Traditional Detection Methods Research Articles

User Engagement Detection-Based Financial Technology Advertising Video Effectiveness Evaluation

With the rapid advancement of financial technology, an increasing number of related advertisements have received widespread attention. User engagement detection during the advertisement viewing process directly reflects the effectiveness of the advertising video. Therefore, detecting user engagement during the advertisement viewing process has become a crucial issue. However, traditional engagement detection methods often require significant computational resources, significantly reducing their practicality. To address this issue, the authors propose a method to effectively detect user engagement by fully integrating multiple relatively practical models. Specifically, the authors extract key frame images from user face video and perform super-resolution reconstruction of them. Then image pyramid matching is used to achieve user engagement detection. Finally, the authors establish a reasonable database and conduct sufficient experiments based on it. Experimental results demonstrate that this proposed method has realistic engagement detection accuracy, and the design of multiple steps is also valid.

Wearable cassette for paper-based aptasensor designed to detection of illicit drug; ketamine- a wearable cassette prototype comparison study

The advent of 3D printing technology has spurred innovation, particularly in healthcare and biosensing. One notable application is the creation of wearable biosensors for detecting substances like ketamine, a potent anesthetic and pain reliever with medical and recreational uses. Monitoring ketamine levels is crucial due to potential misuse and health risks. Utilizing 3D printing, manufacturers can produce intricate and customizable wearable biosensors designed for ketamine detection. This flexibility permits the incorporation of various sensor types, enhancing accuracy. Traditional detection methods are often cumbersome, making 3D printing a transformative tool for real-time monitoring. The application of 3D printing in wearable biosensors has the potential to revolutionize personalized healthcare, ensuring the safe and effective usage of ketamine. In this paper 3D printed paper-based wearable aptamer cassette (3DP-PWC) has been developed by immobilizing Ketamine Aptamer on ZnO-NPs electrodes. Electrochemical techniques such as cyclic voltammetry (CV), linear sweep voltammetry (LSV) and electrochemical impedance spectroscopy (EIS) were employed for validating results. The sensor’s versatility was demonstrated across beverages encompassing both alcoholic and non-alcoholic options. Two prototypes—a bracelet and a pendant—were developed and compared, showing promising results. Here, we reported a 3D-printing paper based wearable aptasensor for the ketamine detection. This pioneering developed sensor showed a low limit detection (LOD) of 0.01 μg/mL (lower than the physiological detection threshold 0.084 μg/mL) with linear-range was between 0.01 and 5 μmL and an optimal response time of 25 s.

Advancing Phishing Email Detection: A Comparative Study of Deep Learning Models.

Phishing is one of the most dangerous attacks targeting individuals, organizations, and nations. Although many traditional methods for email phishing detection exist, there is a need to improve accuracy and reduce false-positive rates. Our work investigates one-dimensional CNN-based models (1D-CNNPD) to detect phishing emails in order to address these challenges. Additionally, further improvement is achieved with the augmentation of the base 1D-CNNPD model with recurrent layers, namely, LSTM, Bi-LSTM, GRU, and Bi-GRU, and experimented with the four resulting models. Two benchmark datasets were used to evaluate the performance of our models: Phishing Corpus and Spam Assassin. Our results indicate that, in general, the augmentations improve the performance of the 1D-CNNPD base model. Specifically, the 1D-CNNPD with Bi-GRU yields the best results. Overall, the performance of our models is comparable to the state of the art of CNN-based phishing email detection. The Advanced 1D-CNNPD with Leaky ReLU and Bi-GRU achieved 100% precision, 99.68% accuracy, an F1 score of 99.66%, and a recall of 99.32%. We observe that increasing model depth typically leads to an initial performance improvement, succeeded by a decline. In conclusion, this study highlights the effectiveness of augmented 1D-CNNPD models in detecting phishing emails with improved accuracy. The reported performance measure values indicate the potential of these models in advancing the implementation of cybersecurity solutions to combat email phishing attacks.

Advancements and Clinical Significance of Breast Cancer Exosome Detection

Exosomes, which are lipid bilayer vesicles actively secreted by cells with a diameter of 30-100 nm, play a crucial role in the occurrence, development, and formation of the tumor microenvironment in breast cancer. They have become an important component of liquid biopsies for breast cancer and ideal targets for early diagnosis, treatment assessment, and clinical therapy. Breast cancer is the most common malignancy among women. However, traditional exosome detection methods have several limitations, and there are still many difficulties and challenges in achieving widespread clinical application of exosomes. Nowadays, significant progress has been made in exosome detection and analysis techniques, including optical methods such as colorimetry, fluorescence, surface-enhanced Raman scattering, as well as various electrochemical and thermal signal methods. This review summarizes the current biomarkers and novel developments in detection techniques applied to liquid biopsy of exosomes in breast cancer.

