An intelligent detection method and system for LED internal welding defects

Abstract

The present invention provides an intelligent detection method and system for internal welding defects of LEDs, which relates to the field of data processing technology. The method and system perform image sampling of a sample set by configuring X-ray sampling control parameters with a sample parameter set to generate a sample image set, perform grayscale conversion of the sample image set and grayscale authentication of adjacent pixels to generate authentication attention results, perform adaptive enhancement of the sample image set and defect recognition results, and complete batch sample detection based on the defect recognition results. The method solves the technical problem that the existing technology for detecting internal welding defects of LEDs is high in cost, and the tolerance of internal welding defects of LEDs is usually based on the quality control rate, resulting in difficulty in balancing the quality and production costs of LED products. The method achieves the technical effect of quickly and efficiently extracting the characteristics of internal welding defects of LEDs in the same batch and comparing and removing LEDs with welding defects in the same batch based on the defect characteristics, thereby improving the quality control of mass-produced LEDs.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an intelligent detection method and system for LED internal welding defects. Background Art

[0002] In LED production, internal soldering quality is directly related to product performance and lifespan. However, due to the high cost of detecting internal soldering defects in LEDs using existing technologies, manufacturers are forced to set an economically acceptable defect tolerance threshold based on quality control rates. While this approach may appear to reduce production costs in the short term, it actually introduces a potential trade-off between quality and cost.

[0003] In summary, the existing technology for detecting internal soldering defects in LEDs is costly and usually tolerates internal soldering defects based on the quality control rate, leading to technical problems such as difficulty in balancing the quality and production costs of LED products. Summary of the Invention

[0004] The present application provides an intelligent detection method and system for LED internal welding defects, which is used to solve the technical problem that the existing technology has high costs for LED internal welding defect detection and usually tolerates LED internal welding defects based on the quality control rate, resulting in difficulty in balancing the quality and production costs of LED products.

[0005] In view of the above problems, the present application provides an intelligent detection method and system for internal welding defects of LEDs.

[0006] The first aspect of the present application provides an intelligent detection method for internal welding defects of LEDs, the method comprising: establishing a batch sampling ratio, and when any batch of samples performs defect detection, performing random sample sampling according to the batch sampling ratio to establish a sample set; establishing a parameter set of the sample, the parameter set including a structural parameter set and a size parameter set, configuring X-ray sampling control parameters with the parameter set, the sampling control parameters including exposure time and ray energy, and configuring enhancement parameters according to the sampling control parameter configuration and sample data; performing image sampling of the sample set by X-rays, and performing batch processing of the image sampling results with the enhancement parameters to generate a sample image set; performing grayscale conversion of the sample image set, and performing grayscale authentication of adjacent pixels of the grayscale conversion result based on the position point to generate an authentication focus result, performing adaptive enhancement of the sample image set with the authentication focus result, and inputting the adaptive enhancement result into a defect detection network to generate a defect recognition result, and completing intelligent detection of batch samples according to the defect recognition result.

[0007] According to a second aspect of the present application, an intelligent detection system for internal welding defects of LEDs is provided, the system comprising: a sample set establishment unit, for establishing a batch sampling ratio, and when defect detection is performed on any batch of samples, random sampling is performed according to the batch sampling ratio to establish a sample set; a parameter set establishment unit, for establishing a parameter set of the sample, the parameter set comprising a structural parameter set and a size parameter set, and the sampling control parameters of the X-ray are configured with the parameter set, the sampling control parameters comprising exposure time and ray energy, and enhancement parameters are configured according to the sampling control parameter configuration and the sample data; an image sampling execution unit, for performing image sampling of the sample set by X-rays, and performing batch processing of the image sampling results with the enhancement parameters to generate a sample image set; a grayscale conversion execution unit, for performing grayscale conversion of the sample image set, and performing grayscale authentication of the grayscale conversion results based on the adjacent pixel grayscale of the position point to generate an authentication focus result, and performing adaptive enhancement of the sample image set with the authentication focus result, and inputting the adaptive enhancement result into the defect detection network to generate a defect recognition result, and completing intelligent detection of the batch samples according to the defect recognition result.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] By establishing a batch sampling ratio, when any batch of samples is subjected to defect detection, random sampling is performed according to the batch sampling ratio to establish a sample set; a sample parameter set is established, the parameter set including a structural parameter set and a size parameter set, and the parameter set is used to configure the X-ray sampling control parameters, the sampling control parameters including exposure time and ray energy, and enhancement parameters are configured according to the sampling control parameter configuration and sample data; image sampling of the sample set is performed by X-ray, and batch processing of the image sampling results is performed with the enhancement parameters to generate a sample image set; grayscale conversion of the sample image set is performed, and the grayscale conversion results are authenticated based on the grayscale of adjacent pixels at the position point to generate an authentication focus result, and the sample image set is adaptively enhanced with the authentication focus result, and the adaptive enhancement result is input into the defect detection network to generate a defect recognition result, and the intelligent detection of the batch samples is completed based on the defect recognition result. The existing technology for detecting internal soldering defects in LEDs is relatively costly, and tolerance for internal soldering defects in LEDs is usually based on the quality control rate, resulting in a technical problem of difficulty in balancing the quality and production cost of LED products. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flow chart of an intelligent detection method for internal soldering defects of LEDs provided in this application;

