Search

In industry, the task of defect classification and defect localization is an important part of defect detection system. However, existing studies only focus on one task and it is difficult to ensure the accuracy of both tasks. This paper proposes a defect detection system based on improved Yolo_v4, which greatly improves the detection ability of minor defects. For K_Means algorithm clustering prianchors question with strong subjectivity, the paper proposes the Density Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm to determine the number of Anchors. To solve the problem of low detection rate of small targets caused by insufficient reuse rate of low-level features in CSPDarknet53 feature extraction network, this paper proposes an ECA-DenseNet-BC-121 feature extraction network to improve it. And the Dual Channel Feature Enhancement (DCFE) module is proposed to improve the local information loss and gradient propagation obstruction caused by quad chain convolution in PANet networks to improve the robustness of the model. The experimental results on the fabric surface defect detection datasets show that the mAP of the improved Yolo_v4 is 98.97%, which is 7.67% higher than SSD, 3.75% higher than Faster_RCNN, 10.82% higher than Yolo_v4 tiny, and 5.35% higher than Yolo_v4, and the detection speed reaches 39.4 fps. It can meet the real-time monitoring needs of industrial sites. [ABSTRACT FROM AUTHOR]

Fabric defect detection is an important part of the textile industry to ensure product quality. To solve the problems such as the difficulty of detecting small defects and the coexistence of multi-scale defects in fabric defect detection, a fabric defect detection method based on the feature purification fusion structure is proposed in this paper. Specifically, we improve the feature extraction network to enhance the network's ability to focus on small defective features and effectively reduce the model parameters. The existing methods use direct fusion between multi-level feature maps, which will lead to feature confusion. Therefore, we propose the feature purification fusion structure (FPF), which includes the semantic information supplementation strategy (SIS) and the detail information supplementation strategy (DIS). SIS extracts valid information from the deep feature map and supplements it to the shallow feature map, weakening feature information irrelevant to the shallow feature map detection task. DIS adaptively supplements the feature information required by the deep feature map detection task from the shallow feature map. FPF improves the ability of the network to detect small defects and effectively mitigates the aliasing effect generated during feature fusion. The experimental results show that compared to the baseline model YOLOv5s algorithm, our model achieved a 6.8% improvement in detection accuracy, with an average inference frame rate of 37.6 FPS, demonstrating better detection performance in fabric defect detection. Furthermore, extending this model to aluminum profile defect datasets also demonstrates strong performance. [ABSTRACT FROM AUTHOR]

3D point cloud registration is a basic task in computer vision. In recent years, various of learning-based methods have been proposed to solve this problem. These methods effectively overcome the original problem of over-reliance on initial conditions and enhance the ability of obtaining the corresponding relationship. However, few methods pay enough attention to local features and tend to cause some mismatches. Therefore, this paper proposes two networks to extract local features sufficiently. To obtain a more accurate correspondence relationship between the point clouds, we propose a feature weight allocation network (FWANet), in which the expression ability of feature is enhanced using the proposed significant feature extraction module. Besides that, we utilize an interference elimination module to remove the interference points and enhance the internal correlation of point clouds. We also propose a spatial structural generation network (SSGNet), which fully utilizes the spatial location information to determine the spatial correspondence and generate a reliable connection after concatenating multi-dimensional features. At last, a complete feature space can be effectively captured after combining our FWANet with SSGNet together. We conducted multiple experiments on ModelNet40 datasets and achieved excellent results. Experimental results on four types of data demonstrate the superiority of our algorithm against the state-of-the-art ones. Our code will be available at https://github.com/liu-zikang/registration as soon as the paper is accepted. [ABSTRACT FROM AUTHOR]

The geometric distortion in panoramic images significantly mediates the performance of saliency detection method based on traditional CNN. The strategy of dynamically expanding convolution kernel can achieve good results, but it also produces a lot of computational overhead in the process of reading the adjacency list, which decreases the computational efficiency. The appearance of graph convolution provides a new way to solve such problems. Although using graph convolution can effectively extract the structural features of the graph, it reduces the accuracy of the model resulting from ignoring the spatial features of the image signal. To this end, this paper proposes a construction method of the multi-scale graph structure of the panoramic image and a panoramic image saliency detection model composed of an image saliency feature extraction network and multi-scale saliency feature fusion network combining the image structure information and spatial information in the panoramic image. First, we establish a graph structure consisting of root and leaf nodes obtained by super-pixel segmentation at different scales and spherical Fibonacci sampling, respectively. Then, a feature extraction network composed of two graph convolution layers and two one-dimensional auto-encoders with the same parameterization is used to extract the salient features of the multi-scale graph structure. Finally, the U-Net network fuses the multi-scale saliency features to get the final saliency map. The results show that the proposed model performs better than the state-of-the-art models in terms of calculation speed and accuracy. [ABSTRACT FROM AUTHOR]

