Circuit hole plate detection method and device, electronic equipment and storage medium
By performing super-resolution processing on the transmittance image, the problem of high cost in orifice plate inspection in existing technologies is solved, and efficient and low-cost orifice plate quality inspection is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHENGDU INTELLIFUSION TECH CO LTD
- Filing Date
- 2022-12-31
- Publication Date
- 2026-07-21
Smart Images
Figure CN116109580B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence, and in particular to a method, apparatus, electronic device, and storage medium for detecting circuit orifice plates. Background Technology
[0002] In current circuit board manufacturing processes, standardized production is generally used to ensure the quality of holes on the circuit board. However, this method is highly dependent on the precision of the drilling equipment. When the precision of the drilling equipment decreases, the yield rate of circuit boards printed through perforated plates will be low. Before printing, the perforated plates need to be inspected. Existing hole inspection machines generally use a varying number of image sensors and non-contact detection methods such as transmitted light to acquire images. The acquired images are then processed such as grayscale and binarization. The principle of light transmission through holes and light blocking by non-holes is used to inspect the holes on the perforated plate. This requires high precision from the image sensors. However, the higher the precision of the image sensor, the higher its price, and the higher the inspection cost of the perforated plate. Summary of the Invention
[0003] This invention provides a method for detecting perforated plates, aiming to solve the problem of high image acquisition costs in existing technologies that use images for perforated plate detection. By performing a first super-resolution processing on the transmittance image, the image resolution of the first processed image is improved. Furthermore, by performing a second super-resolution processing on the region of interest (ROI) in the first processed image, the image resolution of the ROI is further improved. This eliminates the need for a high-precision image sensor to acquire the transmittance image of the perforated plate under test, enabling the acquisition of a high-resolution second processed image for the quality inspection of the perforated plate, thus reducing the cost of perforated plate quality inspection.
[0004] In a first aspect, embodiments of the present invention provide a method for detecting circuit vias, the method comprising:
[0005] A transmittance image of the plate to be tested is obtained, and the transmittance image is subjected to a first super-resolution processing to obtain a first processed image;
[0006] The region of interest in the first processed image is subjected to a second super-resolution processing to obtain a second processed image. The first super-resolution processing and the second super-resolution processing are different super-resolution processing methods.
[0007] Based on the second processed image, the quality of the perforated plate is inspected to obtain the quality inspection result of the perforated plate under test.
[0008] Optionally, performing a first super-resolution process on the transmittance image to obtain a first processed image includes:
[0009] Determine the center pixel of the transmittance image;
[0010] Using the center pixel as the interpolation starting point, bilinear interpolation is performed on the light transmission image to obtain the first processed image.
[0011] Optionally, the second super-resolution processing of the region of interest in the first processed image includes:
[0012] The first processed image is segmented to obtain segmented image regions;
[0013] The region of interest in the first processed image is determined based on the image segmentation region;
[0014] The region of interest is subjected to a second super-resolution process to obtain a second processed image.
[0015] Optionally, performing a second super-resolution processing on the region of interest to obtain a second processed image includes:
[0016] The region of interest is reconstructed using a trained image reconstruction model to obtain a second processed image.
[0017] Optionally, before performing super-resolution reconstruction of the region of interest using the trained image reconstruction model to obtain the second processed image, the method further includes:
[0018] Obtain a first dataset and a model to be trained. The first dataset includes a first sample image and a discrimination image, wherein the first sample image is a low-resolution image of the discrimination image, and the model to be trained includes a generative network and a discrimination network.
[0019] The first sample image is input into the generator network to obtain a generated image, and the generated image is then compared with the discriminator network to obtain a discriminator result.
[0020] Based on the identification results, the parameters of the generator network and the identification network are adjusted. The parameter adjustment process is iterated until the number of iterations reaches the preset number, at which point the training is stopped and the trained generator network is obtained.
[0021] Based on the trained generative network, an image reconstruction model is determined.
[0022] Optionally, the step of performing orifice plate quality inspection based on the second processed image to obtain the quality inspection result of the orifice plate under test includes:
[0023] The second processed image is classified using a trained classification model to obtain the classification result of the test plate.