[0011] Figure 2A schematic diagram of a process for generating certification focus results in an intelligent detection method for internal soldering defects of LEDs provided in this application;

[0012] Figure 3 This is a structural diagram of an intelligent detection system for internal welding defects of LEDs provided in this application.

[0013] Explanation of the reference numerals: sample set establishing unit 1, parameter set establishing unit 2, image sampling execution unit 3, grayscale conversion execution unit 4. DETAILED DESCRIPTION

[0014] This application provides an intelligent detection method and system for internal soldering defects in LEDs. This method addresses the high cost of existing LED internal soldering defect detection techniques, which typically tolerate internal soldering defects based on quality control rates, leading to a difficult balance between LED product quality and production costs. The system achieves the technical effect of quickly and efficiently extracting the characteristics of internal soldering defects in LEDs from the same batch and, based on these characteristics, comparing and removing soldering defective LEDs from the same batch, thereby improving the quality control of mass-produced LEDs.

[0015] The acquisition, storage, use, and processing of data in the technical solution of the present invention comply with relevant regulations.

[0016] Below, the technical solutions of the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments described herein. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention, rather than all, are shown in the accompanying drawings.

[0017] Example 1

[0018] like Figure 1 As shown, the present application provides an intelligent detection method for internal soldering defects of LEDs, the method comprising:

[0019] A100: Establishing a batch sampling ratio. When performing defect detection on any batch of samples, random sampling is performed according to the batch sampling ratio to establish a sample set.

[0020] Specifically, it should be understood that in quality management, in order to test the quality level of an entire batch of products, a portion of the products can be randomly selected as samples for testing, so that the quality of the entire batch of products can be evaluated through samples without having to inspect each product.

[0021] The objects of sampling inspection in this embodiment are LEDs (light-emitting diodes) produced in batches, and the specific inspection direction is to detect whether there are defects in the internal welding process or connection points of the LED components. The batch sampling ratio for LEDs depends on the batch size, quality requirements and cost factors. This embodiment does not limit the specific numerical setting of the batch sampling ratio, which is customized according to the manufacturer's considerations based on batch size, quality requirements and cost factors.

[0022] This embodiment interactively obtains the batch sampling ratio predefined by the manufacturer, and then performs random sampling of a plurality of LEDs in the same batch according to the batch sampling ratio to establish the sample set.

[0023] A200: Establishing a parameter set for a sample, the parameter set including a structural parameter set and a size parameter set, configuring X-ray sampling control parameters using the parameter set, the sampling control parameters including exposure time and ray energy, and configuring enhancement parameters based on the sampling control parameter configuration and sample data;

[0024] Specifically, it should be understood that since the detection target of this embodiment is to detect whether there are defects in the welding process or connection points inside the LED component, this embodiment selects X-ray detection for non-destructive detection inside the component as the detection method.

[0025] Specifically, X-ray testing is a non-destructive testing method that can obtain internal welding images of LED devices without damaging them, so as to detect and analyze the internal welding conditions of LEDs.