In the process of facial expression recognition, face detection is the prerequisite, image preprocessing is the foundation, facial expression feature extraction is the key, and facial expression classification is the target. Effective feature extraction in this process can improve the accuracy of facial expression classifications. On the other hand, traditional facial expression recognition methods are not only complicated in the feature extraction process, but also unable to obtain more in-depth high-semantic features and deep features from the original image. To solve the above problems, this paper proposes a facial expression recognition method based on multi-channel fusion and lightweight neural network. First, a cascade classifier based on Haar features is used to detect the face region of the facial expression image. Second, the local binary pattern (LBP) is used to extract the local texture features from the face region. Third, face edge features are simultaneously obtained by performing edge detection in the face region based on the Canny edge detection algorithm. Fourth, the obtained face image, LBP texture feature image, and edge detection Canny image are fused, and the fused image is input into the constructed lightweight neural network for training and recognition. Experiments are carried out on the public image databases Facial Expression Recognition 2013 (Fer2013) and extended Cohn–Kanade (CK +) using the hold-out cross-validation method. The experimental results show that the proposed method effectively extracts more complete image features by combining traditional feature extraction algorithms with deep learning feature extraction algorithms, improves the accuracy and robustness of facial expression recognition, and has better recognition rate and generalization ability compared to other mainstream methods. [ABSTRACT FROM AUTHOR]

Image segmentation can be used to monitor some unconventional operations to ensure the quality of the assembled product in the product assembly process. The current supervised image segmentation method requires much time for labeling. And, the assembly segmentation feature is difficult to extract from color images, due to the influence of color, illumination, and texture factors. To solve these problems, this paper proposes a domain adaptive segmentation method for mechanical assembly with normal images as input. The method involves an adversarial learning stage and a self-training stage. The adversarial learning stage includes a segmentation network and a discriminator network. In the segmentation network, transformer and PPM layers are added to improve the segmentation accuracy of small parts. Meanwhile, the discriminator network is used to align the output features for the source and target data. In the self-training stage, the segmentation accuracy of the part edges is further improved by fine-tuning the parameter model. In addition, to ensure the effectiveness of assembly monitoring, this study establishes a domain adaptive semantic segmentation dataset based on normal image, and evaluate the proposed method. Compared with other domain adaptation methods, our two-stage training method has clearer edge processing performance, and the mIoU index reaches 83.67%, about 3.5% higher than the existing advanced method. In general, our network structure is more suitable for mechanical assembly segmentation and monitoring. [ABSTRACT FROM AUTHOR]

Surface defect detection of industrial products has become a promising area of research. Among the existing defect detection algorithms, most of the CNN-based methods can achieve the task of defect detection under ideal experimental conditions. However, the accuracy of defect detection is easily affected by the different lighting conditions of the environment and the inconsistency of the defect scale. Therefore, general deep learning methods have difficulties in solving the problem of defect detection in complex scenes. In this paper, a defect detection network based on low-light enhancement and multi-scale feature extraction (LE–MSFE–DDNet) is proposed. There are two blocks in the proposed network, including a low-light enhancement block and a SE-FP block. In the low-light enhancement block, the deep network is applied to enhance light adaptation of the deconstructed low-light feature map in our network. The influence of illumination inconsistency is weakened by the introduction of this block. In the SE-FP block, the dependencies between different channels are combined with the multi-scale feature extraction. The defects with different scales are accurately located through the combination of this block and our network. In addition, a Fine Cans Defect dataset based on the surface of fine cans is collected by this paper to verify the feasibility of the proposed network. The proposed model is compared with the state-of-the-art object detection network and the proposed method achieves 94.3% average accuracy on the Fine Cans Defect dataset. The experimental results show that the proposed method outperforms the state-of-the-art method for surface defect detection. [ABSTRACT FROM AUTHOR]

