LED lamp packaging detection method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of LED lamps, and provides an LED lamp packaging detection method and device, electronic equipment and a storage medium. Performing spatial pose adjustment and sorting of the LED detection targets according to the image of the detection area to obtain a detection sequence, performing spectrum scanning on the LED detection targets in the detection area according to the detection sequence to obtain three-dimensional data of the LED detection targets, and performing defect detection on the three-dimensional data to obtain defect distribution data of the LED detection targets. And performing reliability prediction according to the defect distribution data to obtain reliability prediction data, performing failure probability prediction according to the reliability prediction data to obtain packaging failure probability prediction data, and performing hierarchical decision according to a quality threshold and the packaging failure probability prediction data to generate a packaging quality inspection report. According to the invention, through attitude normalization, depth correlation detection, reliability-failure probability prediction and hierarchical decision, the accuracy and stability of LED packaging defect detection are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of LED lamp, in particular to a packaging detection method and device of LED lamp, electronic equipment and storage medium. BACKGROUND

[0002] As a core optoelectronic device widely used in lighting, display and signal indication scenes, the packaging quality of LED lamp directly determines its light efficiency, heat dissipation capacity, service life and electrical reliability. Microscopic defects such as bubbles, cracks, misplacement, pollution and poor welding may occur during the packaging process. Although these defects may not affect normal use at the initial stage, they may induce failure under long-term power-on or complex environment, leading to light decay, flicker, and even failure of the entire lamp. Therefore, in order to ensure product reliability and consistency, reduce after-sales maintenance cost and improve user experience.

[0003] The common LED lamp packaging detection methods at present mainly include manual visual inspection, optical microscopic imaging, X-ray perspective, infrared thermal imaging and electrical testing. Manual visual inspection relies on operator experience, is low in efficiency and strong in subjectivity; microscopic imaging has advantages in magnifying details, but has limited recognition ability for hidden structural defects; X-ray imaging can observe internal structure, but the device is expensive and has high radiation protection requirements; infrared thermal imaging is mainly used for thermal anomaly detection, but it is difficult to locate the specific type of packaging defects; and electrical testing can reflect the overall performance of the packaged LED, but cannot accurately locate potential physical defects. These methods are mostly static and post-hoc analysis, and cannot effectively predict the dynamic changes and future performance trends during the packaging process, which is insufficient in the intelligent manufacturing scene with fast production rhythm and high quality requirements. SUMMARY

[0004] Therefore, the present application provides a packaging detection method and device of LED lamp, electronic equipment and storage medium to solve the problem of deep correlation analysis between defects, performance and service life.

[0005] The first aspect of the present application provides a packaging detection method of LED lamp, which comprises: adjusting and sorting the spatial pose of LED detection target according to the collected image in the detection area to obtain a detection order; performing spectral scanning on each LED detection target in the detection area according to the detection order to obtain three-dimensional data of each LED detection target; performing defect detection on the three-dimensional data to obtain defect distribution data of each LED detection target, and performing reliability prediction processing according to the defect distribution data to obtain reliability prediction data; According to the reliability prediction data, failure probability prediction is performed to obtain package failure probability prediction data, and according to a preset quality threshold and the package failure probability prediction data, a grading decision is made to generate a package quality inspection report.

[0006] In an optional implementation, the space pose adjustment and sequencing of the LED detection targets according to the collected image in the detection area includes: In step S11, the image of the collected detection area is subjected to LED detection target number identification to obtain a detection target distribution map. In step S12, according to a preset initial vibration parameter set and the detection target distribution map, pose adjustment simulation processing is performed through a preset discrete element optimization model to obtain motion-distribution data of each LED detection target. In step S13, when the motion-distribution data is greater than a preset pose adjustment threshold, the initial vibration parameter set is updated according to a preset parameter optimization mode, and the step S12 is repeatedly executed according to the updated initial vibration parameter set until the motion-distribution data of all LED detection targets in the detection area is less than / equal to the pose adjustment threshold. In step S14, the LED detection targets in the detection area are subjected to pose adjustment through vibration according to the updated initial vibration parameter set, and the LED detection targets are subjected to number identification according to a preset number identification sequence to obtain a detection sequence in the detection area.

[0007] In an optional implementation, the spectral scanning of each LED detection target in the detection area according to the detection sequence to obtain three-dimensional data of each LED detection target includes: According to a preset light source detection parameter set and the detection sequence, each LED detection target is scanned one by one to obtain a spectral image set of each LED detection target. The spectral image set is subjected to multi-scale Retinex image enhancement processing to obtain an illumination correction image set of each LED detection target. According to a preset light source weight coefficient set, the illumination correction image set is subjected to image fusion processing to obtain three-dimensional data of each LED detection target.

[0008] In an optional implementation, the defect detection of the three-dimensional data to obtain defect distribution data of each LED detection target includes: The three-dimensional data is subjected to normalization and standardization processing to obtain a standard gray scale image. performing defect detection and positioning on the standard gray image through a preset FCOS detection model to obtain a bounding box coordinate set of defects in each LED detection target and defect type probability data; performing correlation analysis calibration according to a preset correlation loss algorithm, the bounding box coordinate set, and the defect type probability data to obtain a coordinate set of each type of defect and a corresponding defect type; performing type-positioning coupling processing on the coordinate set of each defect and the corresponding defect type to obtain defect distribution data.