[0026] In this embodiment, the X-ray parameter configuration method for improving the resolution of the X-ray imaging obtained by subsequent X-ray-based sample set image sampling is as follows:

[0027] Interactively obtain multiple sets of historical sampling control data used by historical manufacturers to perform high-resolution X-ray internal imaging of single LEDs, where each set of historical sampling control data includes sample LED structural parameters (LED lamp bead type, packaging structure, internal chip arrangement), sample LED size parameters (physical size parameters include LED outer dimensions, lamp bead size), and sample sampling control parameters (X-ray exposure time and ray energy). Based on the knowledge graph, the associated storage of multiple sets of historical sampling control data is performed to obtain an X-ray control parameter library.

[0028] Interactively obtain the sample LED structure design information (LED lamp bead type, packaging structure, internal chip arrangement) of the sample to establish the structural parameter set, interactively obtain the LED size design information of the sample set (physical size parameters include the external size of the LED and the lamp bead size), establish the size parameter set, and constitute the parameter set of the sample.

[0029] The parameter set is used as a data screening criterion to traverse the X-ray control parameter library to obtain a group of historical sampling control data whose data structure is consistent with the parameter set, and the sample sampling control parameters of this group of historical sampling control data are used as the sampling control parameter configuration of the X-ray equipment currently participating in non-destructive testing.

[0030] Furthermore, it should be understood that when multiple samples need to be inspected, appropriate adjustment and enhancement of X-ray imaging parameters can improve the clarity and resolution of subsequent sample images, thereby more accurately observing and analyzing the details and internal structure of the samples.

[0031] This embodiment interactively obtains multiple groups of samples, including sample quantity, sample X-ray sampling control parameters, and sample X-ray enhancement parameters, when the manufacturer historically collects multiple sample LED images, and then constructs an enhancement parameter analysis model based on a back-propagation neural network. The input of the enhancement parameter analysis model is the sample quantity and the sampling control parameters of the single LED sampling, and the output result is the enhancement parameters for achieving the high-resolution image sampling of multiple LEDs with the sample quantity.

[0032] Based on the existing conventional back-propagation neural network training method, multiple groups of sample quantity-sample X-ray sampling control parameters-sample X-ray enhancement parameters are used as training data to train the enhanced parameter analysis model until the model output accuracy meets the preset requirements.

[0033] The sampling control parameter configuration obtained above and the number of samples in the current sample set to be detected are input into the enhanced parameter analysis model as input data, and the enhanced parameters are obtained based on model analysis.

[0034] A300: performing image sampling of the sample set by X-rays, and performing batch processing of the image sampling results using the enhancement parameters to generate a sample image set;

[0035] Specifically, in this embodiment, after adjusting the parameters of the X-ray imaging device using the sampling control parameter configuration, the sample set is moved to a specific area of the X-ray imaging device for performing X-ray irradiation and imaging acquisition.

[0036] Start the X-ray imaging device to emit X-rays to perform image sampling of the sample set, obtain the image sampling results, the image sampling results include multiple sample LED internal images of multiple samples in the sample set, perform batch processing of the multiple sample LED internal images in the image sampling results with the enhancement parameters, and generate the sample image set.

[0037] A400: Perform grayscale conversion of the sample image set, and perform grayscale authentication of the grayscale conversion results based on the adjacent pixels of the position points to generate authentication attention results, perform adaptive enhancement of the sample image set with the authentication attention results, and input the adaptive enhancement results into the defect detection network to generate defect recognition results, and complete intelligent detection of batch samples based on the defect recognition results.

[0038] In one embodiment, Figure 2 As shown, the method step A400 of generating the authentication attention result provided by this application further includes:

[0039] A410: Obtain the welding process corresponding to the sample set and establish a welding interval;

[0040] A420: Using the welding interval as a contour recognition space, performing edge contour search on the grayscale conversion result, and establishing a proofreading contour space based on the search results;

[0041] A430: Performing grayscale authentication of adjacent pixels in the proofreading contour space to generate a first authentication focus result;

[0042] A440: generating a second authentication focus result based on the contour edge corresponding to the proofread contour space;

[0043] A450: Establish an authentication focus result based on the first authentication focus result and the second authentication focus result.

[0044] In one embodiment, the grayscale authentication of adjacent pixels is performed in the proofreading contour space to generate a first authentication result. Step A430 of the method provided in this application further includes:

[0045] A431: Identifying the number of spatial pixels in the proofreading contour space, and configuring a local verification window according to the identification result;

[0046] A432: performing grayscale anomaly check of adjacent pixels in the proofreading contour space using the local verification window;

[0047] A433: Complete grayscale authentication based on the verification results to generate the first authentication focus result.