[0024] Based on the classification results, the quality inspection results of the test plate are obtained.
[0025] Optionally, before classifying the second processed image using a trained classification model to obtain the classification result of the test plate, the method further includes:
[0026] Obtain a second dataset and a classification model to be trained. The second dataset includes second sample images and category labels.
[0027] The second sample image is input into the classification model to be trained for classification, and the sample classification result is obtained.
[0028] Calculate the error loss between the sample classification result and the category label, and adjust the parameters of the classification model to be trained based on the error loss between the sample classification result and the category label. After training is completed, a trained classification model is obtained.
[0029] In a second aspect, embodiments of the present invention provide a circuit board inspection device, the device comprising:
[0030] The first processing module is used to acquire the transmittance image of the plate to be tested, and to perform a first super-resolution processing on the transmittance image to obtain a first processed image.
[0031] The second processing module is used to perform a second super-resolution processing on the region of interest in the first processed image to obtain a second processed image, wherein the first super-resolution processing and the second super-resolution processing are different super-resolution processing methods.
[0032] The detection module is used to perform quality detection of the perforated plate based on the second processed image, and obtain the quality detection result of the perforated plate to be tested.
[0033] Thirdly, embodiments of the present invention provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in the circuit hole plate detection method provided in embodiments of the present invention.
[0034] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the circuit board detection method provided in the embodiments of the present invention.
[0035] In this embodiment of the invention, a transmittance image of the perforated plate to be tested is acquired, and the transmittance image is subjected to a first super-resolution processing to obtain a first processed image; a second super-resolution processing is performed on the region of interest in the first processed image to obtain a second processed image, wherein the first super-resolution processing and the second super-resolution processing are different super-resolution processing methods; the perforated plate quality is detected based on the second processed image to obtain the quality detection result of the perforated plate to be tested. By performing the first super-resolution processing on the transmittance image, the image resolution of the first processed image is improved; by performing the second super-resolution processing on the region of interest in the first processed image, the image resolution of the region of interest is further improved. A second processed image with higher image resolution can be obtained for the quality detection of the perforated plate without the need for a high-precision image sensor to acquire the transmittance image of the perforated plate, thus reducing the detection cost of the perforated plate quality. Attached Figure Description
[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0037] Figure 1 This is a flowchart of a circuit orifice plate detection method provided in an embodiment of the present invention;
[0038] Figure 2 This is a schematic diagram of the structure of a circuit orifice plate detection device provided in an embodiment of the present invention;
[0039] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0041] Please see Figure 1 , Figure 1 This is a flowchart of a circuit board detection method provided in an embodiment of the present invention, such as... Figure 1 As shown, the circuit board inspection method includes the following steps:
[0042] 101. Obtain the transmittance image of the plate to be tested, and perform a first super-resolution processing on the transmittance image to obtain a first processed image.
[0043] In this embodiment of the invention, the test plate is a perforated plate with irregularly shaped holes, and the perforated plate is used for printing on a circuit board.
[0044] The test plate is fixed at a preset photographing station. A backlight is set below the photographing station, and an image acquisition device is set above the photographing station. After the backlight is turned on, the image acquisition device takes a picture of the test plate at the photographing station to obtain the light transmission image of the test plate.
[0045] The image acquisition device mentioned above can be an industrial camera or other types of camera. It should be noted that the image acquisition device mentioned above can be a camera with a lower image sensor accuracy.
[0046] After obtaining the transmittance image of the plate to be tested, the transmittance image can be subjected to first super-resolution processing to improve the resolution of the transmittance image, thereby obtaining the first processed image.
[0047] It should be noted that super-resolution (SR) refers to the process of restoring image details and other data information based on known image information using optics and related optical knowledge; simply put, it is increasing the resolution of an image.
[0048] 102. Perform second super-resolution processing on the region of interest in the first processed image to obtain the second processed image.
[0049] In this embodiment of the invention, the region of interest is a hole region. After obtaining the first processed image, the region of interest can be extracted from the first processed image, and then the region of interest is subjected to a second super-resolution processing to improve the resolution of the region of interest and enhance the image details of the region of interest.