[0009] In an optional implementation, the reliability prediction data includes a heat generation reliability prediction increment, a light attenuation reliability prediction coefficient, and a photoelectric conversion reliability prediction coefficient; and the performing reliability prediction processing according to the defect distribution data to obtain reliability prediction data includes: performing geometric feature extraction on the defect distribution data according to preset defect type data to obtain a feature parameter set of each type of defect in each LED detection target; performing reliability prediction processing on the feature parameter set through a preset electro-thermal coupling model to obtain a heat generation reliability prediction increment, a light attenuation reliability prediction coefficient, and a photoelectric conversion reliability prediction coefficient of each type of defect in each LED detection target.

[0010] In an optional implementation, the performing hierarchical decision making according to a preset quality threshold and the packaging failure probability prediction data to generate a packaging quality inspection report includes: performing grade evaluation on each LED detection target according to a preset quality threshold and the packaging failure probability prediction data to obtain a quality grade of each LED detection target; determining a sorting decision corresponding to each LED detection target in the detection area according to a preset grade sorting strategy and the quality grade; performing packaging processing on the quality grade, the sorting decision, and the coordinate set of each type of defect according to a preset data format to generate a packaging quality inspection report of each LED detection target.

[0011] The second aspect of the present application provides an LED lamp packaging detection device, the device comprising: a detection sequencing module configured to perform spatial pose adjustment and sequencing of LED detection targets according to images in a collected detection area to obtain a detection order; a spectrum scanning module configured to perform spectrum scanning on each LED detection target in the detection area according to the detection order to obtain three-dimensional data of each LED detection target; a reliable prediction module configured to perform defect detection on the three-dimensional data to obtain defect distribution data of each LED detection target, and perform reliability prediction processing according to the defect distribution data to obtain reliability prediction data; a hierarchical decision module configured to perform failure probability prediction according to the reliability prediction data to obtain package failure probability prediction data, and perform hierarchical decision according to a preset quality threshold and the package failure probability prediction data to generate a package quality inspection report.

[0012] The third aspect of the present application provides an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the package detection method of the LED lamp when executing the computer program.

[0013] The fourth aspect of the present application provides a computer readable storage medium, which stores a computer program, and the processor implements the steps of the package detection method of the LED lamp when executing the computer program.

[0014] In summary, the present application at least has the following beneficial technical effects: 1. The reliability is predicted based on the statistical model of defect distribution, which can identify potential failure risks in the early stage, realize the leap from "visible defects" to "predicted future failure", and greatly improve the credibility of the detection results.

[0015] 2. The hierarchical quality inspection report can be automatically generated by quantifying the output of package failure probability and combining the preset quality threshold for hierarchical decision, which provides a scientific basis for subsequent repair processes, product screening and batch release.

[0016] 3. The three-dimensional data of each LED target is obtained by high-resolution spectral scanning, and then combined with the fine defect detection algorithm to accurately locate various package defects such as size deviation, deformation, micro-cracks, etc. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0018] Figure 1 is a flowchart of a package detection method of an LED lamp provided by the embodiments of the present application; Figure 2is a functional module diagram of an LED lamp packaging detection device provided by an embodiment of the present application; Figure 3 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the scope of protection of the present application.

[0020] As shown in Figure 1 is a flowchart of an LED lamp packaging detection method provided by an embodiment of the present application. The LED lamp packaging detection method provided by the embodiment of the present application includes the following steps.

[0021] Step S1, adjusting and sorting the spatial pose of the LED detection target according to the collected image in the detection area to obtain a detection order.

[0022] After obtaining the image of the detection area, first, the collected image is subjected to LED detection target number identification. The embodiment of the present application uses a pre-trained OCR model to scan the batch number (such as "Chip2024-B5") engraved on the surface of the chip, and simultaneously locates the center coordinates and inclination angle of each chip through an edge detection algorithm. When a certain chip is not visible due to reverse buckling, back geometry feature matching (such as a heat dissipation substrate groove pattern) is enabled for auxiliary identification. All identification results are integrated into a detection target distribution map, that is, a rasterized electronic map marking each chip ID, position coordinates (x, y) and attitude angle θ. For example, the chip with ID B5 at coordinates (102, 58) is marked with an inclination angle of 15°, indicating that it is in an overturned state.

[0023] Subsequently, a discrete element optimization model is initiated based on the acquired distribution map of the detection targets to perform pose adjustment simulation. The model abstracts the LED chips as a combination of double-mass spheres (i.e., a bottom sphere with a diameter of 1.8 mm (simulating a thick substrate) and a top sphere with a diameter of 0.6 mm (simulating a light-emitting layer)), and collision responses are calculated through the Hertz-Mindlin contact theory. A preset initial vibration parameter set (e.g., an amplitude of 20 μm, a frequency of 200 Hz, and a track inclination of 2°) is loaded to simulate the motion trajectory of the chip group in the vibration field. The output motion-distribution data set of each LED detection target includes, but is not limited to, key indicators such as a spacing variation coefficient (i.e., a standard deviation / mean of the distance between adjacent chips) and an attitude angle offset (i.e., an angle between the chip normal and the vertical direction). For example, simulation shows that the final spacing variation of the B5 chip is 0.12, and the inclination angle is 8°, indicating uneven distribution and incomplete overturning.