[0048] In one embodiment, the grayscale anomaly check of adjacent pixels in the proofreading contour space using the local verification window is performed, and step A432 of the method provided by the present application further includes:

[0049] A432-1: placing the local verification window in the proofreading contour space, and extracting the grayscale values of pixels within the window through the local verification window;

[0050] A432-2: Perform extreme value removal on the grayscale values of the pixels at the position, calculate a grayscale mean based on the extreme value removal result, and verify and update the grayscale value of the pixel at the center position using the grayscale mean;

[0051] A432-3: Perform grayscale value calibration of pixels around the center based on the verification update results to complete the grayscale anomaly verification of adjacent pixels.

[0052] In one embodiment, the adaptive enhancement of the sample image set based on the authentication focus result, the method step A400 provided in the present application further includes:

[0053] A460: performing pixel grayscale sampling of a sample image set based on the authentication focus result, and establishing a focus pixel grayscale set and an associated pixel grayscale set;

[0054] A470: Input the grayscale set of the focus pixel and the grayscale set of the associated pixels into the adaptive enhancement processing model, generate enhancement parameters, and complete the adaptive enhancement with the enhancement parameters.

[0055] Specifically, it should be understood that the welding process of the same batch of LEDs is the same and the positions of the welding areas are consistent. Based on this, the interactive production party of this embodiment directly obtains the welding process adopted by the standard LED of the sample set, and then determines the specific area or range that needs to be welded during the internal welding of the LED by the welding process, and obtains the welding interval.

[0056] The color sample image set is converted into a grayscale image by performing weighted averaging on the red, green, and blue channels of the RGB image, and the grayscale conversion result including multiple grayscale converted sample images is obtained. The purpose of grayscale conversion is to ensure that the grayscale image in the welding area in the sample image can be accurately selected.

[0057] The welding interval is used as a contour recognition space, and the contour recognition space refers to a curve contour connecting continuous points with the same color or intensity. The contour recognition space is used as a matching reference, and a shape matching algorithm is used to perform edge contour search (contour matching) on multiple grayscale conversion sample images of the grayscale conversion results one by one to obtain search results, which are actual contours in the multiple grayscale conversion sample images that are similar to the contour recognition space.

[0058] It should be understood that the reason why this embodiment uses the shape matching algorithm to identify similar contours is that there is often a deviation between the actual welding spot size and the weld spot size in the welding interval. The weld spot contour may be larger or smaller than the weld spot contour in the welding interval, but the weld spot shape is similar.

[0059] In this embodiment, multiple actual contours of the multiple grayscale conversion sample images in the search results obtained by contour matching are used as multiple proofreading contour spaces of the multiple grayscale conversion sample images.

[0060] Based on the consistency of the method for identifying welding defects for each sample, this embodiment takes the defect identification based on the grayscale conversion sample image of an unspecified sample as an example to elaborate on the technical solution in detail.

[0061] Furthermore, this embodiment identifies the number of pixel points in the proofreading contour space of a random grayscale conversion sample image, obtains the identification result representing the number of pixel points, and then configures the division size constraint when dividing the relative contour space into several local images according to the number of pixels in the proofreading contour space.

[0062] An LBP feature calculation model is pre-constructed, and the partition size constraint is used as a model adjustment parameter of the LBP feature calculation model to perform parameter adjustment configuration of the LBP feature calculation model to obtain the local verification window.

[0063] The proofreading contour space is gridded based on the division size constraint, and the local verification window is placed in multiple grid local neighborhoods within the proofreading contour space. The grayscale values of the central pixel and the surrounding pixels are compared in multiple grid local neighborhoods through the LBP feature calculation model of the local verification window, and multiple binary numerical values are generated according to the comparison results as the grayscale values of multiple position pixels in the multiple grid local neighborhoods.

[0064] Maximum and minimum value removal is performed on the grayscale values of multiple pixels at the positions, and the grayscale mean is calculated based on the extreme value removal result. The grayscale value of the pixel at the center position of the grayscale conversion sample image is verified and updated using the grayscale mean. The specific verification and update method is to calculate the grayscale difference between the grayscale mean and the grayscale value of the pixel at the center position of the grayscale conversion sample image. If the grayscale difference is less than the preset update grayscale deviation, the grayscale value of the pixel at the center position is retained. Otherwise, the grayscale mean is used to replace the grayscale value of the pixel at the center position.