A feature selection method based on three-way decision is proposed to solve the problem of target recognition under high-dimension and few-shot data. The main research results of this paper include the following aspects: (1) aiming at the uncertainty of existing algorithm caused by uncertain data in high-dimension and few-shot data, based on the traditional filtering feature selection ReliefF algorithm, the algorithm is improved, according to the three-way decision thought, the single threshold of original algorithm is extended into two thresholds, to increase the rate of fault tolerance of the original algorithm, avoid features is too much or too little to eliminate. (2) To solve the problem of "dimension disaster" of high-dimension and few-shot data, the features are divided into three domains according to two thresholds and feature weights, and then the features in the boundary domain are further selected by combining the wrapper feature selection idea to reduce the dimension and obtain the optimal feature subset. Finally, the experimental results show that the recognition accuracy and F1 have been improved. Experiment 1 proved the fault tolerance rate and stability of the proposed algorithm, and Experiment 2 verified the advantages of the proposed algorithm in small sample conditions. [ABSTRACT FROM AUTHOR]

The main goal of person re-identification (ReID) is to identify human images captured by different cameras. However, people are often occluded by various obstacles. To solve the occlusion problem, this paper proposes a novel method that occluded person re-identification based on embedded graph matching network for contrastive feature relation. It involves three modules. In the contrast feature construction (C) module, we use the trained pose estimation (PE) model to extract the local features shown by the key points of person image. In addition, we propose the global discrepant contrastive pooling (GDCP) layer to capture global features after noise removal. At this point, the extracted features are only the node feature information,and what is not connected is a single one. In the adaptive node relationship (R) module, the node relation learning (NRL) layer is proposed to establish the contrast feature relationship between the individual information of feature extraction and the joint structure of monitoring the key points of the feature relationship, so as to automatically update the relationship between features, promote the transmission of meaningful features and inhibit the transmission of meaningless features. In this way, the relationship between the feature information can also be extracted. In the embedded graph matching (M) module, the embedded feature alignment (EFA) layer is proposed to directly predict the similarity score of each feature of the two images. In addition, the information collected from the non-occluded part of one image is used to compensate the occluded part corresponding to the other image, which reduces the dispersion of image information. We avoid using a sensitive one-to-one hard alignment instead of many-to-many soft alignment to enhance robustness. On Occluded_Duke, DukeMTMC and Market1501,this method is superior to the existing person re-identification methods. Particularly on Occluded_Duke, the Rank-1 reaches 60.5%, and mAP reaches 49.2%. [ABSTRACT FROM AUTHOR]

Low-light images usually contain high noise and low contrast. This brings bad visual feelings and hinders subsequent computer vision work. At present, many algorithms have been proposed to enhance low-light images. However, the existing methods still have some problems, such as insufficient enhancement, color distortion, or overexposure. In this paper, we propose a low-light image enhancement network based on the spatial pyramid to solve the problems existing in other methods, so as to make the enhancement result closer to the normal illumination image in brightness and color. The network is divided into two parts. Firstly, the decomposition network is designed based on Retinex theory, and the image is decomposed into the illumination image and reflection image. Then, the illumination image is processed through the three convolution kernels on the spatial pyramid module to obtain three sets of features with different scales. Next, we concatenate these three groups of features together. And the concatenated features are extracted through a convolution kernel to obtain the enhanced illumination image. Finally, the enhanced illumination image and the decomposed reflection image are multiplied pixel by pixel to obtain an enhanced image. In addition, we introduce a color loss function to solve the problem of color distortion. The experimental results show that the proposed algorithm has better visual feelings than other algorithms. We also calculate the peak signal-to-noise ratio, structural similarity index and average brightness of the enhanced results of different algorithms, and the results show that the proposed algorithm performs better. [ABSTRACT FROM AUTHOR]