Abstract 2304: Detection of early-disseminated cancer cells with deep learning-enabled holographic imaging

Abstract Recent research increasingly validates the early dissemination of tumor cells in peripheral blood during the nascent stages of cancer, often preceding clinical identification of the primary tumor. This emerging evidence emphasizes the potential of Circulating Tumor Cells (CTCs) as early indicators for cancer. Despite their significance, CTCs are exceptionally rare compared to normal blood cells, which presents a considerable challenge in detection, particularly during early stages of cancer. Traditional detection methods, such as those based on nucleated cell enrichment, antibody labeling and fluorescence imaging, encounter notable limitations when applied in the context of early detection. These limitations arise from the inherent heterogeneity of CTCs, and the multistep process involved in labeling and manual enumeration, compromising cell viability and resulting in cell loss. This loss becomes particularly problematic in the context of early cancer detection scenarios where the tumor burden is extremely low, and any cell loss can significantly impact assay sensitivity. To overcome these challenges, we propose an innovative CTC detection approach employing a deep learning framework combined with holographic imaging. This technique utilizes coherent light to generate holograms from cell samples, capturing detailed 3D morphological and optical properties of individual cells. We developed a custom convolutional neural network designed to facilitate high-throughput cell sorting in real-time. Trained on healthy blood samples and diverse cancer cell lines, the dataset consists of more than 50 million cellular holograms enhanced by proprietary Gaussian annotation and image augmentation strategies. Moreover, by combining the morphological signatures from holography with antigen signatures (e.g. HER2/ER/PR) in real-time, our model demonstrates the ability to detect less than 1 false positive per 1 million nucleated cells, highlighting its potential as a robust tool for targeted early-stage breast cancer detection. Our technology represents highly sensitive alternative to traditional screening methods, with the promise of enhancing patient outcomes. Citation Format: Nicholas Heller, Kevin Mallery, Nathaniel Bristow, Yash Travadi, Eng Hock Lee, Jiarong Hong. Detection of early-disseminated cancer cells with deep learning-enabled holographic imaging [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2304.

Microbiomes detected by cerebrospinal fluid metagenomic next-generation sequencing among patients with and without HIV with suspected central nervous system infection.

Opportunistic infections in the central nervous system (CNS) can be a serious threat to people living with HIV. Early aetiological diagnosis and targeted treatment are crucial but difficult. Metagenomic next-generation sequencing (mNGS) has significant advantages over traditional detection methods. However, differences in the cerebrospinal fluid (CSF) microbiome profiles of patients living with and without HIV with suspected CNS infections using mNGS and conventional testing methods have not yet been adequately evaluated. We conducted a retrospective cohort study in the first hospital of Changsha between January 2019 and June 2022 to investigate the microbiomes detected using mNGS of the CSF of patients living with and without HIV with suspected CNS infections. The pathogens causing CNS infections were concurrently identified using both mNGS and traditional detection methods. The spectrum of pathogens identified was compared between the two groups. Overall, 173 patients (140 with and 33 without HIV) with suspected CNS infection were enrolled in our study. In total, 106 (75.7%) patients with and 16 (48.5%) patients without HIV tested positive with mNGS (p = 0.002). Among the enrolled patients, 71 (50.7%) with HIV and five (15.2%) without HIV tested positive for two or more pathogens (p < 0.001). Patients with HIV had significantly higher proportions of fungus (20.7% vs. 3.0%, p = 0.016) and DNA virus (59.3% vs. 21.2%, p < 0.001) than those without HIV. Epstein-Barr virus (33.6%) was the most commonly identified potential pathogen in the CSF of patients living with HIV using mNGS, followed by cytomegalovirus (20.7%) and torque teno virus (13.8%). The top three causative pathogens identified in patients without HIV were Streptococcus (18.2%), Epstein-Barr virus (12.1%), and Mycobacterium tuberculosis (9.1%). In total, 113 patients living with HIV were diagnosed as having CNS infections. The rate of pathogen detection in people living with HIV with a CNS infection was significantly higher with mNGS than with conventional methods (93.8% vs. 15.0%, p < 0.001). CSF microbiome profiles differ between patients living with and without HIV with suspected CNS infection. mNGS is a powerful tool for the diagnosis of CNS infection among people living with HIV, especially in those with mixed infections.