[0024] When the motion-distribution data exceeds the pose adjustment threshold, a parameter optimization mechanism is triggered. In this embodiment of the present application, the pose adjustment threshold is set to a spacing variation coefficient ≤ 0.05 and an attitude angle offset ≤ 1°, which is based on the physical limitations of optical detection: when the chip spacing is uneven by more than 5%, multispectral imaging will overlap (e.g., two chips with a distance < 0.5 mm will cause the ultraviolet image to stick together), and when the inclination angle > 1°, the internal wire imaging distortion rate reaches 60%. The optimization uses a gradient descent algorithm: the track inclination is reduced by 0.1°, the friction coefficient is increased by 0.05, and the frequency is fine-tuned by 0.5 Hz. The updated parameter set (e.g., an inclination angle of 1.6°, a friction coefficient of 0.3, and a frequency of 223 Hz) drives the simulation again until all chip data meets the threshold. For example, after three iterations, the spacing variation of the B5 chip is 0.03, and the inclination angle is 0.8°, meeting the threshold requirements.

[0025] Finally, the physical vibration platform is driven according to the optimized parameters. The electromagnetic coil works in pulse mode at a frequency of 223 Hz, and the hopper twists at an amplitude of 15 μm, causing the chips to automatically overturn using the mass difference between the front and back surfaces (the substrate is heavier and tends to fall downward) in the process of jumping. At the same time, the pose is verified by a high-speed camera: when the B5 chip reaches the coordinates (205, 310) and the inclination angle is 0.8°, it is marked as the 7th detection target in the queue according to the preset numbering identification order (Z-shaped scanning according to the spatial position). The output detection order is an ordered array ["Chip2024-A1", "Chip2024-C3", …, "Chip2024-B5"], corresponding to the spatial arrangement order of the chips after vibration.

[0026] Step S2, according to the detection order, performing spectral scanning on each LED detection target in the detection area to obtain three-dimensional data of each LED detection target.

[0027] According to the above-mentioned detection sequence, the first step is to preset the light source detection parameter set according to the type of each LED detection target (such as chip or packaging material), which includes but is not limited to the wavelength, illumination intensity and exposure time configuration of the three light sources, to ensure that different defect types can obtain sufficient signal-to-noise ratio in their respective imaging channels. The preset light source detection parameter set usually includes a visible light band (450-650 nm) with 1000 lux, 20 μs illumination and collection settings to reveal chip surface colloidal bubbles and scratch marks; a near-infrared 850 nm laser with 50 mW, 50 μs pulse excitation to penetrate the packaging colloid and present the gold wire structure; and a 100 μs excitation time for the ultraviolet 365 nm light source to promote the fluorescent powder layer to emit a captureable fluorescent signal. The parameters of the above three channels need to be calibrated for different packaging materials and thicknesses to ensure that bubbles, gold wire offset and uneven fluorescent powder can be clearly distinguished in their respective channels. After the LED detection target is positioned in the imaging area and the position sensor is triggered, the three groups of light sources will be turned on in a predetermined sequence, each time only activating a single channel light source and simultaneously starting the corresponding high-speed camera to capture images. For example, when the exposure conditions calibrated in the light source detection parameter set are applied to the detection target numbered "Chip2024-B5", the visible light channel will obtain a high-contrast image of the surface scratch, the near-infrared channel will present a gold wire grid structure, and the ultraviolet channel will record uneven fluorescent patches.

[0028] After obtaining the spectral image set, the images of each channel need to be processed by multi-scale Retinex image enhancement to eliminate environmental light interference and improve the visibility of defect features. The Retinex algorithm is an imaging model that simulates the human eye's adaptive adjustment to different lighting environments. It extracts the reflection component by dividing the original image by the version after multiple Gaussian blurring, thereby suppressing background and scene illumination changes while preserving details. Specifically, each channel image will be applied with three different Gaussian filter radii (σ = 15, 80, 250 pixels), and then the ratio of the original pixel value to the corresponding blurred image will be used as the enhanced pixel. Finally, the weighted average of the multi-scale results is obtained to get the illumination corrected image set. It should be understood that the multi-scale strategy of the embodiments of the present application can not only amplify the gray difference of the edge of a small scratch (for example, the edge of a bubble with an original gray value of 30 is raised to more than 120), but also suppress overexposed areas (such as reducing the saturation value from the original 250 to about 200), significantly improving the signal-to-noise ratio of defect recognition in all channels. The signal-to-noise ratio of the gold wire offset can be increased from about 3:1 to about 8:1.