[0050] The first and second super-resolution processing methods are different. The first super-resolution processing can be an interpolation-based method, while the second super-resolution processing can be a deep learning-based method.
[0051] Interpolation-based super-resolution processing methods can include nearest neighbor interpolation, bilinear interpolation, and bicubic interpolation. Interpolation-based super-resolution processing methods have the advantage of fast processing speed. Therefore, interpolation-based super-resolution processing methods can be used to perform the first super-resolution processing on the transmittance image.
[0052] Deep learning-based super-resolution processing methods can include SR methods based on convolutional neural networks, SR methods based on residual networks (ResNet), and SR methods based on generative adversarial networks (GAN).
[0053] 103. Perform quality inspection of the orifice plate based on the second processed image to obtain the quality inspection result of the orifice plate to be tested.
[0054] In this embodiment of the invention, after obtaining the second processed image, the quality of the orifice plate can be detected by a trained quality detection model. The second processed image is input into the trained quality detection model, and the quality detection model outputs the quality detection result of the orifice plate to be tested.
[0055] The quality inspection results of the above-mentioned orifice plate include quality inspection passed and quality inspection failed. If the quality inspection passes, it means that the quality of the orifice plate is qualified, and if the quality inspection fails, it means that the quality of the orifice plate is unqualified.
[0056] In this embodiment of the invention, a transmittance image of the perforated plate to be tested is acquired, and the transmittance image is subjected to a first super-resolution processing to obtain a first processed image; a second super-resolution processing is performed on the region of interest in the first processed image to obtain a second processed image, wherein the first super-resolution processing and the second super-resolution processing are different super-resolution processing methods; the perforated plate quality is detected based on the second processed image to obtain the quality detection result of the perforated plate to be tested. By performing the first super-resolution processing on the transmittance image, the image resolution of the first processed image is improved; by performing the second super-resolution processing on the region of interest in the first processed image, the image resolution of the region of interest is further improved. A second processed image with higher image resolution can be obtained for the quality detection of the perforated plate without the need for a high-precision image sensor to acquire the transmittance image of the perforated plate, thus reducing the detection cost of the perforated plate quality.
[0057] Optionally, in the step of performing the first super-resolution processing on the transmittance image to obtain the first processed image, the center pixel of the transmittance image can be determined; and bilinear interpolation can be performed on the transmittance image with the center pixel as the interpolation starting point to obtain the first processed image.
[0058] In this embodiment of the invention, the transmittance pattern is composed of m×n pixels. After obtaining the transmittance pattern of the plate to be tested, a center pixel is found in the transmittance pattern. If there is no center pixel, for example, if the center point of the transmittance pattern is in the middle of four pixels, then any one of the four pixels can be taken as the center pixel.
[0059] The bilinear interpolation described above first determines the resolution of the target image after interpolation, then maps the transmittance image to the target image, and uses the four nearest pixels of the interpolation point in the transmittance image to perform interpolation calculation, thereby improving the resolution and obtaining the first processed image.
[0060] Optionally, in the step of performing second super-resolution processing on the region of interest in the first processed image, the first processed image may be segmented to obtain segmented regions; the region of interest in the first processed image may be determined based on the segmented regions; and the region of interest may be subjected to second super-resolution processing to obtain the second processed image.
[0061] In this embodiment of the invention, the region of interest is the light-transmitting hole region. The geometric center point of the light-transmitting hole in the plate to be tested can be determined. The geometric center point is used as the seed point for region growth. The seed point is used as the growth starting point. Then, the pixels in the surrounding area of the seed point are grown and merged according to the growth rules that the pixel value error is within the preset range until there are no pixels that can meet the growth point, thus obtaining the light-transmitting hole region.
[0062] In one possible embodiment, the light transmittance image is converted to grayscale to obtain a grayscale image of the light transmittance image. A grayscale threshold is calculated by mean iterative segmentation. The light transmittance image is then divided into a light-transmitting aperture region and a background region by using an accuracy threshold.
[0063] After obtaining the region of the light-transmitting aperture, a masking process is applied to the region of the non-light-transmitting aperture, representing the non-light-transmitting aperture region with a mask value. After masking, the region of the light-transmitting aperture is selected as the region of interest, and a second super-resolution processing is performed on the region of interest to obtain the second processed image.
