基于轻量化重构网络的表面缺陷视觉检测

基于深度学习的方法在某些工业产品的表面缺陷识别和分类方面表现出优异的性能,然而大多数工业产品缺陷样本稀缺,而且特征差异大,导致这类需要大量缺陷样本训练的检测方法难以适用.提出一种基于重构网络的无监督缺陷检测算法,仅使用容易大量获得的无缺陷样本数据实现对异常缺陷的检测.提出的算法包括两个阶段:图像重构网络训练阶段和表面缺陷区域检测阶段.训练阶段通过一种轻量化结构的全卷积自编码器设计重构网络,仅使用少量正常样本进行训练,使得重构网络能够生成无缺陷重构图像,进一步提出一种结合结构性损失和L1损失的函数作为重构网络的损失函数,解决自编码器检测算法对不规则纹理表面缺陷检测效果较差的问题;缺陷检测阶段以重构图像与待测图像的残差作为缺陷的可能区域,通过常规图像操作即可实现缺陷的定位.对所提出的重构网络的无监督缺陷检测算法的网络结构、训练像素块大小、损失函数系数等影响因素进行了详细的实验分析,并在多个缺陷图像样本集上与其他同类算法做了对比,结果表明重构网络的无监督缺陷检测算法有较强的鲁棒性和准确性.由于重构网络的无监督缺陷检测算法的轻量化结构,检测1 024×1 024像素图像仅仅耗时2.82 ms,...     

基于改进CenterNet的竹条表面缺陷检测方法

在竹条表面缺陷检测中,竹条表面缺陷形状各异,成像环境脏乱,现有基于卷积神经网络(CNN)的目标检测模型面对这样特定的数据时并不能很好地发挥神经网络的优势;而且竹条来源复杂且有其他条件限制,因此没办法采集所有类型的数据,导致竹条表面缺陷数据量少到CNN不能充分学习.针对这些问题,提出一种专门针对竹条表面缺陷的检测网络.该...     

样本仿真结合迁移学习的声呐图像水雷检测

水雷检测对于国防安全具有重要意义,然而,由于水下目标声呐成像实验代价较大,通常难以获得足够的水雷声呐图像样本,因此导致难以应用深度神经网络提高水雷等目标的检测精度.针对这一问题,提出样本仿真结合迁移学习的侧扫声呐图像水雷目标检测与识别方法.首先,根据侧扫声呐成像机理,建立水雷目标的仿真模型,进而仿真得到大量水雷目标样本...     

Quality detection and classification for ultrasonic welding of carbon fiber composites using time-series data and neural network methods

Ultrasonic welding is a novel and efficient technique for joining carbon fiber composites in the automotive industry. Weld quality detection and classification is important to its adoption and deep neural network models are a promising method for this purpose. However, it is difficult to collect the large volume of data needed to train these models with laboratory experiments due to the cost of the materials and cost of weld experiments. Using a limited set of experimental data, a copula multivariate Monte Carlo simulation is proposed to generate large data sets of time-series process signals with similar statistical distributions as the experimental data. The experimental data and simulated data are used to train Bayesian regularized neural network (BRNN) and convolutional neural network (CNN) models to predict weld quality classifications in ultrasonic welding. The results show that BRNN and CNN have similar classification accuracy. But CNN has an advantage in training efficiency compared with BRNN. Both neural-network-based methods were found to be more accurate than support vector machine and k-nearest neighbor methods, when applied to both features extracted from signals and full time-series-based process signals.     

基于U-P-Net的手机玻璃屏幕缺陷分割

随着科技的发展及电子设备的普及,玻璃屏幕质量成为电子设备和其他产品的重要考虑因素;而玻璃外观缺陷检测是玻璃质量检测中最重要的环节,这也是保证产出高品质、高性能的玻璃产品的关键环节;目前玻璃表面缺陷检测方法存在无目标训练图像资源消耗、检测精度较低、复杂特征信息难以提取等问题;因此,为了解决上述问题,提出了一种基于U-pyramid pooling module-Net(U-P-Net)的手机玻璃屏幕缺陷分割模型;采用超像素预处理,有效地降低了原始图像的复杂度;采用ResNet50作为分类网络,减少无目标训练图像造成的资源浪费,提高训练效率;U-P-Net被提出,有效地聚合了不同区域的上下文信息,提高了获取全局信息的能力;实验结果表明,所设计的基于U-P-Net玻璃缺陷分割算法分割精度明显优于其它传统卷积神经网络分割方法,证明了该框架在移动屏幕数据集上的有效性.     