[0029] After obtaining the set of illumination corrected images, pixel-level multi-channel fusion is performed to generate an image cube representing the three-dimensional volumetric data. The fusion process linearly weights and superimposes the gray scale values of the three sets of corrected images at each pixel location (x, y) according to a set of light source weight coefficients calibrated in advance through a large number of experiments, where the visible light channel weight is 0.4 (mainly focusing on surface details), the near-infrared channel weight is 0.3 (highlighting internal structures), and the ultraviolet channel weight is 0.3 (enhancing fluorescent distribution). For example, at the coordinate (205, 310), the visible light pixel value is 180, the near-infrared value is 120, and the ultraviolet value is 90, then the synthesized gray scale value after fusion is 0.4 x 180 + 0.3 x 120 + 0.3 x 90 = 138. The synthesized gray scale value after fusion represents the three-dimensional volumetric data of the image cube at this coordinate. By performing the same operation on all pixels of the entire image, a tensor with a size of H x W x 3 is finally output, where the bottom layer corresponds to the visible light channel, the second layer corresponds to the near-infrared, and the top layer corresponds to the ultraviolet, which can be regarded as a three-layer transparent film combined to form a stereoscopic image. The three-dimensional volumetric data generated in this way not only retains the defect characteristics of each channel, but also improves the robustness and accuracy of the overall detection through the complementarity between different wavebands.

[0030] Step S3, performing defect detection on the three-dimensional volumetric data to obtain defect distribution data of each LED detection target, and performing reliability prediction processing according to the defect distribution data to obtain reliability prediction data.

[0031] When further extracting defect information from the multispectral image fused into three-dimensional volumetric data, first, the tensor obtained by superimposing different wavebands is normalized and standardized to obtain a unified standard gray scale image. Normalization is to linearly map the pixel value from the original [0, 255] to the [0, 1] interval, and standardization is to calculate the gray scale mean and standard deviation of the entire image. The Z-score algorithm is used in the embodiments of the present application to subtract the mean value from each pixel and divide it by the standard deviation, thereby correcting the brightness inconsistency caused by the differences in light source intensity and exposure time in the multispectral imaging process; under the geometric premise guaranteed by the previous pose adjustment threshold (extracted from the spatial arrangement optimization stage, requiring that the adjacent chip spacing fluctuation rate is ≤0.05 mm and the attitude angle offset is ≤1° to avoid imaging overlap and gold wire distortion), this standard gray scale operation can maximize the elimination of lighting interference and achieve high-fidelity presentation of bubble edges, gold wire structures, and fluorescent patches in subsequent network detection.

[0032] Subsequently, the obtained standard gray image is input into a pre-trained FCOS (Fully Convolutional One-Stage Object Detection) full convolutional detection model to extract defect features. The model generates a defect probability map at each pixel point through a classification branch without relying on preset anchor boxes, and outputs a corresponding set of bounding box coordinates from a positioning branch; when the classification confidence is high but the intersection over union (IoU) is low, the pre-embedded correlation loss algorithm, proxy squared error (PSE), will dynamically adjust the classification branch gradient, so that the classification probability and positioning accuracy tend to be consistent. For example, if the classification confidence in a certain bubble area is 0.88 and the IoU is only 0.4, the corrected probability after back propagation will be reduced to 0.75, thereby avoiding false detection with high confidence and low accuracy. With the help of this built-in calibration mechanism, the bounding box coordinate set of the defect in each LED detection target and the corresponding defect type probability data can be aligned in space. Further, the matrix containing defect type probability and bounding coordinate information is spliced into a multi-channel tensor along the depth direction, and then the type-position coupling algorithm is used to remove low-confidence and insufficient overlap candidate boxes and merge multiple responses of the same defect, so as to finally establish the accurate coordinate set of all bubble, gold wire offset and phosphor unevenness defects on each LED detection target and the corresponding type. The obtained defect distribution data not only labels the spatial position of each defect entity in the chip plane, but also records the confidence and bounding coordinates of each defect, and forms structured information in the form of "defect type-coordinate" description.

[0033] After obtaining comprehensive defect distribution data, reliability prediction is performed based on a pre-set electro-thermal coupling model. First, geometric feature parameter sets are extracted for each type of defect: the equivalent diameter and the Euclidean distance from the chip center for bubble defects, the inclination angle for gold wire offset defects, and the gray variance for phosphor unevenness defects. Subsequently, based on the energy conservation and carrier recombination mechanism of the packaging structure, the above feature parameters are mathematically mapped through three inverse equations of thermal resistance increment, third-order non-radiation coefficient and radiation coefficient. Among them, the thermal resistance increment (i.e., the reliability prediction increment of heat generation) quantifies the hindering effect of bubbles on the heat dissipation path, for example, when the bubble diameter is 0.2 mm and the center distance is 0.8 mm, the additional thermal resistance increment is calculated to be about 0.029 K / W; the third-order non-radiation coefficient (i.e., the light attenuation reliability prediction coefficient) reflects the Auger recombination loss caused by gold wire offset, and if the gold wire offset angle is 5°, the non-radiation coefficient is calculated to be about 0.0037; the radiation coefficient (i.e., the photoelectric conversion reliability prediction coefficient) represents the influence of phosphor unevenness on photon scattering and conversion efficiency, and when the gray variance is 0.35, the radiation coefficient is about 1.82.