[0064] Optionally, in the step of performing a second super-resolution processing on the region of interest to obtain a second processed image, the region of interest can be reconstructed using a trained image reconstruction model to obtain the second processed image.
[0065] In this embodiment of the invention, the above-mentioned image reconstruction model can be an SRResNet generator network, which processes low-resolution images by inputting them into the SRResNet generator network and outputs high-resolution images.
[0066] Specifically, after obtaining the region of interest, the non-regions of interest are masked to obtain a mask image containing the region of interest. The mask image is then input into the SRResNet generator network for processing to obtain a high-resolution second processed image.
[0067] It should be noted that the first super-resolution processing uses bilinear interpolation, which can quickly perform super-resolution processing on the light transmission image and improve the global resolution of the light transmission image. The second super-resolution processing uses an image reconstruction model to perform super-resolution processing on the region of interest, which can improve the local image resolution of the light transmission hole and make the light transmission hole have more image details.
[0068] Optionally, before performing super-resolution reconstruction of the region of interest using a trained image reconstruction model to obtain the second processed image, a first dataset and a model to be trained can be acquired. The first dataset includes a first sample image and a discrimination image, wherein the first sample image is a low-resolution image of the discrimination image, and the model to be trained includes a generator network and a discrimination network. The first sample image is input into the generator network to obtain a generated image, and the generated image and the discrimination image are distinguished by the discrimination network to obtain a discrimination result. The parameters of the generator network and the discrimination network are adjusted according to the discrimination result, and the parameter adjustment process is iterated until the number of iterations reaches a preset number, at which point training stops to obtain a trained generator network. Based on the trained generator network, an image reconstruction model is determined.
[0069] In this embodiment of the invention, the model to be trained can be a generative adversarial model, which includes a generator network and a discriminator network. Specifically, the generative adversarial model can be a generative adversarial model based on an SRGAN network, where SRResNet is the generator network of the SRGAN network, and the discriminator network is used to distinguish between real HR images and HR images reconstructed by SRResNet. The ultimate goal of the SRGAN network is to make the discriminator network unable to distinguish between real HR images and HR images reconstructed by SRResNet.
[0070] Specifically, the first sample image is input into the generator network to obtain a high-resolution HR (High-Resolution Image) generated image. This HR generated image is then input into the discriminator network for comparison with the discriminator image. If the discriminator image is true, it indicates an error in the discriminator network's identification, and its parameters can be adjusted until the discriminator network's identification result is false. If the discriminator image is false, it indicates that the generated image produced by the generator network is not realistic enough, and its parameters can be adjusted until the discriminator network's identification result is true. This adversarial process is iterated until a preset number of iterations is reached, at which point the iteration stops, resulting in a trained generator network. The discriminator network is only used during the training process and is not used in practice. The trained generator network is then designated as the image reconstruction model.
[0071] Optionally, in the step of performing quality detection of the perforated plate based on the second processed image to obtain the quality detection result of the perforated plate to be tested, the second processed image can be classified by a trained classification model to obtain the classification result of the perforated plate to be tested; and the quality detection result of the perforated plate to be tested can be obtained based on the classification result.
[0072] In this embodiment of the invention, the classification model can be a classification model based on a deep convolutional neural network, such as a classification model based on networks like the YOLO series, ResNet series, VggNet series, and MobileNet series.
[0073] Specifically, the second processed image is input into the classification model, which processes the second processed image and outputs the classification result of the second processed image. The classification result can include quality detection passing and quality detection failing. If the quality detection passes, it means that the quality of the test plate is qualified; if the quality detection fails, it means that the quality of the test plate is unqualified.
[0074] Optionally, before classifying the second processed image using the trained classification model to obtain the classification result of the test plate, a second dataset and a classification model to be trained can be obtained. The second dataset includes second sample images and category labels. The second sample images are input into the classification model to be trained for classification to obtain sample classification results. The error loss between the sample classification results and the category labels is calculated, and the parameters of the classification model to be trained are adjusted based on the error loss between the sample classification results and the category labels. After training is completed, a trained classification model is obtained.
