CN117422681A - PCB surface defect detection method based on improved YOLOv8 algorithm - Google Patents
PCB surface defect detection method based on improved YOLOv8 algorithm Download PDFInfo
- Publication number
- CN117422681A CN117422681A CN202311378097.5A CN202311378097A CN117422681A CN 117422681 A CN117422681 A CN 117422681A CN 202311378097 A CN202311378097 A CN 202311378097A CN 117422681 A CN117422681 A CN 117422681A
- Authority
- CN
- China
- Prior art keywords
- yolov8
- network
- pcb
- module
- defect detection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 67
- 230000007547 defect Effects 0.000 title claims abstract description 48
- 238000000034 method Methods 0.000 claims abstract description 24
- 230000004927 fusion Effects 0.000 claims abstract description 19
- 230000007246 mechanism Effects 0.000 claims abstract description 15
- 230000008569 process Effects 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 15
- 238000005070 sampling Methods 0.000 claims abstract description 11
- 238000012360 testing method Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 3
- 210000000988 bone and bone Anatomy 0.000 claims description 2
- 230000009466 transformation Effects 0.000 claims description 2
- 238000010200 validation analysis Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000002457 bidirectional effect Effects 0.000 description 3
- 238000002679 ablation Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000003416 augmentation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000005476 soldering Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30148—Semiconductor; IC; Wafer
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a PCB surface defect detection method based on an improved YOLOv8 algorithm; the method adopts an improved model based on YOLOv8 to detect defects; in the model, an up-sampling process is added in a Yolov8 Neck network, original FPN and PAN structures are replaced by BiFPN structures, and a multi-level feature network is obtained by means of weighted feature fusion and cross-scale connection; providing a C2f-CAM module, wherein the module fuses a CAM attention mechanism into the C2f module, and replaces the C2f module in the up-sampling process in a Neck network with the C2f-CAM module; meanwhile, a loss function is optimized based on the normalized Wasserstein distance is introduced during model training; the method can improve the detection precision of the surface defects of the PCB, has higher accuracy and recall rate, and solves the problem of poor effect of the YOLOv8 algorithm in the detection process of the small target.
Description
Technical Field
The invention belongs to the technical field of printed circuit board defect detection, and particularly relates to a PCB surface defect detection method based on an improved YOLOv8 algorithm.
Background
The PCB (Printed Circuit Board ) is an important component in electronic devices for enabling the connection and support of electronic components. Surface defects such as short circuits, open circuits, soldering problems, etc. may cause circuit failures and performance degradation during PCB manufacturing. Therefore, the surface defect detection of the PCB is important to improve the quality and reliability of the electronic product.
In recent years, deep learning technology has made remarkable progress in the field of computer vision and has made remarkable results in the task of target detection. YOLO (You Only Look Once) is a real-time target detection algorithm based on deep learning, which is receiving a great deal of attention at high speed and accuracy. YOLOv8 is the latest version in the YOLO series algorithm. However, the conventional YOLOv8 algorithm has poor detection effect on small targets.
Disclosure of Invention
The technical problem to be solved by the invention is to provide the PCB surface defect detection method based on the improved YOLOv8 algorithm, which improves the detection precision of the PCB surface defect, improves the YOLOv8 target detection algorithm aiming at the conventional scale target, realizes effective detection of the PCB surface defect, and has higher accuracy and recall rate.
The invention provides a PCB surface defect detection method based on an improved YOLOv8 algorithm. According to the method, an up-sampling process is increased by improving a feature fusion network, a weighted bidirectional feature pyramid network (Bi-directional Feature Pyramid Network, biFPN) is introduced, a multi-level feature network is obtained by utilizing weighted feature fusion and cross-scale connection, and the detection efficiency of a small target is improved. The invention also provides a C2f-CAM module, which fuses a context information attention mechanism (Context Augmentation Module, CAM) into the original C2f module of the YOLOv8, selects more useful context information and extracts finer and meaningful features. Meanwhile, a measurement method of normalized Wasserstein distance (Normalized Wasserstein Distance, NWD) is introduced, the position, shape and visual similarity of the target frame are comprehensively considered, and the accuracy of similarity measurement is improved. The technical scheme of the invention is as follows.