[0034] Finally, the three reliability prediction data are integrated into an output tensor, in which the main data dimension stores the thermal resistance increment, non-radiation coefficient and radiation coefficient values, and the covariance matrix describes the positive and negative correlations between factors (for example, the thermal resistance increment and the non-radiation coefficient have a strong positive correlation, with a coefficient of about 0.7; the non-radiation coefficient and the radiation coefficient have a negative correlation, with a coefficient of about -0.5).

[0035] Step S4, failure probability prediction is performed according to the reliability prediction data to obtain packaging failure probability prediction data, and a grading decision is made according to a preset quality threshold and the packaging failure probability prediction data to generate a packaging quality inspection report.

[0036] After obtaining the reliability prediction data of each LED detection target, the physical quantitative indicators are first input into a failure probability prediction model to establish a mapping relationship from working conditions to failure probability. The model combines the Arrhenius accelerated failure theory and the current degradation law, and calculates the transition probability between states by real-time solving the junction temperature, electrical power and carrier recombination characteristics of each chip. The fourth-order Runge-Kutta numerical integration method is used to solve the Kolmogorov forward differential equation of Markov chain, thereby generating a time series curve of failure probability changing with time. The reason for modeling the light attenuation coefficient and the thermal resistance increment in parallel is that poor heat dissipation caused by bubble defects will accelerate the temperature rise of the PN junction, thereby amplifying the Auger recombination loss in a nonlinear manner, and the decline of the electroluminescent conversion efficiency caused by uneven distribution of phosphor will further change the junction temperature distribution, thereby affecting the subsequent current-temperature coupling behavior. After mapping the heat generation reliability prediction increment to the equivalent junction temperature change, the light attenuation reliability prediction coefficient to the output power decay curve with time, and the photoelectric conversion reliability prediction coefficient to the photoelectric conversion efficiency degradation model, the probability distribution of the LED detection target in the four states of no defect, single defect, compound defect and failure at any time point t can be calculated, thereby obtaining the packaging failure probability prediction data.

[0037] After each failure probability curve (i.e., the packaging failure probability prediction data of each LED detection target) is generated, it needs to be compared with the pre-calibrated quality threshold set to complete the quality grade evaluation. Among them, the quality threshold set is the key boundary value set according to the large-scale statistical data obtained from the industry accelerated aging experiment, such as the LM-80 standard, combined with the reliability requirements of the target application scene. With 20,000 hours as the node, the failure probability of 5% is defined as the boundary between A and B levels, and this threshold corresponds to the high quality requirement of 95% reliability within 20,000 hours under normal use conditions; 20% is defined as the boundary between B and C levels, which means that when this threshold is exceeded, the risk of more than 30% light decay within the next 50,000 hours exceeds 90%. The reason why the embodiment of the present application uses 20,000 hours as the evaluation milestone is that for general lighting and backlight applications, this time length is sufficient to reflect the thermal stability and light output decay characteristics of the device during medium and long-term use.

[0038] When comparing the packaging failure probability prediction data of each LED detection target with the quality threshold, the instantaneous failure probability F(20,000) at the 20,000 hour point can be obtained. If F(20,000) is less than or equal to 5%, the chip is rated as A level, which means that it has high reliability guarantee in medium and long-term use and can directly enter the mounting pipeline; if F(20,000) is between 5% and 20%, it is rated as B level, which triggers the repair process to perform special repair operations for the main defect types detected, such as enabling a vacuum glue injection machine to fill internal cavities for bubble defects, and restoring the wire network to the design position through laser fine tuning for gold wire offset; if F(20,000) is greater than 20%, it is identified as C level and needs to enter the disassembly recycling line to recover precious metals and reject unqualified components. The grade sorting strategy of the embodiment of the present application not only considers the failure probability itself, but also combines the application reliability requirements and repair economy to ensure that chips of the same quality level have consistent work specifications and incentive thresholds in the subsequent processing process.

[0039] After the grading and sorting decisions are completed, the core data of quality grade, sorting instruction, and defect coordinate set and type information of each LED detection target need to be packaged in the quality map format defined in the ISO / IEC 15434 standard. The standardized format is composed of a basic layer, a spatial layer, and a traceability layer. The basic layer records the quality grade and sorting decision code, for example, "B-REPAIR-BUBBLE-102-58" represents performing repair on the B-grade bubble defect at the coordinate (102, 58); the spatial layer embeds the bounding box coordinate set generated in the previous stage for precise positioning of the defect entity; and the traceability layer adds the chip unique ID and blockchain hash value to ensure the non-tamperability and full life cycle traceability of the quality data. When generating the packaged quality inspection report, the system will package the above three layers of information into a structured data record according to the embedded communication protocol, and store it through a distributed storage system, while driving the laser marking machine to engrave the corresponding two-dimensional code on the back of the chip. After scanning the two-dimensional code, the user or the post-production equipment can retrieve the complete set of data including the original multispectral image, the standard grayscale result, the FCOS detection heat map, the associated loss calibration record, and the Markov failure probability curve, etc., to realize intelligent quality management from point detection to full-process closed-loop traceability.