Multi-scale convolutional neural network for multi-focus image fusion

In this study, we present new deep learning (DL) method for fusing multi-focus images. Current multi-focus image fusion (MFIF) approaches based on DL methods mainly treat MFIF as a classification task. These methods use a convolutional neural network (CNN) as a classifier to identify pixels as focused or defocused pixels. However, due to unavailability of labeled data to train networks, existing DL-based supervised models for MFIF add Gaussian blur in focused images to produce training data. DL-based unsupervised models are also too simple and only applicable to perform fusion tasks other than MFIF. To address the above issues, we proposed a new MFIF method, which aims to learn feature extraction, fusion and reconstruction components together to produce a complete unsupervised end-to-end trainable deep CNN. To enhance the feature extraction capability of CNN, we introduce a Siamese multi-scale feature extraction module to achieve a promising performance. In our proposed network we applied multiscale convolutions along with skip connections to extract more useful common features from a multi-focus image pair. Instead of using basic loss functions to train the CNN, our model utilizes structure similarity (SSIM) measure as a training loss function. Moreover, the fused images are reconstructed in a multiscale manner to guarantee more accurate restoration of images. Our proposed model can process images with variable size during testing and validation. Experimental results on various test images validate that our proposed method yields better quality fused images that are superior to the fused images generated by compared state-of-the-art image fusion methods.     

A semi-supervised convolutional neural network-based method for steel surface defect recognition

Automatic defect recognition is one of the research hotspots in steel production, but most of the current methods focus on supervised learning, which relies on large-scale labeled samples. In some real-world cases, it is difficult to collect and label enough samples for model training, and this might impede the application of most current works. The semi-supervised learning, using both labeled and unlabeled samples for model training, can overcome this problem well. In this paper, a semi-supervised learning method using the convolutional neural network (CNN) is proposed for steel surface defect recognition. The proposed method requires fewer labeled samples, and the unlabeled data can be used to help training. And, the CNN is improved by Pseudo-Label. The experimental results on a benchmark dataset of steel surface defect recognition indicate that the proposed method can achieve good performances with limited labeled data, which achieves an accuracy of 90.7% with 17.53% improvement. Furthermore, the proposed method has been applied to a real-world case from a Chinese steel company, and obtains an accuracy of 86.72% which significantly better than the original method in this workshop.     

A sub-region one-to-one mapping (SOM) detection algorithm for glass passivation parts wafer surface low-contrast texture defects

Glass Passivation Parts (GPP) wafer texture defects are one of the most important factors affecting the accuracy of wafer defect detection. Template matching has local errors and low efficiency, and deep learning requires many training samples. In the early stage, defect training sample sets cannot be provided. This paper discusses the design of an effective GPP wafer grain region texture defect detection algorithm using a sub-region one-to-one mapping. A set of standard wafer datum is selected as the reference of grain region segmentation detection, and then the standard wafer images and test GPP wafer images are automatically calibrated and segmented, respectively. Then, a series of pre-processes were performed to equalize the sizes of the two grain-region images. Then the grain region was divided into an equal number of rectangular sub-regions of the same size according to the measurement precision requirement. The correlation degree of each test sub-region is judged by the designed three-channel RGB gray-scale similarity decision functions. Experiments show that the algorithm successfully achieved the necessary calibration and segmentation for the grain region. Compared with the template and histogram matching algorithms, the proposed method does not require a training set, the detection accuracy is significantly improved and the detection efficiency is up to 29.74 times better on average using the proposed algorithm.

     

基于双线性CNN与DenseBlock的导光板标记线缺陷检测

导光板标记线检测是导光板制造品控中的一个重要步骤, 但在使用传统图像算法进行检测的过程中, 有大量的气泡、严重污染和无标记线的情况存在. 因有大量气泡, 严重污染和无标记线的情况造成人工特征难以设计, 因此, 使用基于卷积网络的方法来代替人工特征设计进行缺陷检测. DenseNet 卷积神经网络较其他分类神经网络具有参数较少, 梯度收敛稳定等特点. 因DenseNet 卷积神经网络中使用特征融合的思想, 保证了图片分类准确率. 通过迁移学习的方法, 将训练得到的DenseNet 网络权重迁移到Bilinear-CNN算法进行训练, 提升卷积神经网络局部注意力, 提高图像分类准确率. 通过实现结果表明, 所提方法具有可行性, 相比于V2-ResNet-101网络结构, 准确率提升至95.53%, 参数减少了97.2%, 平均单张图像检测时间减少25%.     