[0065] It should be understood that when the welding is qualified, the solder joint produced by LED welding is a smooth curved surface, and the grayscale value deviation of each pixel point in the grayscale image of the corresponding solder joint is small.

[0066] Based on this, this embodiment presets a grayscale deviation threshold, and traverses the grayscale value calibration of pixels around the comparison center (all pixel points in the calibration contour space) according to the similarity grayscale value of the verification update result, marks the pixel points whose grayscale value deviation is greater than the preset grayscale deviation threshold as abnormal pixel points, completes the abnormal verification of the grayscale of adjacent pixels, and identifies the multiple abnormal pixel points obtained by the abnormal verification as the first authentication focus result of the grayscale conversion sample image. The first authentication focus result is an image with a high probability of welding defects in the weld shape area of the weld, which serves as the key focus area for subsequent weld shape area defect detection.

[0067] In addition to identifying suspected defects in the weld shape area of the solder joint, this embodiment also identifies suspected defects in the weld contour of the solder joint (the boundary line where the solder contacts the inner surface of the LED).

[0068] In this embodiment, the weld contour is the contour edge corresponding to the proofreading contour space. The measurement software is used to obtain the longest interval between two pixel points in the proofreading contour space as the diameter of the proofreading contour space, and then a circle is drawn with the center position pixel of the proofreading contour space as the center and the diameter as the standard weld contour of the proofreading contour space. Then, the standard weld contour and the proofreading contour space are overlapped based on the center position pixel, and the non-overlapping weld contour is identified as the second authentication focus result. The second authentication focus result is an image with a high probability of welding defects in the weld contour area of the weld point.

[0069] Using the same method, multiple first authentication focus results and multiple second authentication focus results of multiple grayscale conversion sample images are obtained, and multiple authentication focus results of multiple grayscale conversion sample images are established based on the multiple first authentication focus results and multiple second authentication focus results.

[0070] Based on the multiple authentication focus results, pixel point positioning and pixel grayscale sampling are performed on multiple sample LED internal images in the sample image set to obtain multiple focus pixel grayscale sets, and then pixel grayscale sampling of adjacent pixel points of the focus pixels in the multiple sample LED internal images is performed based on the positioned pixel points to obtain multiple associated pixel grayscale sets.

[0071] After obtaining the grayscale values of the abnormal pixel and the pixels adjacent to the abnormal pixel, the abnormal pixel (the area where welding defects may exist) is made more prominent by adjusting the contrast (enhancement parameter).

[0072] To improve the scientific nature of contrast adjustment, this embodiment pre-constructs an adaptive enhancement processing model based on a back-propagation neural network. The input data of the adaptive enhancement processing model are the grayscale values of abnormal pixels and the grayscale values of adjacent normal pixels, and the output result is the enhancement parameter for pixel contrast adjustment.

[0073] The training data of the adaptive enhancement processing model are multiple groups of sample abnormal pixel grayscale values-sample adjacent normal pixel grayscale values-sample enhancement parameters that effectively improve the visibility of abnormal pixels.

[0074] Based on a conventional back propagation network training method, the obtained multiple sets of data are used as training data to train the adaptive enhancement processing model until the enhancement parameter output accuracy of the adaptive enhancement processing model is higher than 98%.

[0075] After calculating the pixel grayscale mean of the focus pixel grayscale set and the associated pixel grayscale set of each sample LED internal image, the calculation results are input into the adaptive enhancement processing model to generate multiple enhancement parameters for multiple sample LED internal images, and adaptive enhancement is performed on multiple sample LED internal images corresponding to the multiple enhancement parameters to obtain multiple sample LED internal images with improved visualization of areas with a high probability of welding defects as the adaptive enhancement results.

[0076] The defect detection network is also constructed based on the back propagation neural network, and the defect detection network can detect welding defects in the weld shape area and the weld contour area of the weld point.

[0077] The training data for the defect detection network consists of multiple sets of raw weld spot images—welding defect identification images—based on manually labeled welding defects and produced using the same welding process as the samples in this example. The network undergoes supervised training using this training data until it can fully identify and identify welding defects.