[0075] In this embodiment of the invention, a second dataset and a classification model to be trained can be obtained. The second dataset includes second sample images processed using the same processing method as the second processed image, and classification labels corresponding to the second sample images. The classification model to be trained is then subjected to supervised training using the first dataset to obtain a trained classification model. The classification labels can be 0 (representing a failed quality check) and 1 (representing a passed quality check).
[0076] The second sample images mentioned above are the transmission images of various existing types of perforated plates, processed by the first and second super-resolution methods to obtain the second processed images. The second sample images can be manually labeled to obtain corresponding classification labels, with each second sample image corresponding to one classification label. The classification model to be trained can be constructed based on networks such as the YOLO series, ResNet series, VggNet series, and MobileNet series.
[0077] During training, the second sample image is input into the classification model to be trained. The classification model outputs the sample classification result and calculates the error loss between the sample classification result and the corresponding classification label. The optimization objective is to minimize the error loss between the sample classification result and the corresponding classification label. The parameters of the classification model to be trained are adjusted by backpropagation. The above parameter adjustment process is iterated until the number of iterations reaches the preset number or the classification model to be trained converges at the minimum error loss. At this point, training stops and a trained classification model is obtained.
[0078] After obtaining the trained classification model, the second processed image of the test plate is input into the trained classification model for processing, and the classification result is output. In the classification result, 1 indicates that the quality detection has passed and 0 indicates that the quality detection has failed.
[0079] It should be noted that the circuit board detection method provided in this embodiment of the invention can be applied to devices such as smart cameras, smartphones, computers, and servers that are capable of performing circuit board detection.
[0080] Optional, please see Figure 2 , Figure 2 This is a schematic diagram of the structure of a circuit board detection device provided in an embodiment of the present invention, as shown below. Figure 2 As shown, the device includes:
[0081] The first processing module 201 is used to acquire the transmittance image of the plate to be tested, and to perform a first super-resolution processing on the transmittance image to obtain a first processed image.
[0082] The second processing module 202 is used to perform a second super-resolution processing on the region of interest in the first processed image to obtain a second processed image, wherein the first super-resolution processing and the second super-resolution processing are different super-resolution processing methods.
[0083] The detection module 203 is used to perform orifice plate quality detection based on the second processed image to obtain the quality detection result of the orifice plate to be tested.
[0084] Optionally, the acquisition module 201 includes:
[0085] A determination submodule is used to determine the center pixel of the light transmittance image;
[0086] The first processing submodule is used to perform bilinear interpolation on the light transmission image with the center pixel as the interpolation starting point to obtain the first processed image.
[0087] Optionally, the processing module 202 includes:
[0088] The segmentation submodule is used to segment the first processed image to obtain segmented image regions.
[0089] A determination submodule is used to determine the region of interest in the first processed image based on the image segmentation region;
[0090] The second processing submodule is used to perform a second super-resolution processing on the region of interest to obtain a second processed image.
[0091] Optionally, the second processing submodule includes:
[0092] The reconstruction unit is used to perform super-resolution reconstruction of the region of interest using a trained image reconstruction model to obtain a second processed image.
[0093] Optionally, the device further includes:
[0094] The first acquisition module is used to acquire a first dataset and a model to be trained. The first dataset includes a first sample image and a discrimination image, wherein the first sample image is a low-resolution image of the discrimination image, and the model to be trained includes a generative network and a discrimination network.
[0095] The discrimination module is used to input the first sample image into the generation network to obtain a generated image, and to use the discrimination network to distinguish between the generated image and the discrimination image to obtain a discrimination result;
[0096] The adjustment module is used to adjust the parameters of the generator network and the discrimination network according to the discrimination result, and iterate the parameter adjustment process until the number of iterations reaches a preset number, then stop training and obtain the trained generator network.
[0097] The determination module is used to determine the image reconstruction model based on the trained generative network.
[0098] Optionally, the detection module 203 includes:
[0099] The classification submodule is used to classify the second processed image using a trained classification model to obtain the classification result of the test plate.