Rap-Densenet Framework for Network Attack Detection and Classification

Cybersecurity is becoming increasingly important with the rise in Internet usage. The two most frequent cyberattacks that can seriously harm a website or a server and render them inaccessible to other customers are denial of service (DoS) and distributed denial of service (DDoS) attacks. These attacks are so common and take many different forms, it is difficult to identify and respond to them with previous methods. Furthermore, computational complexity, inconsistency, and irrelevant data are problems for traditional intrusion detection methods. As a result, a powerful deep learning-based technique is applied in this study for the identification and categorisation of DoS and DDoS attacks. Refined Attention Pyramid Network (RAPNet)-based feature extraction is used in this proposed framework to extract features from the input data. Then, Binary Pigeon Optimisation Algorithm (BPOA) is used to determine the best features. After choosing optimal characteristics, Densenet201-based deep learning is deployed to categorise the assaults in Bot-IoT, CICIDS2017, and CICIDS2019 datasets. Furthermore, the Conditional Generative Adversarial Network (CGAN) is used to provide extra data samples for minority classes to address the issue of imbalanced data. The findings show that the proposed model can precisely identify and categorise DoS and DDoS assaults in comparison to the existing intrusion detection approaches with 99.43%, 99.26%, and 99.38% accuracy for CICIDS2019, CICIDS2017, and BoT-IoT correspondingly.

FCFormer: fish density estimation and counting in recirculating aquaculture system

In intelligent feeding recirculating aquaculture system, accurately estimating fish population and density is pivotal for management practices and survival rate assessments. However, challenges arise due to mutual occlusion among fish, rapid movement, and complex breeding environments. Traditional object detection methods based on convolutional neural networks (CNN) often fall short in fully addressing the detection demands for fish schools, especially for distant and small targets. In this regard, we introduce a detection framework dubbed FCFormer (Fish Count Transformer). Specifically, the Twins-SVT backbone network is employed first to extract global features of fish schools. To further enhance feature extraction, especially in the fusion of features at different levels, a Bi-FPN aggregation network model with a CAM Count module is incorporated (BiCC). The CAM module aids in focusing more on critical region features, thus rendering feature fusion more cohesive and effective. Furthermore, to precisely predict density maps and elevate the accuracy of fish counting, we devised an adaptive feature fusion regression head: CRMHead. This approach not only optimizes the feature fusion process but also ensures superior counting precision. Experimental results shown that the proposed FCFormer network achieves an accuracy of 97.06%, with a mean absolute error (MAE) of 6.37 and a root mean square error (MSE) of 8.69. Compared to the Twins transformer, there's a 2.02% improvement, outperforming other transformer-based architectures like CCTrans and DM_Count. The presented FCFormer algorithm can be effectively applied to fish density detection in intelligent feeding recirculating aquaculture system, offering valuable input for the development of intelligent breeding management systems.

A Real-Time Spoofing Detection Method Using Three Low-Cost Antennas in Satellite Navigation

The vulnerability of civil receivers of the Global Satellite Navigation System (GNSS) to spoofing jamming has raised significant concerns in recent times. Traditional multi-antenna spoofing detection methods are limited in application scenarios and come with high hardware costs. To address this issue, this paper proposes a novel GNSS spoofing detection method utilizing three low-cost collinear antennas. By leveraging the collinearity information of the antennas, this method effectively constrains the observation equation, leading to improved estimation accuracy of the pointing vector. Furthermore, by employing a binary statistical detection model based on the sum of squares (SSE) between the observed value and the estimated value of the pointing vector, real-time spoofing signal detection is enabled. Simulation results confirm the efficacy of the proposed statistical model, with the error of the skewness coefficient not exceeding 0.026. Experimental results further demonstrate that the collinear antenna-based method reduces the standard deviation of the angle deviation of the pointing vector by over 55.62% in the presence of spoofing signals. Moreover, the experiments indicate that with a 1 m baseline, this method achieves 100% spoofing detection.