Attention-based convolutional neural network deep learning approach for robust malware classification

Recently, transforming windows files into images and its analysis using machine learning and deep learning have been considered as a state-of-the art works for malware detection and classification. This is mainly due to the fact that image-based malware detection and classification is platform independent, and the recent surge of success of deep learning model performance in image classification. Literature survey shows that convolutional neural network (CNN) deep learning methods are successfully employed for image-based windows malware classification. However, the malwares were embedded in a tiny portion in the overall image representation. Identifying and locating these affected tiny portions is important to achieve a good malware classification accuracy. In this work, a multi-headed attention based approach is integrated to a CNN to locate and identify the tiny infected regions in the overall image. A detailed investigation and analysis of the proposed method was done on a malware image dataset. The performance of the proposed multi-headed attention-based CNN approach was compared with various non-attention-CNN-based approaches on various data splits of training and testing malware image benchmark dataset. In all the data-splits, the attention-based CNN method outperformed non-attention-based CNN methods while ensuring computational efficiency. Most importantly, most of the methods show consistent performance on all the data splits of training and testing and that illuminates multi-headed attention with CNN model's generalizability to perform on the diverse datasets. With less number of trainable parameters, the proposed method has achieved an accuracy of 99% to classify the 25 malware families and performed better than the existing non-attention based methods. The proposed method can be applied on any operating system and it has the capability to detect packed malware, metamorphic malware, obfuscated malware, malware family variants, and polymorphic malware. In addition, the proposed method is malware file agnostic and avoids usual methods such as disassembly, de-compiling, de-obfuscation, or execution of the malware binary in a virtual environment in detecting malware and classifying malware into their malware family.     

基于深度自编码器网络的压盖缺陷检测

口服液压盖过程,会出现压盖不良等情况,瓶盖可能会出现划痕、刮花、表面卷曲、压盖破损等缺陷,为保证食品药品安全必须在出厂前进行检测.在基于深度学习的口服液瓶压盖缺陷检测的研究过程中,使用传统卷积神经网络对口服液压盖缺陷数据集进行训练,需要进行人工标注,效率较低.为有效解决上述问题,设计出一种无监督学习的深度卷积去噪自编码...     

基于CNN的改进行人重识别技术

针对行人重识别研究中训练样本的不足,为提高识别精度及泛化能力,提出一种基于卷积神经网络的改进行人重识别方法。首先对训练数据集进行扩充,使用生成对抗网络无监督学习方法生成无标签图像;然后与原数据集联合作半监督卷积神经网络训练,通过构建一个Siamese网络,结合分类模型和验证模型的特点进行训练;最后加入无标签图像类别分布方法,计算交叉熵损失来进行相似度量。实验结果表明,在Market-1501、CUHK03和DukeMTMC-reID数据集上,该方法相比原有的Siamese方法在Rank-1和mAP等性能指标上有近3～5个百分点的提升。当样本较少时,该方法具有一定应用价值。     

一种端到端的自然场景文本检测与识别模型

提出了一种结合卷积神经网络和递归神经网络的有效的端到端场景文本识别方法。首先使用特征金字塔(FPN)提取图像的多尺度特征,然后将引入残差网络(ResNet)的深度双向递归网络(Bi-LSTM)对这些特征进行编码,获得文本序列特征,进而引入注意力机制(Attention)对文本序列特征进行解码达到识别效果。在ICDAR2013、ICDAR2015数据集实验验证了该算法的有效性,该方法不仅降低了训练难度,而且提升了网络的收敛速度,提高了文本识别准确率。该方法的有效性在ICDAR2013、ICDAR2015数据集上得到了充分验证。     

基于三维卷积神经网络的肺结节假阳性筛查

从 CT 影像中检测肺结节在肺癌的早期诊断中至关重要,而肺结节假阳性的筛查 是提高肺结节检测准确度的重要一步。为了从大量候选结节中快速准确地区分出真正的肺结节, 设计了一个 3D 卷积神经网络(CNN)筛查肺结节假阳性。提出了网络模型,通过恒等映射和残差 单元来加速模型训练,采用单连接路径重复利用特征并重组新特征。基于该模型的肺结节假阳 性筛查方法,与基于 2D CNN 的方法相比,不仅可以省略数据切片步骤,而且能够充分利用 CT 影像的空间信息; 与其他基于 3D CNN 的方法相比,具有参数量小、模型训练快的优点。该方 法在 LUNA16 数据集中的假阳性筛查中取得了较高的敏感度。     

Deep learning based neural network application for automatic ultrasonic computed tomographic bone image segmentation

Deep-learning techniques have led to technological progress in the area of medical imaging segmentation especially in the ultrasound domain. In this paper, the main goal of this study is to optimize a deep-learning-based neural network architecture for automatic segmentation in Ultrasonic Computed Tomography (USCT) bone images in a short time process. The proposed method is based on an end to end neural network architecture. First, the novelty is shown by the improvement of Variable Structure Model of Neuron (VSMN), which is trained for both USCT noise removal and dataset augmentation. Second, a VGG-SegNet neural network architecture is trained and tested on new USCT images not seen before for automatic bone segmentation. Therefore, we offer a free USCT dataset. In addition, the proposed model is implemented on both the CPU and the GPU, hence overcoming previous works by a value of 97.38% and 96% for training and validation and achieving high segmentation accuracy for testing with a small error of 0.006, in a short time process. The suggested method demonstrates its ability to augment USCT data and then to automatically segment USCT bone structures achieving excellent accuracy outperforming the state of the art.