[0078] The multiple sample LED internal images with enhanced visualization in the adaptive enhancement results are input into the defect detection network one by one to generate the defect recognition results composed of multiple defect feature images. The defect recognition results are used as all types of welding defects that may appear in the current batch of samples, and all LEDs in the current batch are intelligently detected, thereby removing the LEDs in the current batch with welding defects.

[0079] This embodiment achieves the technical effect of quickly and efficiently extracting the internal welding defect features of LEDs in the same batch and comparing and removing LEDs with welding defects in the same batch based on the defect features, thereby improving the quality control of batch-produced LEDs.

[0080] In one embodiment, the method steps provided by the present application further include:

[0081] A510: Establishing coverage space coordinates based on the proofreading contour space;

[0082] A520: performing welding coverage evaluation of the sample using the coverage space coordinates, wherein the welding coverage evaluation results include coverage integrity evaluation results and coverage position accuracy evaluation results;

[0083] A530: Add the weld coverage evaluation results to the defect identification results.

[0084] Specifically, in this embodiment, the coordinate origin and the coordinate establishment direction are located according to the internal structure of the LED, and a qualified LED internal image of the qualified LED is interactively obtained. The coordinate origin is located and the coordinate establishment direction is determined in the qualified LED internal image to construct a standard coordinate system, and the contour coordinates of the qualified solder joints and the center coordinates of the solder joints are extracted in the standard coordinate system as a benchmark for solder joint coverage evaluation.

[0085] Then, in the internal image of the sample LED, the coordinate origin is located and the coordinate establishment direction is determined to construct a standard coordinate system, and then the coverage space coordinates of the calibration contour space are extracted based on the standard coordinate system. The coverage space coordinates include the contour coordinates of the actual solder joint and the center coordinates of the actual solder joint.

[0086] The percentage of the overlap between the contour coordinates of the actual solder joint and the contour coordinates of the qualified solder joint in the total contour coordinate data of the qualified solder joint is calculated as the coverage integrity evaluation result; the distance between the center coordinates of the actual solder joint and the center coordinates of the qualified solder joint is calculated as the coverage position accuracy evaluation result.

[0087] By analogy, multiple groups of coverage integrity evaluation results and coverage position accuracy evaluation results of multiple sample LED internal images are calculated and their averages are calculated respectively to obtain the coverage welding evaluation results of the sample set. The coverage welding evaluation results are used as a reference for optimizing the welding process. This embodiment achieves the technical effect of providing a reference for optimizing the welding process.

[0088] In one embodiment, the method steps provided by the present application further include:

[0089] A610: Determine whether the defect identification result satisfies the abnormal defect threshold for batch abnormality;

[0090] A620: If present, perform optical inspection of random sample cross sections;

[0091] A630: Perform batch anomaly certification based on cross-sectional optical inspection results, and complete intelligent inspection of batch samples based on the certification results.

[0092] Specifically, in this embodiment, the batch abnormality threshold includes multiple types of defects. If the defect type composition of the defect identification result exceeds the abnormal defect type composition of the batch abnormality threshold, it indicates that there is an operational defect of the welding equipment. Then, the welds obtained by welding based on the operational defect of the welding equipment may have internal welding defects that are not external welding defects, including but not limited to pores, cracks or uneven distribution of solder inside the welds. Based on this, this embodiment performs destructive cross-sectional optical inspection of the welds of random samples.

[0093] Using the existing cross-sectional optical analysis method, the cross-sectional optical inspection results are used to analyze and evaluate whether the detected defects affect the performance and reliability of LED products, and to determine whether internal welding defects will affect the performance and reliability of the product.

[0094] When internal solder defects affect product performance and reliability, batch abnormality certification is performed based on the cross-sectional optical inspection results. Based on this certification result, intelligent testing of batch samples is performed to automatically identify products from the same batch that contain the same or similar internal solder defects. This embodiment eliminates the risk of soldering equipment failure and ensures LED quality control.

[0095] Example 2

[0096] Based on the same inventive concept as the intelligent detection method for internal welding defects of LEDs in the aforementioned embodiment, Figure 3 As shown, the present application provides an intelligent detection system for internal welding defects of LEDs, wherein the system includes:

[0097] The sample set establishment unit 1 is used to establish a batch sampling ratio. When any batch of samples is subjected to defect detection, random samples are taken according to the batch sampling ratio to establish a sample set.