The invention provides a PCB surface defect detection method based on an improved YOLOv8 algorithm, which comprises the following steps:
step one: performing data enhancement processing on the disclosed PCB data set to obtain an extended data set; dividing the extended data set into a training set and a testing set;
step two: establishing a modified YOLOv8 PCB surface defect detection model
Adding an up-sampling process in a Neck network of Yolov8, replacing the original FPN and PAN structures of the Neck network in Yolov8 with a weighted bidirectional feature pyramid network BiFPN structure, and obtaining a multi-stage feature network through the BiFPN network by utilizing weighted feature fusion and cross-scale connection; fusing a context information attention mechanism CAM into a C2f module, providing a C2f-CAM module, and replacing the C2f module in the up-sampling process in a Neck network with the C2f-CAM module; constructing a PCB surface defect detection model based on improved YOLOv8, wherein the PCB surface defect detection model comprises an input end, a back bone network, a Neck network and a detection head;
step three: training the PCB surface defect detection model based on the improved YOLOv8 in a training set sample to obtain a target detection optimal model;
step four: and taking sample data in the test set as input of a target detection optimal model, carrying out defect detection through the target detection optimal model, outputting a defect detection result, and evaluating the detection result.
In the first step of the invention, the data enhancement processing comprises random overturn, random mirror image, random brightness change and random scaling processing; proportion 8 of training set and validation set: 2.
in the second step, in the improved YOLOv8 PCB surface defect detection model, an image is sent into a backhaul network from an input end to perform feature extraction, then a Neck network performs feature fusion on the extracted features, and the fused features are sent into three detection heads with different scales to perform detection, so that a detection result is obtained.
In the second step, the Neck network in the YOLOv8 is improved, a sampling process for the shallow feature map is added, the original FPN+PAN structure is replaced by the BiFPN structure, multi-scale feature fusion is carried out through the BiFPN network, different weights are given to each layer for fusion, the network is focused on important layers, node connection of unnecessary layers is reduced, the model has better performance when detecting PCB surface defects with different scales, and the problems of small target missing detection and false detection are solved.
In the second step, the C2f-CAM module inserts the CAM attention mechanism before the last convolution layer in the C2f module, namely the convolution layer performs CAM attention conversion on the output feature map, and then the converted feature map is obtained as the input of the convolution layer, so that the convolution layer receives more useful channel information, and the expression capability and the discrimination of the feature map are improved.
In the third step, a small target detection evaluation method based on normalized Wasserstein distance, namely NWD measurement, is introduced during training. Obtaining a boundary frame of target detection, wherein the boundary frame comprises a prediction frame and a real frame, and based on the prediction frame and the real frame, a Wasserstein distance is obtained; normalizing the Wasserstein distance; the loss function of the YOLOv 8-based PCB surface defect detection model was improved according to the normalized wasperstein distance. The Loss functions of yolov8 include classification Loss (VFL Loss) and regression Loss (CIOU loss+df Loss), which are the Loss functions responsible for localization, and the present method improves CIOU Loss by NWD.
Compared with the prior art, the invention has the beneficial effects that:
the invention aims at PCB surface defect detection, improves the YOLOv8 algorithm, comprises improving a feature fusion network, fusing a context information attention mechanism and optimizing a YOLOv8 model with an improved loss function, improves the feature fusion capability, aims at the problem of poor effect in the small target detection process, improves the detection precision of the model, and increases the generalization capability and robustness of the model. The PCB surface defect detection method based on the improved YOLOv8 algorithm has higher practical value and application prospect. By the method, the efficiency and the accuracy of PCB manufacture and quality control can be improved, so that the quality and the reliability of electronic products are improved.