[0100] The third processing submodule is used to obtain the quality inspection result of the test plate based on the classification result.
[0101] Optionally, the device further includes:
[0102] The second acquisition module is used to acquire the second dataset and the classification model to be trained. The second dataset includes second sample images and category labels.
[0103] The classification module is used to input the second sample image into the classification model to be trained for classification and obtain the sample classification result;
[0104] The calculation module is used to calculate the error loss between the sample classification result and the category label, and to adjust the parameters of the classification model to be trained based on the error loss between the sample classification result and the category label, so that a trained classification model is obtained after training is completed.
[0105] It should be noted that the circuit board detection device provided in this embodiment of the invention can be applied to devices such as smart cameras, smartphones, computers, and servers that can perform circuit board detection methods.
[0106] The circuit board detection device provided in this embodiment of the invention can realize all the processes implemented by the circuit board detection method in the above-described method embodiments, and can achieve the same beneficial effects. To avoid repetition, it will not be described again here.
[0107] See Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention, such as... Figure 3 As shown, it includes: a memory 302, a processor 301, and a computer program for a circuit board inspection method stored in the memory 302 and executable on the processor 301, wherein:
[0108] The processor 301 is used to call the computer program stored in the memory 302 and perform the following steps:
[0109] A transmittance image of the plate to be tested is obtained, and the transmittance image is subjected to a first super-resolution processing to obtain a first processed image;
[0110] The region of interest in the first processed image is subjected to a second super-resolution processing to obtain a second processed image. The first super-resolution processing and the second super-resolution processing are different super-resolution processing methods.
[0111] Based on the second processed image, the quality of the perforated plate is inspected to obtain the quality inspection result of the perforated plate under test.
[0112] Optionally, the processor 301 performs the first super-resolution processing on the transmittance image to obtain a first processed image, including:
[0113] Determine the center pixel of the transmittance image;
[0114] Using the center pixel as the interpolation starting point, bilinear interpolation is performed on the light transmission image to obtain the first processed image.
[0115] Optionally, the second super-resolution processing of the region of interest in the first processed image performed by the processor 301 includes:
[0116] The first processed image is segmented to obtain segmented image regions;
[0117] The region of interest in the first processed image is determined based on the image segmentation region;
[0118] The region of interest is subjected to a second super-resolution process to obtain a second processed image.
[0119] Optionally, the processor 301 performs a second super-resolution processing on the region of interest to obtain a second processed image, including:
[0120] The region of interest is reconstructed using a trained image reconstruction model to obtain a second processed image.
[0121] Optionally, before performing super-resolution reconstruction of the region of interest using the trained image reconstruction model to obtain the second processed image, the method executed by the processor 301 further includes:
[0122] Obtain a first dataset and a model to be trained. The first dataset includes a first sample image and a discrimination image, wherein the first sample image is a low-resolution image of the discrimination image, and the model to be trained includes a generative network and a discrimination network.
[0123] The first sample image is input into the generator network to obtain a generated image, and the generated image is then compared with the discriminator network to obtain a discriminator result.
[0124] Based on the identification results, the parameters of the generator network and the identification network are adjusted. The parameter adjustment process is iterated until the number of iterations reaches the preset number, at which point the training is stopped and the trained generator network is obtained.
[0125] Based on the trained generative network, an image reconstruction model is determined.
[0126] Optionally, the process of performing orifice plate quality detection based on the second processed image, executed by processor 301, to obtain the quality detection result of the orifice plate under test, includes:
[0127] The second processed image is classified using a trained classification model to obtain the classification result of the test plate.
[0128] Based on the classification results, the quality inspection results of the test plate are obtained.
[0129] Optionally, before classifying the second processed image using the trained classification model to obtain the classification result of the test plate, the method executed by the processor 301 further includes:
[0130] Obtain a second dataset and a classification model to be trained. The second dataset includes second sample images and category labels.
[0131] The second sample image is input into the classification model to be trained for classification, and the sample classification result is obtained.
[0132] Calculate the error loss between the sample classification result and the category label, and adjust the parameters of the classification model to be trained based on the error loss between the sample classification result and the category label. After training is completed, a trained classification model is obtained.