[0098] A parameter set establishing unit 2 is configured to establish a parameter set for the sample, the parameter set including a structural parameter set and a size parameter set, and to configure X-ray sampling control parameters using the parameter set, the sampling control parameters including exposure time and ray energy, and to configure enhancement parameters based on the sampling control parameter configuration and sample data;

[0099] An image sampling execution unit 3 is configured to execute image sampling of the sample set by X-rays, and perform batch processing of the image sampling results using the enhancement parameters to generate a sample image set;

[0100] The grayscale conversion execution unit 4 is used to execute the grayscale conversion of the sample image set, and perform grayscale authentication of the grayscale conversion results based on the adjacent pixels of the position point to generate an authentication attention result, adaptively enhance the sample image set with the authentication attention result, and input the adaptive enhancement result into the defect detection network to generate a defect recognition result, and complete the intelligent detection of the batch samples according to the defect recognition result.

[0101] In one embodiment, the grayscale conversion execution unit 4 further includes:

[0102] Obtaining the welding process corresponding to the sample set and establishing a welding interval;

[0103] Using the welding interval as a contour recognition space, performing edge contour search on the grayscale conversion result, and establishing a proofreading contour space based on the search results;

[0104] Performing grayscale authentication of adjacent pixels in the proofreading contour space to generate a first authentication focus result;

[0105] generating a second authentication focus result based on the contour edge corresponding to the proofread contour space;

[0106] An authentication focus result is established based on the first authentication focus result and the second authentication focus result.

[0107] In one embodiment, the grayscale conversion execution unit 4 further includes:

[0108] Identifying the number of spatial pixels in the proofreading contour space and configuring a local verification window according to the identification result;

[0109] Using the local verification window to perform grayscale anomaly check on adjacent pixels in the proofreading contour space;

[0110] Grayscale authentication is completed according to the verification result to generate a first authentication focus result.

[0111] In one embodiment, the grayscale conversion execution unit 4 further includes:

[0112] Placing the local verification window in the proofreading contour space, and extracting the grayscale values of pixels at positions within the window through the local verification window;

[0113] Perform extreme value removal on the grayscale values of the pixels at the position, calculate a grayscale mean based on the extreme value removal result, and verify and update the grayscale value of the pixel at the center position with the grayscale mean;

[0114] According to the verification update result, grayscale value calibration is performed with the pixels at the center and surrounding positions to complete the grayscale abnormality verification of adjacent pixels.

[0115] In one embodiment, the system further comprises:

[0116] Establishing coverage space coordinates based on the calibration contour space;

[0117] Performing welding coverage evaluation of the sample using the coverage space coordinates, wherein the welding coverage evaluation results include coverage integrity evaluation results and coverage position accuracy evaluation results;

[0118] The weld coverage evaluation results are added to the defect identification results.

[0119] In one embodiment, the system further comprises:

[0120] Determining whether the defect identification result satisfies a batch abnormality threshold for abnormal defects;

[0121] If present, perform optical inspection of cross sections of random samples;

[0122] Batch abnormality certification is performed based on the cross-sectional optical inspection results, and intelligent inspection of batch samples is completed based on the certification results.

[0123] In one embodiment, the grayscale conversion execution unit 4 further includes:

[0124] Perform pixel grayscale sampling of the sample image set based on the authentication focus result, and establish a focus pixel grayscale set and an associated pixel grayscale set;

[0125] The grayscale set of the pixel of interest and the grayscale set of the associated pixels are input into an adaptive enhancement processing model to generate enhancement parameters, and the adaptive enhancement is completed with the enhancement parameters.

[0126] Any of the methods or steps described above may be stored as computer instructions or programs in various types of computer memories, and the computer instructions or programs may be recognized by various types of computer processors to implement any of the methods or steps described above.

[0127] Based on the above specific embodiments of the present invention, any improvements and modifications made to the present invention by those skilled in the art without departing from the principles of the present invention shall fall within the scope of patent protection of the present invention.