Drawings
FIG. 1 is a schematic diagram of a modified YOLOv8 network model.
Fig. 2 is a schematic diagram of an improved feature fusion network.
FIG. 3 is a schematic diagram of a CAM module.
FIG. 4 is a schematic diagram of a C2f-CAM module.
Detailed Description
The invention will be further described with reference to the drawings and examples, which should not be construed as limiting the scope of the invention.
Example 1
1. Data preprocessing: using the disclosed PCB defect dataset, write python code, data enhance the dataset to 4899 sheets and convert it to YOLO training format, following 8: the scale of 2 is divided into training and test sets.
2. And improving the YOLOv8 network model by improving a feature fusion network, fusing a context information attention mechanism and optimizing a loss function, and constructing a PCB surface defect detection model based on the improved YOLOv 8. The improved YOLOv8 network model structure is shown in fig. 1. The specific improvement steps can be divided into three steps:
2.1 improved feature fusion network
An up-sampling process is added, a weighted bidirectional feature pyramid network (BiFPN) is introduced into a Neck network part of a YOLOv8 model, an original Concat module is replaced by a BiFPN module, and an improved feature fusion network schematic diagram is shown in figure 2; biFPN is a feature fusion mechanism with weights, each path is assigned with a learnable weight, the weights are continuously updated through learning of the data features, so that more important information is obtained, and the network feature fusion capability is further enhanced through the learnable weights. BiFPN realizes trans-scale feature fusion through the following fusion formula:
wherein w is i Representing the weight value obtained after the network training; i i Representing the characteristics of the input.
2.2 fusion context information attention mechanism
The invention provides a C2f-CAM module, wherein the CAM attention mechanism is a neural network module based on a context information mechanism, and the CAM module consists of a space attention module and a channel attention module which are respectively used for extracting context information of space and channel dimensions, and the structure schematic diagram of the CAM module is shown in figure 3. For the design of the module, the invention adopts the CAM attention mechanism added in the original C2f module of the YOLOv8 to enhance the characteristics of the context information among different channels. The schematic structure of the C2f-CAM module is shown in fig. 4, and a CAM attention mechanism is inserted before the last convolution layer in the C2f module, that is, the convolution layer performs CAM attention transformation on the output feature map, and then the transformed feature map is used as the input of the convolution layer. This allows the convolutional layer to receive more useful channel information, improving the expressive power and discrimination of feature maps. The C2f module of the Neck network up-sampling process is replaced by a C2f-CAM module.
The C2f-CAM module can improve the detection effect on small targets, dense targets, shielding targets and other targets difficult to detect. Features with different dimensions and positions can be selected in a self-adaptive manner and are dynamically fused so as to adapt to defects with different dimensions and shapes; more useful channel information can be selected by the attention mechanism, extracting finer and meaningful features.
2.3 optimizing the loss function
The NWD is introduced to measure and optimize the loss function, and has the advantages of insensitivity to objects with different scales and suitability for measuring the similarity between tiny objects. In the defect detection process, a boundary box is modeled as two-dimensional Gaussian distribution, and the second-order Wasserstein distance between the boundary boxes is
Wherein,representing the second order Wasserstein distance, cx between bounding box A and bounding box B A Is the central abscissa, cy, of bounding box A A Is the central ordinate, w, of the bounding box A A For the width of bounding box A, h A To the height of bounding box A, cx B Is the center abscissa, cy, of the bounding box B B Is the central ordinate, w, of the bounding box B B For the width of the boundary box B, h B For the height of bounding box B, TT represents the transpose.
The second order wasperstein distance is normalized and a so-called normalized gaussian wasperstein distance is obtained:
in the formula, NWD (mu) A ,μ B ) Representing the normalized Gaussian Wasserstein distance between bounding box A and bounding box B, C is the normalization constant.