An image‐based runway detection method for fixed‐wing aircraft based on deep neural network

AbstractVisual information is important in final approach and landing phases for an approaching aircraft, it presents supplementary source for navigation system, and provides backup guidance when radio navigation fails, or even supports a complete vision‐based landing. Relative position and attitude can be solved from the runway features in the image. Traditional runway detection methods have high latency and low accuracy, which is unable to satisfy the requirements for a safe landing. This paper proposes a real‐time runway detection model, efficient runway feature extractor (ERFE), based on deep convolutional neural network, generating semantic segmentation and feature lines output. In order to evaluate the model's effectiveness, a benchmark is proposed to calculate the actual error between predicted feature line and ground truth one. A novel runway dataset which is based on pictures from Microsoft Flight Simulator 2020 (FS2020), is also proposed in this paper to train and test the model. The dataset will be released at https://www.kaggle.com/datasets/relufrank/fs2020‐runway‐dataset. ERFE shows excellent performance in FS2020 dataset, it gives satisfactory results even for real runway images excluded from our dataset.

Comparing Microcrystal Electron Diffraction (MicroED) and X-ray crystallography as methods for structure determination of Oseltamivir phosphate

Oseltamivir phosphate is a pharmaceutical used in the treatment of influenza viruses. Panče Naumov et al. took two years of dedicated single crystal cultivation and successfully obtained low-quality single crystals through a diffusion method. Cultivation of Oseltamivir phosphate single crystals was proved to be difficult and can extend over several months or more. We employed microcrystalline electron diffraction (MicroED) for direct powder measurements, eliminating the need for lengthy single-crystal cultivation, and the time to obtain the molecular structure of Oseltamivir phosphate was shortened from several months to five hours. Comparative analyses of single crystal (Crystal 1) and MicroED (Crystal 2) results revealed equivalent crystal structures, with minimal differences in hydrogen bond distances, identical powder X-ray diffraction patterns, and negligible variations in Hirshfeld surface analysis. The differential scanning calorimetry (DSC) analysis showcased a higher melting point of Oseltamivir phosphate (206.0 °C) than Oseltamivir (110.7 °C), indicating that Oseltamivir phosphate had high temperature stability, and may result in the selection of Oseltamivir phosphate as a formulation form. The study establishes and verifies a congruence between MicroED detection and traditional single-crystal detection methods and may be applied in the elucidation of compound structures and the surveillance of drug polymorphism.

On-chip bioluminescence biosensor for the detection of microbial surface contamination

Detection of microbial pathogens is important for food safety reasons, and for monitoring sanitation in laboratory environments and health care settings. Traditional detection methods such as culture-based and nucleic acid-based methods are time-consuming, laborious, and require expensive laboratory equipment. Recently, ATP-based bioluminescence methods were developed to assess surface contamination, with commercial products available. In this study, we introduce a biosensor based on a CMOS image sensor for ATP-mediated chemiluminescence detection. The original lens and IR filter were removed from the CMOS sensor revealing a 12 MP periodic microlens/pixel array on an area of 6.5 mm × 3.6 mm. UltraSnap swabs are used to collect samples from solid surfaces including personal electronic devices, and office and laboratory equipment. Samples mixed with chemiluminescence reagents were placed directly on the surface of the image sensor. Close proximity of the sample to the photodiode array leads to high photon collection efficiency. The population of microorganisms can be assessed and quantified by analyzing the intensity of measured chemiluminescence. We report a linear range and limit of detection for measuring ATP in UltraSnap buffer of 10–1000 nM and 225 fmol, respectively. The performance of the CMOS-based device was compared to a commercial luminometer, and a high correlation with a Pearson's correlation coefficient of 0.98589 was obtained. The Bland-Altman plot showed no significant bias between the results of the two methods. Finally, microbial contamination of different surfaces was analyzed with both methods, and the CMOS biosensor exhibited the same trend as the commercial luminometer.

Bi2WO6/TiO2-based visible light-driven photoelectrochemical enzyme biosensor for glucose measurement.

Nowadays, the frequent occurrence of food adulteration makes glucose detection particularly important in food safety and quality management. The quality and taste of honey are closely related to the glucose content. However, due to the drawbacks of expensive equipment, complex operating procedures, and time-consuming processes, the application scope of traditional glucose detection methods is limited. Hence, this study developed a photoelectric chemical (PEC) sensor, which is composed of a photoactive material of bismuth tungstate (Bi2WO6) with titanium dioxide (TiO2) and glucose oxidase (GOD), for simple and rapid detection of glucose. Notably, the composites' absorption prominently increased in the visible light region, and the photo-generated electron-hole pairs were efficiently separated by virtue of the unique nanostructure system, thus playing a crucial role in facilitating PEC activity. In the presence of dissolved oxygen, the photocurrent intensity was enhanced by H2O2 generated from glucose under electro-oxidation specifically catalyzed by GOD fixed on the modified electrode. When the working potential was 0.3V, the changes of photocurrent response indicated that the PEC enzyme biosensor providesalow detection limit (3.8µM), and a wide linear range (0.008-8mM). This method has better selectivity in honey samples and broad application prospects in clinical diagnosis for future.