Claims

1. An intelligent detection method for LED internal welding defects, characterized in that: The method comprises: Establishing a batch sampling ratio, when performing defect detection on any batch of samples, performing random sampling according to the batch sampling ratio to establish a sample set; Establishing a parameter set for the sample, the parameter set including a structural parameter set and a size parameter set, configuring X-ray sampling control parameters using the parameter set, the sampling control parameters including exposure time and ray energy, and configuring enhancement parameters based on the sampling control parameter configuration and the sample data; performing image sampling of the sample set by X-rays, and performing batch processing of the image sampling results with the enhancement parameters to generate a sample image set; Perform grayscale conversion on the sample image set, perform grayscale authentication on the grayscale conversion results based on the grayscale of adjacent pixels at the location point, generate an authentication focus result, perform adaptive enhancement on the sample image set using the authentication focus result, input the adaptive enhancement result into the defect detection network, generate a defect recognition result, and complete intelligent detection of batch samples based on the defect recognition result; The generating of the authentication attention result includes: Obtaining the welding process corresponding to the sample set and establishing a welding interval; Using the welding interval as a contour recognition space, performing edge contour search on the grayscale conversion result, and establishing a proofreading contour space based on the search results; Performing grayscale authentication of adjacent pixels in the proofreading contour space to generate a first authentication focus result; generating a second authentication focus result based on the contour edge corresponding to the proofread contour space; Establishing an authentication focus result according to the first authentication focus result and the second authentication focus result; The step of performing grayscale authentication of adjacent pixels in the proofreading contour space to generate a first authentication focus result includes: Identifying the number of spatial pixels in the proofreading contour space and configuring a local verification window according to the identification result; Using the local verification window to perform grayscale anomaly check on adjacent pixels in the proofreading contour space; Complete grayscale authentication based on the verification results to generate a first authentication focus result; The performing grayscale abnormality check of adjacent pixels in the proofreading contour space using the local verification window includes: Placing the local verification window in the proofreading contour space, and extracting the grayscale values of pixels at positions within the window through the local verification window; Perform extreme value removal on the grayscale values of the pixels at the position, calculate a grayscale mean based on the extreme value removal result, and verify and update the grayscale value of the pixel at the center position with the grayscale mean; According to the verification update result, grayscale value calibration is performed with the pixels at the center and surrounding positions to complete the grayscale abnormality check of adjacent pixels; Establishing coverage space coordinates based on the calibration contour space; Performing welding coverage evaluation of the sample using the coverage space coordinates, wherein the welding coverage evaluation results include coverage integrity evaluation results and coverage position accuracy evaluation results; The weld coverage evaluation results are added to the defect identification results.

2. The method according to claim 1, wherein The method comprises: Determining whether the defect identification result satisfies a batch abnormality threshold for abnormal defects; If present, perform optical inspection of cross sections of random samples; Batch abnormality certification is performed based on the cross-sectional optical inspection results, and intelligent inspection of batch samples is completed based on the certification results.

3. The method according to claim 1, wherein The adaptive enhancement of the sample image set based on the authentication focus result includes: Perform pixel grayscale sampling of the sample image set based on the authentication focus result, and establish a focus pixel grayscale set and an associated pixel grayscale set; The grayscale set of the pixel of interest and the grayscale set of the associated pixels are input into an adaptive enhancement processing model to generate enhancement parameters, and the adaptive enhancement is completed with the enhancement parameters.

4. An intelligent detection system for LED internal welding defects, characterized in that: The steps for implementing the method according to any one of claims 1 to 3 include: A sample set establishment unit is used to establish a batch sampling ratio. When any batch of samples is subjected to defect detection, random samples are taken according to the batch sampling ratio to establish a sample set. a parameter set establishing unit, configured to establish a parameter set for the sample, the parameter set including a structural parameter set and a size parameter set, and to configure X-ray sampling control parameters using the parameter set, the sampling control parameters including exposure time and ray energy, and to configure enhancement parameters based on the sampling control parameter configuration and sample data; an image sampling execution unit, configured to execute image sampling of the sample set by X-rays, and perform batch processing of the image sampling results using the enhancement parameters to generate a sample image set; A grayscale conversion execution unit is used to execute grayscale conversion of a sample image set, and perform grayscale authentication of the grayscale conversion results based on the grayscale of adjacent pixels at the position points to generate an authentication focus result, adaptively enhance the sample image set using the authentication focus result, and input the adaptive enhancement result into a defect detection network to generate a defect recognition result, and complete intelligent detection of batch samples based on the defect recognition result.

Images