature range; if the color temperature result deviates from the set standard value, the LED light source may have color quality problems, and a color temperature abnormality removal instruction is generated; Otherwise, the LED lamp beads with the corresponding color temperature are screened according to the predetermined color temperature range.
7. The LED lamp bead production detection system based on machine vision according to claim 6 is characterized in that: The specific process of performing the functional detection is as follows: Analyze the stability of LED lamp beads, set up a standard environment to measure the initial state of LED lamp beads, and record the initial brightness, color temperature, current, and voltage parameters; place the LED lamp beads in a temperature and humidity controlled environment box and work for a certain period of time under different working conditions; During the test, the brightness, color temperature, and electrical characteristic parameters of the LED lamp beads are measured regularly, and the performance data at each time point is recorded; then the stability index is calculated; At each time point, calculate the percentage change of each performance parameter relative to the initial state: Lt is the brightness after t hours, Tt is the color temperature after t hours, It and Vt are the current and voltage after t hours; the brightness change rate, color temperature change rate, current change rate and voltage change rate are weighted to calculate the output LED stability coefficient S, and the LED stability coefficient S is compared with the preset stability threshold. If the LED stability coefficient S is less than the preset stability threshold, a function abnormality removal instruction is generated.
Citation Information
Cited By
Collaborative visual detection method and system for SMT production line
CN121304613A
Smt production line-oriented collaborative visual inspection method and system
CN121304613B