     

A new deep learning pipeline to detect Covid-19 on chest X-ray images using local binary pattern,dual tree complex wavelet transform and convolutional neural networks

In this study, which aims at early diagnosis of Covid-19 disease using X-ray images, the deep-learning approach, a state-of-the-art artificial intelligence method, was used, and automatic classification of images was performed using convolutional neural networks (CNN). In the first training-test data set used in the study, there were 230 X-ray images, of which 150 were Covid-19 and 80 were non-Covid-19, while in the second training-test data set there were 476 X-ray images, of which 150 were Covid-19 and 326 were non-Covid-19. Thus, classification results have been provided for two data sets, containing predominantly Covid-19 images and predominantly non-Covid-19 images, respectively. In the study, a 23-layer CNN architecture and a 54-layer CNN architecture were developed. Within the scope of the study, the results were obtained using chest X-ray images directly in the training-test procedures and the sub-band images obtained by applying dual tree complex wavelet transform (DT-CWT) to the above-mentioned images. The same experiments were repeated using images obtained by applying local binary pattern (LBP) to the chest X-ray images. Within the scope of the study, four new result generation pipeline algorithms having been put forward additionally, it was ensured that the experimental results were combined and the success of the study was improved. In the experiments carried out in this study, the training sessions were carried out using the k-fold cross validation method. Here the k value was chosen as 23 for the first and second training-test data sets. Considering the average highest results of the experiments performed within the scope of the study, the values of sensitivity, specificity, accuracy, F-1 score, and area under the receiver operating characteristic curve (AUC) for the first training-test data set were 0,9947, 0,9800, 0,9843, 0,9881 and 0,9990 respectively; while for the second training-test data set, they were 0,9920, 0,9939, 0,9891, 0,9828 and 0,9991; respectively. Within the scope of the study, finally, all the images were combined and the training and testing processes were repeated for a total of 556 X-ray images comprising 150 Covid-19 images and 406 non-Covid-19 images, by applying 2-fold cross. In this context, the average highest values of sensitivity, specificity, accuracy, F-1 score, and AUC for this last training-test data set were found to be 0,9760, 1,0000, 0,9906, 0,9823 and 0,9997; respectively.

     

基于改进深度度量学习算法的表面缺陷检测

为了解决小批量、多品种工业产品的表面质量检测问题,提出一种基于改进深度度量学习的缺陷检测算法.该算法对VGG16网络模型做改进,更有利于原始图像的隐空间映射.针对产品表面缺陷检测的任务,提出条件三元组损失函数以加强神经网络的拟合能力.同时,在隐空间中进行缺陷判定时,抛弃原始度量学习中基于KNN算法的归类方法,提出基于高...     

Efficient edge detection in digital images using a cellular neural network optimized by differential evolution algorithm

A cellular neural network (CNN) based edge detector optimized by differential evolution (DE) algorithm is presented. Cloning template of the proposed CNN is adaptively tuned by using simple training images. The performance of the proposed edge detector is evaluated on different test images and compared with popular edge detectors from the literature. Simulation results indicate that the proposed CNN operator outperforms competing edge detectors and offers superior performance in edge detection in digital images.     

Land-use scene classification based on a CNN using a constrained extreme learning machine

ABSTRACT

Classifying land-use scenes from high-resolution remote-sensing imagery with high quality and accuracy is of paramount interest for science and land management applications. In this article, we proposed a new model for land-use scene classification by integrating the recent success of convolutional neural network (CNN) and constrained extreme learning machine (CELM). In the model, the fully connected layers of a pretrained CNN have been removed. Then, CNN works as a deep and robust convolutional feature extractor. After normalization, deep convolutional features are fed to the CELM classifier. To analyse the performance, the proposed method has been evaluated on two challenging high-resolution data sets: (1) the aerial image data set consisting of 30 different aerial scene categories with sub-metre resolution and (2) a Sydney data set that is a large high spatial resolution satellite image. Experimental results show that the CNN-CELM model improves the generalization ability and reduces the training time compared to state-of-the-art methods.     

基于卷积神经网络的PCB缺陷检测

印刷电路板(PCB)在生产制造中由于生产工序等问题易导致电路板存在瑕疵缺陷,为提高对电路板缺陷的检测效率,提出了一种基于卷积神经网络(CNN)的电路板缺陷检测网络.该检测网络基于YOLO v4网络进行优化改造,针对于PCB制作精密、复杂,各类缺陷不易检测的难点,在优化后的网络中加入了基于细粒度空间域的长距离全局注意力机...