[0040] The application is applied to the technical field of LED lamp, and the spatial pose adjustment and sequencing of the LED detection target are performed according to the image of the detection area to obtain a detection order, the LED detection target in the detection area is scanned according to the detection order to obtain three-dimensional data of the LED detection target, the defect distribution data of the LED detection target are obtained by performing defect detection on the three-dimensional data, the reliability prediction data are obtained by performing reliability prediction according to the defect distribution data, the packaging failure probability prediction data are obtained by performing failure probability prediction according to the reliability prediction data, and the grading decision is generated according to the quality threshold and the packaging failure probability prediction data to generate a packaging quality inspection report. The application improves the accuracy and stability of LED packaging defect detection by performing spatial pose adjustment and sequencing, three-dimensional spectral scanning, defect distribution analysis, reliability prediction, and grading decision on the LED target in the detection area.

[0041] As shown in Figure 2 FIG. 1 is a functional module diagram of an LED lamp packaging detection device according to an embodiment of the application.

[0042] In some embodiments, the LED lamp packaging detection device 2 can include a plurality of functional modules composed of computer program segments. The computer programs of each program segment in the LED lamp packaging detection device 2 can be stored in the memory of the server and executed by at least one processor to perform the functions of the LED lamp packaging detection method (see Figure 1 Description).

[0043] In the embodiment, the packaging detection device 2 of the LED lamp can be divided into multiple functional modules according to the functions performed by the packaging detection device 2. The functional modules can include a detection sequencing module 21, a spectrum scanning module 22, a reliable prediction module 23, and a hierarchical decision module 24. The module referred to in the present application refers to a series of computer program segments capable of being executed by at least one processor and capable of completing a fixed function, which are stored in a memory. In the embodiment, the functions of the modules will be described in detail in subsequent embodiments.

[0044] The detection sequencing module 21 is configured to adjust and sequence the spatial poses of the LED detection targets according to the images collected in the detection area.

[0045] In an optional implementation, the detection sequencing module 21 is specifically configured to: In step S11, the images of the collected detection area are subjected to LED detection target number identification to obtain a detection target distribution map. In step S12, a preset discrete element optimization model is used to perform pose adjustment simulation processing according to a preset initial vibration parameter set and the detection target distribution map to obtain motion-distribution data of each LED detection target. In step S13, when the motion-distribution data is greater than a preset pose adjustment threshold, the initial vibration parameter set is updated according to a preset parameter optimization mode, and the step S12 is repeatedly performed according to the updated initial vibration parameter set until the motion-distribution data of all the LED detection targets in the detection area is less than / equal to the pose adjustment threshold. In step S14, the LED detection targets in the detection area are subjected to pose adjustment by vibration according to the updated initial vibration parameter set, and the LED detection targets are subjected to number identification according to a preset number identification sequence to obtain a detection sequence in the detection area.

[0046] The spectrum scanning module 22 is configured to perform spectrum scanning on each LED detection target in the detection area according to the detection sequence to obtain three-dimensional data of each LED detection target.

[0047] In an optional implementation, the spectrum scanning module 22 is specifically configured to: Each LED detection target is scanned one by one according to a preset light source detection parameter set and the detection sequence to obtain a spectrum image set of each LED detection target. The spectrum image set is subjected to multi-scale Retinex image enhancement processing to obtain an illumination correction image set of each LED detection target. perform image fusion processing on the set of illumination correction images according to the set of preset light source weight coefficients to obtain three-dimensional data of each LED detection target.

[0048] The reliable prediction module 23 is configured to perform defect detection on the three-dimensional data to obtain defect distribution data of each LED detection target, and perform reliability prediction processing according to the defect distribution data to obtain reliability prediction data.

[0049] In an optional implementation, the reliable prediction module 23 is configured to: perform normalization and standardization processing on the three-dimensional data to obtain a standard gray-scale image; perform defect detection and positioning on the standard gray-scale image by using a preset FCOS detection model to obtain a set of bounding box coordinates of defects in each LED detection target and defect type probability data; perform correlation analysis calibration according to a preset correlation loss algorithm, the set of bounding box coordinates, and the defect type probability data to obtain a set of coordinates of defects of various types and corresponding defect types; perform type-positioning coupling processing on the set of coordinates of defects of various types and corresponding defect types to obtain defect distribution data.

[0050] In an optional implementation, the reliable prediction module 23 is further configured to: perform geometric feature extraction on the defect distribution data according to preset defect type data to obtain a set of feature parameters of defects of various types in each LED detection target; perform reliability prediction processing on the set of feature parameters by using a preset electro-thermal coupling model to obtain a heat generation reliability prediction increment, a light attenuation reliability prediction coefficient, and a photoelectric conversion reliability prediction coefficient of defects of various types in each LED detection target.