SNDMI: Spyware network traffic detection method based on inducement operations

Spyware is a type of malware widely used by attackers to steal confidential information from users without their knowledge. It has the characteristics of high latency, high stealth, and high threat. Spyware typically uses trigger-based tactics to implement its malicious behavior, making it a challenge for traditional host-based detection methods and traditional network traffic-based detection methods to implement highly accurate and efficient spyware detection. To address this problem, this article focuses on the network behavior of spyware, analyzes the inducibility of spyware network traffic, and further proposes a spyware network traffic detection method based on inducement operations (SNDMI). SNDMI can cause spyware to generate more network traffic for detection by using inducement operations in the constructed inducement operation set. SNDMI extracts three types of network traffic features: distributional features, statistical features, and trend features, and uses a genetic algorithm to select features. The experimental results show that SNDMI with LightGBM achieves an accuracy of 98.98% and a false positive rate of 0.45% on the traffic of hundreds of benign and spyware samples.

DEM-based topographic change detection considering the spatial distribution of errors

ABSTRACT Digital Elevation Model (DEM) errors tend to be spatially correlated, inevitably affecting DEM-based topographic change detection. Traditional topographic change detection methods often ignore the spatial distribution of the DEM error. This paper aims to develop a workflow that considers the spatial autocorrelation of the error in topographic change detection. Firstly, the DEM of Difference (DoD) is obtained from two-period DEMs, and the Monte Carlo method is employed to evaluate the Spatially Distributed Errors (SDE) in DEMs. Secondly, DoD errors are calculated by propagation based on spatially distributed DEM errors. At the same time, its spatial distribution is quantified using the semi-variance function. Finally, topographic changes (erosion, deposition, and net changes) are calculated based on the spatial distribution analysis and significance detection. The results in two small catchments indicate that DEM errors are spatially correlated, increasing the volume calculation errors. However, using Standard Deviation of Errors (SDE) instead of Root Mean Square Error (RMSE) can effectively reduce the sensitivity of the detection results in the significance threshold. When the significance threshold increases from 68% to 95%, the observations loss using the spatially distributed error is 4.67% −6.92% lower than that using the RMSE. The level of detection has little impact on the net topographic change and significantly influences gross erosion and deposition. In particular, the use of level of detection can effectively reduce the misclassification of erosion or deposition in stable topography areas. The proposed method can be effectively utilized in various applications like surface deformation monitoring, erosion monitoring, and sediment transport assessment.

A Transfer Hydrogenation Approach to Activity-Based Sensing of Formate in Living Cells.

Formate is a major reactive carbon species in one-carbon metabolism, where it serves as an endogenous precursor for amino acid and nucleic acid biosynthesis and a cellular source of NAD(P)H. On the other hand, aberrant elevations in cellular formate are connected to progression of serious diseases, including cancer and Alzheimer's disease. Traditional methods for formate detection in biological environments often rely on sample destruction or extensive processing, resulting in a loss of spatiotemporal information. To help address these limitations, here we present the design, synthesis, and biological evaluation of a first-generation activity-based sensing system for live-cell formate imaging that relies on iridium-mediated transfer hydrogenation chemistry. Formate facilitates an aldehyde-to-alcohol conversion on various fluorophore scaffolds to enable fluorescence detection of this one-carbon unit, including through a two-color ratiometric response with internal calibration. The resulting two-component probe system can detect changes in formate levels in living cells with a high selectivity over potentially competing biological analytes. Moreover, this activity-based sensing system can visualize changes in endogenous formate fluxes through alterations of one-carbon pathways in cell-based models of human colon cancer, presaging the potential utility of this chemical approach to probe the continuum between one-carbon metabolism and signaling in cancer and other diseases.