[0133] The electronic device provided in this embodiment of the invention can implement all the processes of the circuit board detection method in the above-described method embodiments, and can achieve the same beneficial effects. To avoid repetition, further details are omitted here.
[0134] This invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of the circuit orifice plate detection method provided in this invention and achieves the same technical effect. To avoid repetition, it will not be described again here.
[0135] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (RON), or random access memory (RAN), etc.
[0136] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.
Claims
1. A method for detecting circuit orifice plates, characterized in that, Includes the following steps: A transmittance image of the plate to be tested is obtained, and a first super-resolution processing is performed on the transmittance image to obtain a first processed image; specifically, this includes: determining the center pixel of the transmittance image; using the center pixel as the interpolation starting point, performing bilinear interpolation on the transmittance image to obtain the first processed image; The region of interest (ROI) in the first processed image is subjected to a second super-resolution processing to obtain a second processed image. The first super-resolution processing and the second super-resolution processing are different super-resolution processing methods. Specifically, the process includes: image segmentation of the first processed image to obtain image segmentation regions; determining the ROI in the first processed image based on the image segmentation regions; and performing super-resolution reconstruction of the ROI using a trained image reconstruction model to obtain the second processed image. The trained image reconstruction model is an image reconstruction model based on a generative adversarial network. Based on the second processed image, the quality of the perforated plate is inspected to obtain the quality inspection result of the perforated plate under test.
2. The circuit board inspection method as described in claim 1, characterized in that, Before performing super-resolution reconstruction of the region of interest using the trained image reconstruction model to obtain the second processed image, the method further includes: Obtain a first dataset and a model to be trained. The first dataset includes a first sample image and a discrimination image, wherein the first sample image is a low-resolution image of the discrimination image, and the model to be trained includes a generative network and a discrimination network. The first sample image is input into the generator network to obtain a generated image, and the generated image is then compared with the discriminator network to obtain a discriminator result. Based on the identification results, the parameters of the generator network and the identification network are adjusted. The parameter adjustment process is iterated until the number of iterations reaches the preset number, at which point the training is stopped and the trained generator network is obtained. Based on the trained generative network, an image reconstruction model is determined.
3. The circuit board inspection method as described in claim 2, characterized in that, The step of performing orifice plate quality detection based on the second processed image to obtain the quality detection result of the orifice plate under test includes: The second processed image is classified using a trained classification model to obtain the classification result of the test plate. Based on the classification results, the quality inspection results of the test plate are obtained.
4. The circuit board inspection method as described in claim 3, characterized in that, Before classifying the second processed image using the trained classification model to obtain the classification result of the test plate, the method further includes: Obtain a second dataset and a classification model to be trained. The second dataset includes second sample images and category labels. The second sample image is input into the classification model to be trained for classification, and the sample classification result is obtained. Calculate the error loss between the sample classification result and the category label, and adjust the parameters of the classification model to be trained based on the error loss between the sample classification result and the category label. After training is completed, a trained classification model is obtained.
5. A circuit board detection device, characterized in that, The device includes: The first processing module is used to acquire the transmittance image of the plate under test and perform a first super-resolution processing on the transmittance image to obtain a first processed image; specifically, it includes: determining the center pixel of the transmittance image; using the center pixel as the interpolation starting point, performing bilinear interpolation on the transmittance image to obtain the first processed image; The second processing module is used to perform a second super-resolution processing on the region of interest in the first processed image to obtain a second processed image. The first super-resolution processing and the second super-resolution processing are different super-resolution processing methods. Specifically, it includes: performing image segmentation on the first processed image to obtain image segmentation regions; determining the region of interest in the first processed image based on the image segmentation regions; and performing super-resolution reconstruction on the region of interest using a trained image reconstruction model to obtain the second processed image. The trained image reconstruction model is an image reconstruction model based on a generative adversarial network. The detection module is used to perform quality detection of the perforated plate based on the second processed image, and obtain the quality detection result of the perforated plate to be tested.
6. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the circuit orifice plate detection method as described in any one of claims 1 to 4.
7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the circuit orifice plate detection method as described in any one of claims 1 to 4.