[0051] The hierarchical decision module 24 is configured to perform failure probability prediction according to the reliability prediction data to obtain packaging failure probability prediction data, and perform hierarchical decision making according to a preset quality threshold and the packaging failure probability prediction data to generate a packaging quality inspection report.

[0052] In an optional implementation, the hierarchical decision module 24 is specifically configured to: perform grade evaluation on each LED detection target according to a preset quality threshold and the packaging failure probability prediction data to obtain a quality grade of each LED detection target; determine a sorting decision corresponding to each LED detection target in the detection area according to a preset grade sorting strategy and the quality grade; According to the preset data format, the quality grade, the sorting decision and the coordinate set of each type of defect are encapsulated to generate an encapsulation quality inspection report of each LED detection target.

[0053] It should be understood that the various changes and specific embodiments of the method provided by the above embodiments are also applicable to the encapsulation detection device of the LED lamp of the present embodiment. Through the foregoing detailed description of the encapsulation detection method of the LED lamp, those skilled in the art can clearly understand the implementation method of the encapsulation detection device of the LED lamp in the present embodiment. For the sake of brevity of the description, the detailed description is not repeated here.

[0054] As shown in Figure 3 FIG. 1 is a structural schematic diagram of an electronic device according to an embodiment of the present application.

[0055] In the preferred embodiment of the present application, the electronic device 3 can include, but is not limited to, a memory 31, at least one processor 32 and at least one communication bus 33.

[0056] Those skilled in the art should understand that Figure 3 The structure of the electronic device 3 shown is not a limitation of the embodiments of the present application. The electronic device 3 can also include more or less other hardware or software, or different component arrangements.

[0057] In some embodiments, the electronic device 3 is a device capable of automatically performing numerical calculation and / or information processing according to pre-set or stored instructions. The hardware thereof includes, but is not limited to, a microprocessor, an application specific integrated circuit, a programmable gate array, a digital processor and an embedded device, etc.

[0058] It should be noted that the electronic device 3 is only an example. Other existing or future electronic products, such as those that can be adapted to the present application, should also be included within the scope of protection of the present application and are hereby incorporated by reference.

[0059] In some embodiments, the memory 31 stores a computer program which, when executed by the at least one processor 32, implements all or part of the steps of the method for detecting the package of LED lamp as described. The memory 31 includes a Read-Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), a One-time Programmable Read-Only Memory (OTPROM), an Electrically-Erasable Programmable Read-Only Memory (EEPROM), a Compact Disc Read-Only Memory (CD-ROM) or other optical disk storage, magnetic disk storage, magnetic tape storage, or any other medium of computer readable nature which can serve to carry or store data. Further, the computer readable storage medium can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs required by at least one function, etc.

[0060] In some embodiments, the at least one processor 32 is a Control Unit of the electronic device 3, which connects various components of the entire electronic device 3 through various interfaces and lines, and performs various functions of the electronic device 3 and processes data by running or executing programs or modules stored in the memory 31 and calling data stored in the memory 31. For example, the at least one processor 32 implements all or part of the steps of the method for detecting the package of LED lamp as described in the embodiments of the present application when executing the computer program stored in the memory 31, or implements all or part of the functions of the device for detecting the package of LED lamp. The at least one processor 32 can be composed of integrated circuits, for example, can be composed of a single packaged integrated circuit, or can be composed of multiple packaged integrated circuits with the same function or different functions, including one or more combinations of Central Processing Units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips, etc.

[0061] In some embodiments, the at least one communication bus 33 is configured to enable connection communication between the memory 31 and the at least one processor 32. Although not shown, the electronic device 3 can further include a power supply (such as a battery) for powering the various components of the electronic device 3. Preferably, the power supply is logically connected to the at least one processor 32 via a power management device, thereby enabling management of charging, discharging, and power consumption management, etc. by the power management device. The power supply can also include one or more direct current or alternating current power sources, recharging circuits, power failure detection circuitry, power converters or inverters, power status indicators, and the like. The electronic device 3 can further include a variety of sensors, a Bluetooth module, a Wi-Fi module, and the like, which are not described herein.

[0062] The integrated units in the form of software function modules described above can be stored in a computer readable storage medium. The software function modules described above are stored in a storage medium, and include a plurality of instructions for causing an electronic device (which can be a personal computer, an electronic device, or a network device, etc.) or a processor to execute part of the method described in various embodiments of the present application.

[0063] In several embodiments provided in the present application, it should be understood that the disclosed apparatus and method can be implemented in other manners. For example, the described apparatus embodiments are merely illustrative. For example, the division of the modules is merely a logical function division. There can be another division manner in actual implementation.

[0064] The modules illustrated as separated components can or can not be physically separated, and the components illustrated as modules can or can not be physical units. They can be located in one position or distributed on a plurality of network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments.

[0065] The above are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Therefore, any equivalent changes made on the structure, shape, and principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method of detecting the package of an LED lamp, characterized by, The method comprises: According to the image collected in the detection area, the spatial pose adjustment and sorting of the LED detection target are carried out to obtain the detection sequence; According to the detection sequence, the spectral scanning of each LED detection target in the detection area is carried out to obtain the three-dimensional data of each LED detection target; The three-dimensional data is subjected to defect detection to obtain the defect distribution data of each LED detection target, and the reliability prediction processing is carried out according to the defect distribution data to obtain the reliability prediction data; According to the reliability prediction data, the failure probability prediction is carried out to obtain the packaging failure probability prediction data, and the grading decision is carried out according to the preset quality threshold and the packaging failure probability prediction data to generate the packaging quality inspection report.