Cloud detection is an important step in remote sensing image processing and a prerequisite for subsequent analysis and interpretation of remote sensing images. Traditional cloud detection methods are difficult to accurately detect clouds and snow with very similar features such as color and texture. In this paper, the features of cloud and snow in remote sensing images are deeply extracted, and an accurate cloud and snow detection method is proposed based on the advantages of Unet3+ network in feature fusion. Firstly, color space conversion is performed on remote sensing images, RGB images and HIS images are used as input of Unet3+ network. Resnet 50 is used to replace the Unet3+ feature extraction network to extract remote sensing image features at a deeper level, and add the Convolutional Block Attention Module in Resnet50 to improve the network's attention to cloud and snow. Finally, the weighted cross entropy loss is constructed to solve the problem of unbalanced sample number caused by high proportion of background area in the image. The results show that the proposed method has strong adaptability and moderate computation. The mPA value, mIoU value and mPrecision value can reach 92.76%, 81.74% and 86.49%, respectively. Compared with other algorithms, the proposed method can better eliminate all kinds of interference information in remote sensing images of different landforms and accurately detect cloud and snow in images. [ABSTRACT FROM AUTHOR]

Collaborative filtering (CF) is a prevailing technique utilized for recommendation systems and has been comprehensively explored to tackle the problem of information overload particularly in the Big Data context. The traditional CF algorithms are capable to perform adequately under various circumstances, nevertheless, there exist some shortcomings involving cold start and data sparsity. Moreover, a potential breakthrough rests in taking full advantage of any valuable semantic information contained in items. Therefore, for alleviating these defects, in this paper, we propose a two-stage collaborative filtering approach driven by Simhash-based semantic feature analysis, of which the first stage is Simhash-based semantic feature extraction for items and categories, and the second stage is reinforced CF rating prediction driven by intensely compressed category features. The rich semantic features of vast items and their categories can be rapidly extracted and compressed in the first stage by employing the Simhash, with being utilized to promote the traditional collaborative filtering processes. Besides, to solve the problems pertaining to the Big Data context, we design a parallel algorithm on Spark to accelerate the time-consuming process of semantic feature extraction for vast items. Finally, we conduct comprehensive experiments to validate the reinforced CF approach by adopting practical datasets, and the results reveal that compared with the traditional CF algorithms it can accomplish a promising performance. [ABSTRACT FROM AUTHOR]

As a key environment perception technology of autonomous driving or driver assistance systems, lane detection is to ensure vehicles to drive safely in corresponding lane. However, existing lane detection algorithms for two-lane detection focus on using various filtering methods to reduce the impact of useless information, resulting in low accuracy and low efficiency. In this paper, a novel Adaptive Region of Interest (A-ROI) extraction method is proposed to improve the accuracy and real-time performance of the two-lane detection algorithm. Three key technologies are introduced to solve the problems. First, A-ROI, which only focuses on the lane where the vehicle is located, is applied to the Bird's-Eye-View image obtained by using Inverse Perspective Mapping (IPM). Next, based on Bayesian framework and Likelihood models, a lane feature extraction method with a lane-like feature filter is used for edge detection. Finally, an improved Random Sample Consensus (RANSAC) algorithm is introduced by using a filter that can remove noisy lane data. The performance of the proposed A-ROI method together with the improved lane detection method is evaluated via simulation of various scenarios. Experimental results show the proposed method has better accuracy and real-time performance than the traditional lane detection methods. [ABSTRACT FROM AUTHOR]

Intelligence modeling interpretable for incipient fault diagnosis of batch process represents a serious challenge. For the multivariate, nonlinear, and high-dimensionality characteristics of process data, existing fault diagnosis solutions are easily concealed by noise while neglect the low amplitude and noise interference of the incipient faults. In this paper, an intelligent incipient fault diagnosis model based on deep time series feature extraction network is proposed, which integrates denoising autoencoder (DAE), layer normalization and dropout layer added LSTM (LD-LSTM), and stacked autoencoder (SAE) to obtain the efficient features. DAE is applied to restore the data and eliminate noise interference. LD-LSTM is utilized to extract time series features, in which the layer normalization and the dropout layer are added between the layers of the LSTM to solve the problems of slow model convergence and over-fitting. SAE performs the feature extraction to obtain deep time series features and learns more efficient expressions. The softmax function is applied to achieve the incipient fault diagnosis based on the deep time series features. A case study on a fed-batch penicillin fermentation process suggests that the proposed method can obtain an accuracy rate of 98.97%, which has increased an average of 20% than the traditional shallow machine learning. The accuracy of fault recognition and comparative experiments demonstrate the effectiveness and feasibility of this model, and provide a solution for the application of intelligent fault diagnosis in batch industries. [ABSTRACT FROM AUTHOR]