L NWD =1-NWD(N a ,N b )
Wherein L is NWD Is normalized Wasserstein loss, N a Is the gaussian distribution of the predicted bounding box a,N b is a gaussian distribution of a true bounding box B, NWD (N a ,N b ) Is the normalized wasperstein distance of the predicted bounding box a and the real bounding box B.
3. And (3) sending the PCB surface defect training set into the improved YOLOv8 network model for training to obtain the defect detection optimal model.
4. And taking sample data in the test set as input, carrying out target detection through the improved model, outputting a target detection result, and evaluating the detection result. The performance of the improved YOLOv8 model is objectively evaluated, the conventional accuracy (P), recall (R) and average accuracy average value mAP are used as evaluation indexes, an ablation experiment is performed on the model, and the defect detection performance results (average value of 6 defect detection performances) are shown in table 1.
Table 1 ablation experiments based on the modified YOLOv8 algorithm
Algorithm | P | R | mAP0.5 | mAP0.5:0.95 |
YOLOv8 | 98.7% | 96.1% | 98.1% | 74.0% |
YOLOv8+BiFPN | 99.2% | 97.3% | 98.9% | 75.8% |
YOLOv8+BiFPN+C2F-CAM | 99.2% | 97.5% | 99.2% | 78.7% |
YOLOv8+BiFPN+C2F-CAM+NWD | 99.4% | 97.8% | 99.3% | 79.0% |
As can be seen from Table 1, the PCB surface defect detection model based on improved YOLOv8 provided by the invention has the advantages that the accuracy, recall rate and mAP0.5 average precision are respectively improved by 0.7%,1.7%,1.2% and 5% on the mAP0.5:0.95 average precision, the defect detection effect is obviously improved on the severe index of mAP0.5:0.95, and the model has good reference significance for PCB defect detection application.
While the present invention has been described in detail through the foregoing description of the preferred embodiment, it should be understood that the foregoing description is not to be considered as limiting the invention. Many modifications and substitutions of the present invention will become apparent to those of ordinary skill in the art upon reading the foregoing. Accordingly, the scope of the invention should be limited only by the attached claims.
Claims (5)
1. A PCB surface defect detection method based on an improved YOLOv8 algorithm is characterized by comprising the following steps:
step one: performing data enhancement processing on the disclosed PCB data set to obtain an extended data set; dividing the extended data set into a training set and a testing set;
step two: establishing a modified YOLOv8 PCB surface defect detection model
Adding an up-sampling process in a Neck network of the YOLOv8, replacing the original FPN and PAN structures of the Neck network in the YOLOv8 with a BiFPN structure, and obtaining a multi-stage feature network through the BiFPN network by utilizing weighted feature fusion and cross-scale connection; fusing a context information attention mechanism CAM into a C2f module, providing a C2f-CAM module, and replacing the C2f module in the up-sampling process in a Neck network with the C2f-CAM module; constructing a PCB surface defect detection model based on improved YOLOv8, wherein the PCB surface defect detection model comprises an input end, a back bone network, a Neck network and a detection head;
step three: training the PCB surface defect detection model based on the improved YOLOv8 in a training set sample to obtain a target detection optimal model;
step four: and taking sample data in the test set as input of a target detection optimal model, carrying out defect detection through the target detection optimal model, outputting a defect detection result, and evaluating the detection result.
2. The method for detecting surface defects of a PCB based on the modified YOLOv8 algorithm of claim 1, wherein in step one, the data enhancement process includes performing random flipping, random mirroring, random brightness variation, and random scaling; proportion 8 of training set and validation set: 2.
3. the method for detecting surface defects of a PCB based on the modified YOLOv8 algorithm of claim 1, wherein in the second step, the network of neg is modified, a sampling process for shallow feature map is added, and the FPN and PAN structures are replaced with BiFPN structures.