Study on breast cancer image detection and classification based on residual connected convolutional neural network (CNN)

This paper presents a method for breast cancer image detection and classification based on convolutional neural networks. While traditional breast cancer detection methods carry the risk of subjectivity and misclassification, machine learning can improve the accuracy and efficiency of detection by training algorithms to automatically identify breast cancer lesions in patients. The dataset of this paper includes 7909 microscopic images of breast tumor tissues with different magnifications collected from 82 patients, which are divided into two categories: benign and malignant tumors. In this paper, residual connection is designed for convolutional neural networks, which is a cross-layer connection method that can effectively solve the gradient vanishing and gradient explosion problems in deep neural networks, and at the same time, improve the generalization ability and training speed of the model. In this paper, the dataset is divided into training set, validation set and test set according to the ratio of 6:2:2. Through training and validation, the prediction accuracy of the obtained model on the test set reaches 85.5%, which achieves good prediction results in breast cancer image detection and classification. Eventually, the model's loss converged at 0.53 and the AUC was 0.832. The research results in this paper are of great significance for the early diagnosis and treatment of breast cancer. Automatic identification and classification of breast cancer images by machine learning algorithms can reduce the subjective judgment of doctors, improve the accuracy and efficiency of detection, and provide important support for the early diagnosis and treatment of breast cancer. Meanwhile, the research method and results of this paper also provide reference and reference for image detection and classification of other malignant tumors.

Automatic Detection of Cast Billet Dendrite Based on Improved Hough Transform

Primary dendrite information is one of the most important metrics to measure the quality of continuous cast slabs. The contrast of low magnification images is very low under the influence of illumination and sampling devices, so the traditional dendrite detection method has the problem of missed detections. We propose an automatic dendrite detection method based on an improved Hough transform, which effectively improves the accuracy and efficiency of primary dendrite detection. By using the local grayscale features of the image, a genetic algorithm-based local contrast enhancement algorithm is proposed. Compared with the traditional contrast enhancement algorithm, it can retain all the information of the dendrites. Combined with the image binarization method based on Hessian matrix, we can obtain more detailed information about the dendrites. According to the continuity and solidification characteristics of dendrites, the Hough transform is improved to extract dendrite information, which effectively reduces the computational cost of the Hough transform. The experimental results show that the method of this paper has versatility, and the error is four pixels compared with the manual method, which can provide a reliable basis for the subsequent judgement of the quality of cast billets.

CD-MAE: Contrastive Dual-Masked Autoencoder Pre-Training Model for PCB CT Image Element Segmentation

Element detection is an important step in the process of the non-destructive testing of printed circuit boards (PCB) based on computed tomography (CT). Compared with the traditional manual detection method, the image semantic segmentation method based on deep learning greatly improves efficiency and accuracy. However, semantic segmentation models often require a large amount of data for supervised training to generalize better model performance. Unlike natural images, the PCB CT image annotation task is more time-consuming and laborious than the semantic segmentation task. In order to reduce the cost of labeling and improve the ability of the model to utilize unlabeled data, unsupervised pre-training is a very reasonable and necessary choice. The masked image reconstruction model represented by a masked autoencoder is pre-trained on the unlabeled data, learning a strong feature representation ability by recovering the masked image, and shows a good generalization ability in various downstream tasks. In the PCB CT image element segmentation task, considering the characteristics of the image, it is necessary to use a model with strong feature robustness in the pre-training stage to realize the representation learning on a large number of unlabeled PCB CT images. Based on the above purposes, we proposed a contrastive dual-masked autoencoder (CD-MAE) pre-training model, which can learn more robust feature representation on unlabeled PCB CT images. Our experiments show that the CD-MAE outperforms the baseline model and fully supervised models in the PCB CT element segmentation task.

Intelligent detection of foreign objects over coal flow based on improved GANomaly

The mining belt conveyor is one of the most important modules in coal mine, whose safety always be threatened by the foreign objects. Although the traditional target detection methods achieve promising results in various computer vision tasks, the performance heavily depends on sufficient labelled data. However, in real-world production scenario, it is difficult to acquire huge number of images with foreign objects. The obtained datasets lacking of capacity and diversity are not suitable for training supervised learning-based foreign objects detection models. To address this concern, we propose a novel method for detecting the foreign objects on the surface of underground coal conveyor belt via improved GANomaly. The proposed foreign objects detection method employs generative adversarial networks (GAN) with attention gate to capture the distribution of normality in both high-dimensional image space and low-dimensional latent vector space. Only the normal images without foreign object are utilized to adversarially train the proposed network, including a U-shape generator to reconstruct the input image and a discriminator to classify real images from reconstructed ones. Then the combination of the difference between the input and generated images as well as the difference between latent representations are utilized as the anomaly score to evaluate whether the input image contain foreign objects. Experimental results over 707 images from real-world industrial scenarios demonstrate that the proposed method achieves an area under the receiver operating characteristic curve of 0.864 and is superior to the previous GAN-based anomaly detection methods.