2. The LED lamp packaging detection method according to claim 1, wherein, According to the image collected in the detection area, the spatial pose adjustment and sorting of the LED detection target are carried out to obtain the detection sequence, which comprises: Step S11, the image of the collected detection area is subjected to LED detection target number identification to obtain a detection target distribution map; Step S12, through a preset discrete element optimization model, the motion-distribution data of each LED detection target is obtained by carrying out pose adjustment simulation processing according to the preset initial vibration parameter set and the detection target distribution map; Step S13, when the motion-distribution data is greater than a preset pose adjustment threshold, the initial vibration parameter set is updated according to a preset parameter optimization mode, and the step S12 is repeatedly executed according to the updated initial vibration parameter set until the motion-distribution data of all LED detection targets in the detection area is less than / equal to the pose adjustment threshold; Step S14, the LED detection targets in the detection area are subjected to pose adjustment by vibration mode according to the updated initial vibration parameter set, and the LED detection targets are subjected to number identification according to a preset number identification sequence to obtain the detection sequence in the detection area.

3. The LED lamp packaging detection method according to claim 1, wherein, According to the detection sequence, the spectral scanning of each LED detection target in the detection area is carried out to obtain the three-dimensional data of each LED detection target, which comprises: According to a preset light source detection parameter set and the detection sequence, each LED detection target is scanned one by one to obtain a spectral image set of each LED detection target; The spectral image set is subjected to multi-scale Retinex image enhancement processing to obtain an illumination correction image set of each LED detection target; According to a preset light source weight coefficient set, the illumination correction image set is subjected to image fusion processing to obtain the three-dimensional data of each LED detection target.

4. The LED lamp packaging detection method according to claim 1, wherein, The three-dimensional data is subjected to defect detection to obtain the defect distribution data of each LED detection target, which comprises: The three-dimensional data is subjected to normalization and standardization processing to obtain a standard gray scale image; Through a preset FCOS detection model, the standard gray scale image is subjected to defect detection and positioning to obtain a bounding box coordinate set of defects in each LED detection target and defect type probability data; According to the preset correlation loss algorithm, the boundary box coordinate set, and the defect type probability data, correlation analysis calibration is performed to obtain a coordinate set of each type of defect and a corresponding defect type; The coordinate set of each defect and the corresponding defect type are subjected to type-position coupling processing to obtain defect distribution data.

5. The method of claim 1, wherein, The reliability prediction data includes a heat generation reliability prediction increment, a light attenuation reliability prediction coefficient, and a photoelectric conversion reliability prediction coefficient; and the reliability prediction processing according to the defect distribution data to obtain the reliability prediction data includes: According to the preset defect type data, geometric feature extraction is performed on the defect distribution data to obtain a feature parameter set of each type of defect in each LED detection target; According to the preset electro-thermal coupling model, reliability prediction processing is performed on the feature parameter set to obtain a heat generation reliability prediction increment, a light attenuation reliability prediction coefficient, and a photoelectric conversion reliability prediction coefficient of each type of defect in each LED detection target.

6. The LED lamp packaging detection method according to claim 4, wherein, The hierarchical decision making according to the preset quality threshold and the packaging failure probability prediction data to generate a packaging quality inspection report includes: According to the preset quality threshold and the packaging failure probability prediction data, a grade assessment is performed on each LED detection target to obtain a quality grade of each LED detection target; According to the preset grade sorting strategy and the quality grade, a sorting decision corresponding to each LED detection target in the detection area is determined; According to the preset data format, packaging processing is performed on the quality grade, the sorting decision, and the coordinate set of each type of defect to generate a packaging quality inspection report of each LED detection target.

7. An LED lamp package detection device applied to the LED lamp package detection method of claim 1, characterized in that, The device includes: A detection ordering module configured to perform spatial pose adjustment and ordering of LED detection targets according to images in a detection area collected by the image acquisition module to obtain a detection order; A spectrum scanning module configured to perform spectrum scanning on each LED detection target in the detection area according to the detection order to obtain three-dimensional data of each LED detection target; A reliability prediction module configured to perform defect detection on the three-dimensional data to obtain defect distribution data of each LED detection target, and to perform reliability prediction processing according to the defect distribution data to obtain reliability prediction data; A hierarchical decision making module configured to perform failure probability prediction according to the reliability prediction data to obtain packaging failure probability prediction data, and to perform hierarchical decision making according to the preset quality threshold and the packaging failure probability prediction data to generate a packaging quality inspection report.

8. An electronic device, comprising: The electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the steps of the packaging detection method of the LED lamp according to any one of claims 1 to 6 when executing the computer program.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the packaging detection method of the LED lamp according to any one of claims 1 to 6.

Images