4. The method for detecting surface defects of a PCB based on the modified YOLOv8 algorithm of claim 1, wherein in the second step, the C2f-CAM module inserts a CAM attention mechanism before the last convolution layer in the C2f module, i.e. the convolution layer is used to perform CAM attention transformation on the output feature map, and then the transformed feature map is used as an input of the convolution layer, which enables the convolution layer to receive more useful channel information, and improves the expressive power and the distinguishing degree of the feature map.
5. The method for detecting the surface defects of the PCB based on the improved YOLOv8 algorithm according to claim 1, wherein in the third step, when a PCB surface defect detection model based on the improved YOLOv8 is trained, a normalized Wasserstein distance is introduced as a measurement standard, and a loss function is optimized;
L NWD =1-NWD(N a ,N b )
wherein L is NWD Is normalized Wasserstein loss, N a Is the Gaussian distribution of the predicted bounding box A, N b Is a gaussian distribution of a true bounding box B, NWD (N a ,N b ) Is the normalized Wasserstein distance of the predicted bounding box A and the actual bounding box B.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311378097.5A CN117422681A (en) | 2023-10-24 | 2023-10-24 | PCB surface defect detection method based on improved YOLOv8 algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311378097.5A CN117422681A (en) | 2023-10-24 | 2023-10-24 | PCB surface defect detection method based on improved YOLOv8 algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117422681A true CN117422681A (en) | 2024-01-19 |
Family
ID=89529668
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311378097.5A Pending CN117422681A (en) | 2023-10-24 | 2023-10-24 | PCB surface defect detection method based on improved YOLOv8 algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117422681A (en) |
-
2023
- 2023-10-24 CN CN202311378097.5A patent/CN117422681A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108492272B (en) | Cardiovascular vulnerable plaque identification method and system based on attention model and multitask neural network | |
CN108537286B (en) | Complex target accurate identification method based on key area detection | |
Kadam et al. | Detection and localization of multiple image splicing using MobileNet V1 | |
CN112037219A (en) | Metal surface defect detection method based on two-stage convolution neural network | |
CN116258707A (en) | PCB surface defect detection method based on improved YOLOv5 algorithm | |
CN113673509B (en) | Instrument detection classification method based on image text | |
CN112949338A (en) | Two-dimensional bar code accurate positioning method combining deep learning and Hough transformation | |
CN111833310B (en) | Surface defect classification method based on neural network architecture search | |
CN117152123B (en) | Detection positioning optimization method, system and storage medium for solder paste printing | |
Akhtar et al. | Optical character recognition (OCR) using partial least square (PLS) based feature reduction: An application to artificial intelligence for biometric identification | |
CN111242144A (en) | Method and device for detecting abnormality of power grid equipment | |
CN111639697B (en) | Hyperspectral image classification method based on non-repeated sampling and prototype network | |
CN115439458A (en) | Industrial image defect target detection algorithm based on depth map attention | |
CN103886332A (en) | Method for detecting and identifying defects of metallic meshes | |
CN114723833A (en) | Improved YOLOV 5-based deep learning wafer solder joint detection method | |
CN113447771A (en) | Partial discharge pattern recognition method based on SIFT-LDA characteristics | |
CN116087880A (en) | Radar radiation source signal sorting system based on deep learning | |
CN113763364B (en) | Image defect detection method based on convolutional neural network | |
CN114998756A (en) | Yolov 5-based remote sensing image detection method and device and storage medium | |
Yao et al. | Dual-attention transformer and discriminative flow for industrial visual anomaly detection | |
CN111291712B (en) | Forest fire recognition method and device based on interpolation CN and capsule network | |
CN117516937A (en) | Rolling bearing unknown fault detection method based on multi-mode feature fusion enhancement | |
CN112966730A (en) | Vehicle damage identification method, device, equipment and storage medium | |
CN116704526A (en) | Staff scanning robot and method thereof | |
CN117422681A (en) | PCB surface defect detection method based on improved YOLOv8 algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |