Deep Learning-based Detection Method Research Articles

Deep learning-based method for detection and feature quantification of microscopic cracks on the surface of concrete dams

Efficient detection of surface cracks in concrete dams is crucial for maintaining hydraulic engineering and infrastructure. Accurately identifying and diagnosing microscopic cracks in practical engineering remains challenging. An intelligent method based on an improved YOLOv8s-YOLO-DEW combined with Unet is proposed for microscopic crack detection and information quantification. In YOLO-DEW, designing the CSP Bottleneck with Dynamic Snake Convolutions (C2f_DSC) enhances the fusion of features at different scales, Efficient Multi-Scale Attention (EMA) is introduced to focus on the key information of fracture features, and employing WIoUv3 loss function improves crack localization capability. Subsequently, the Unet is constructed to extract morphological features of micro-fractures. The results indicate that the comprehensive performance of the proposed method is optimal, with 83.5 % mAP for YOLO-DEW, while Unet achieving mIoU and mPA of 80.73 % and 86.49 %. Results from UAV field experiments further confirm the method’s effectiveness, which efficiently and accurately detects microscopic cracks on concrete dam faces.

A cosine similarity-based labeling technique for vulnerability type detection using source codes

Vulnerability detection is of great importance in providing reliability to software systems. Although existing methods achieve remarkable success in vulnerability detection, they have several disadvantages as follows: (1) The irrelevant information is removed from source codes, which have a high noise ratio, thereby utilizing deep learning methods and devising experiments featuring high accuracy. However, deep learning-based detection methods necessitate large-scale datasets. This results in computational hardship with respect to vulnerability detection in small-scale software systems. (2) The majority of the studies perform feature selection by processing vulnerability commits. Despite tremendous endeavors, there are few works detecting vulnerability with source codes. To solve these two problems, in this study, a novel labeling and vulnerability detection algorithm is proposed. The algorithm first exploits source codes with the help of a keyword vulnerability matrix. After that, an ultimate encoded matrix is generated by word2vec, thereby combining the labeling vector with the source code matrix to reveal a trainable dataset for a generalized linear model (GLM). Different from preceding studies, our method performs vulnerability detection without requiring vulnerability commits but using source codes. In addition to this, similar studies generally aim to bring sophisticated solutions for just one type of programming language. Conversely, our study develops vulnerability keywords for three programming languages including C#, Java, and C++, and creates the related labeling vectors by regarding the keyword matrix. The proposed method outperformed the baseline approaches for most of the experimental datasets with over 90% of the area under the curve (AUC). Further, there is a 7.7% margin between our method and the alternatives on average for Recall, Precision, and F1-score with respect to five types of vulnerabilities.

Contrasting YOLOv7, SSD, and DETR on Insulator Identification under Small-sample Learning

Background:: Daily inspections of insulators are necessary because they are indispensable components for power transmission lines. Using deep learning to monitor insulators is a newly developed method. However, most deep learning-based detection methods rely on a large training sample set, which consumes computing resources and increases the workload of sample labeling. The selection of learning models to monitor insulators becomes problematic. Objective:: Through comparative analysis, a model suitable for small-sample insulator learning is found to provide a reference for the research and application of insulator detection. Methods:: This paper compares some of the latest deep learning models, YOLOv7, SSD, and DETR, for insulator detection based on small-sample learning. The small sample here means that the number of samples and their proportion to the total sample are relatively small. Two public insulator image sets, InsulatorDataSet with 600 insulator images and Transmission-line-pictures (TLP) with 1230 insulator images in the natural background are selected to test the performance of these models. Results:: Tests on two public insulator image sets, InsulatorDataSet and TLP, show that the recognition rates of YOLOv7, DETR, and SSD are arranged from high to low. The DETR and the YOLOv7 have stable performance, while the SSD lacks stable performance in terms of the learning time and recognition rate. Conclusion:: The in-domain and cross-domain scenario tests show that YOLOv7 is more suitable for insulator detection under small-sample conditions among the three models.

Research Progress on Deep Learning Based Defect Detection Technology for Solar Panels

INTRODUCTION: Based on machine vision technology to carry out photovoltaic panel defect detection technology research to solve the photovoltaic panel production line automation online defect detection and localization problems.
 OBJECTIVES: The goal is to improve the accuracy of defect detection on PV cell production lines, increase the speed of defect detection to meet real-time monitoring needs, and improve production efficiency.
 METHODS: In this paper, three detection methods such as image processing based detection, traditional machine learning based detection and deep learning algorithm based detection are discussed and compared and analyzed respectively. Finally, it is concluded that deep learning based detection methods are more effective in comparison. Then, further analysis and simulation experiments are done by several deep learning based detection algorithms.
 RESULTS: The experimental results show that the YOLOv8 algorithm has the highest precision rate and maintains good results in terms of recall and mAP values. The detection speed is all less than other algorithms, 10.6ms.
 CONCLUSION: The inspection model based on yolov8 algorithm has the highest comprehensive performance and is the most suitable algorithmic model for detecting defects in solar panels in production lines.

VulMPFF: A Vulnerability Detection Method for Fusing Code Features in Multiple Perspectives

Source code vulnerabilities are one of the significant threats to software security. Existing deep learning-based detection methods have proven their effectiveness. However, most of them extract code information on a single intermediate representation of code (IRC), which often fails to extract multiple information hidden in the code fully, significantly limiting their performance. To address this problem, we propose VulMPFF, a vulnerability detection method that fuses code features under multiple perspectives. It extracts IRC from three perspectives: code sequence, lexical and syntactic relations, and graph structure to capture the vulnerability information in the code, which effectively realizes the complementary information of multiple IRCs and improves vulnerability detection performance. Specifically, VulMPFF extracts serialized abstract syntax tree as IRC from code sequence, lexical and syntactic relation perspective, and code property graph as IRC from graph structure perspective, and uses Bi-LSTM model with attention mechanism and graph neural network with attention mechanism to learn the code features from multiple perspectives and fuse them to detect the vulnerabilities in the code, respectively. We design a dual-attention mechanism to highlight critical code information for vulnerability triggering and better accomplish the vulnerability detection task. We evaluate our approach on three datasets. Experiments show that VulMPFF outperforms existing state-of-the-art vulnerability detection methods (i.e., Rats, FlawFinder, VulDeePecker, SySeVR, Devign, and Reveal) in Acc and F1 score, with improvements ranging from 14.71% to 145.78% and 152.08% to 344.77%, respectively. Meanwhile, experiments in the open-source project demonstrate that VulMPFF has the potential to detect vulnerabilities in real-world environments.

Object-Enhanced YOLO Networks for Synthetic Aperture Radar Ship Detection

Synthetic aperture radar (SAR) enables precise object localization and imaging, which has propelled the rapid development of algorithms for maritime ship identification and detection. However, most current deep learning-based algorithms tend to increase network depth to improve detection accuracy, which may result in the loss of effective features of the target. In response to this challenge, this paper innovatively proposes an object-enhanced network, OE-YOLO, designed specifically for SAR ship detection. Firstly, we input the original image into an improved CFAR detector, which enhances the network’s ability to localize and perform object extraction by providing more information through an additional channel. Additionally, the Coordinate Attention mechanism (CA) is introduced into the backbone of YOLOv7-tiny to improve the model’s ability to capture spatial and positional information in the image, thereby alleviating the problem of losing the position of small objects. Furthermore, to enhance the model’s detection capability for multi-scale objects, we optimize the neck part of the original model to integrate the Asymptotic Feature Fusion (AFF) network. Finally, the proposed network model is thoroughly tested and evaluated using publicly available SAR image datasets, including the SAR-Ship-Dataset and HRSID dataset. In comparison to the baseline method YOLOv7-tiny, OE-YOLO exhibits superior performance with a lower parameter count. When compared with other commonly used deep learning-based detection methods, OE-YOLO demonstrates optimal performance and more accurate detection results.

A Deep Learning Based Detection Method for Insulator Defects in High Voltage Transmission Lines

A Deep Learning Based Detection Method for Insulator Defects in High Voltage Transmission Lines

Deep Spatial-Spectral Joint-Sparse Prior Encoding Network for Hyperspectral Target Detection.

Hyperspectral target detection aims to locate targets of interest in the scene, and deep learning-based detection methods have achieved the best results. However, black box network architectures are usually designed to directly learn the mapping between the original image and the discriminative features in a single data-driven manner, a choice that lacks sufficient interpretability. On the contrary, this article proposes a novel deep spatial-spectral joint-sparse prior encoding network (JSPEN), which reasonably embeds the domain knowledge of hyperspectral target detection into the neural network, and has explicit interpretability. In JSPEN, the sparse encoded prior information with spatial-spectral constraints is learned end-to-end from hyperspectral images (HSIs). Specifically, an adaptive joint spatial-spectral sparse model (AS 2 JSM) is developed to mine the spatial-spectral correlation of HSIs and improves the accuracy of data representation. An optimization algorithm is designed for iteratively solving AS 2 JSM, and JSPEN is proposed to simulate the iterative optimization process in the algorithm. Each basic module of JSPEN one-to-one corresponds to the operation in the optimization algorithm so that each intermediate result in the network has a clear explanation, which is convenient for intuitive analysis of the operation of the network. With end-to-end training, JSPEN can automatically capture the general sparse properties of HSIs and faithfully characterize the features of background and target. Experimental results verify the effectiveness and accuracy of the proposed method. Code is available at https://github.com/Jiahuiqu/JSPEN.

Deep learning-based detection method for analysis of high-pressure hydrogen induced damage in acrylonitrile butadiene rubber for hydrogen mobility

The increasing use of high-pressure hydrogen gas has heightened the need to understand material behavior in hydrogen-rich environments. Recent studies have shown that examining the pore-shaped damage in the cross-section of rubber materials exposed to high-pressure hydrogen can provide valuable insights into their resistance to such environments. This paper introduces an approach for training a deep learning model to detect hydrogen-induced pore-shaped damage. The study proposes a semi-automated labeling method and employs a modified faster R-CNN, implementing ResNet50-D, aspectual anchor box optimization, and dataset augmentation. To conduct the testing to validate the proposed method, acrylonitrile butadiene rubber was exposed to hydrogen at 96.6 MPa for 24 h. The dataset was created by analyzing damaged cross-sections using a scanning electron microscope. The detection results demonstrate that the proposed method outperforms both traditional and data-based conventional methods.

Research on Optimization of Boundary Detection and Dangerous Area Warning Algorithms Based on Deep Learning in Campus Security System

This study designs and implements a boundary detection and dangerous area warning algorithm based on deep learning from the perspective of typified campus security situation resources such as data, information, and knowledge. Based on integrating multiple campus security factors, real-time perception and further prediction of campus security situation can be achieved. Through coordinated operation among various algorithm modules, object intrusion in specific areas can be accurately identified and early warning can be given. The research results show that when an object invades a specific area, the difference coefficient will increase, and the larger the change value in the intrusion area, the larger the corresponding difference coefficient. By using this feature, the threshold of the difference coefficient can be determined. When a region is invaded, the contour length of the foreground will sharply increase. Based on the statistical information of the contour length of the foreground, the threshold can be set to determine whether someone has invaded the region. The deep learning algorithm in this study accurately extracts the contour of moving targets and can identify foreground targets. The real-time performance of the algorithm is also guaranteed, and it has high practical value in intelligent video monitoring. This algorithm greatly improves the efficiency of intrusion detection by utilizing the joint constraints of two types of time-domain and scene-space transformations in monitoring images. This method is not affected by the brightness of the regional environment, nor will it cause misjudgment due to significant differences in brightness of the regional environment. The detection and inference time of deep learning-based detection methods is controlled within 2-3ms, and the FPS value of the detection method is always at a high level, which can quickly increase to over 350frames/s after transmission begins. The detection method based on deep learning has higher detection efficiency.

A deep learning-based detection method for pig body temperature using infrared thermography

Body temperature is one of the important indicators that reflect the health of pigs. The traditional rectal temperature measurement is inconvenient and time-consuming, and also easy to cause stress responses to pigs. Infrared Thermography (ITG) was supposed to have potentials to realize non-invasive and rapid temperature detection for pigs in intensive pig farming. In this paper, an automatic temperature detection method base on ITG was developed. First, temperatures on six regions on pig body surface (i.e., forehead (FH), eyes, nose, ear root (ET), back and anus) were measured by ITG for the region of interest (ROI) selection; Then, an improved model of YOLOv5s-BiFPN was developed for automatic ROI detection and temperature extraction. A dataset with 2797 images that collected from sixteen pigs for 30 days was used for model development. It was shown that temperatures on FH and ET were supposed to be the ROI for ITG temperature detection and because of their strong correlations with the rectal temperature among the six parts on pig body surface. Also, the maximum temperature (MaxT) on FH and ET could reflect pig’s body temperature variations the best. The proposed model of YOLOv5s-BiFPN achieved optimal performances (e.g., the mAP was 96.36%, outperformance was 20 MB, target detection speed was up to 100 frame/s), and the mAPs were increased by 4.85%, 4.38%, 1.60%, 1.56%, 31.52%, and 22.42% compared with models of CenterNet, Faster R-CNN, YOLOv4, YOLOv5s, Nanodet and YOLOv5n, respectively. Therefore, it is a feasible way for pig body temperature automatic detection and facilitates early staged disease warning and environmental control.

Lightweight One-Stage Maize Leaf Disease Detection Model with Knowledge Distillation

Maize is one of the world’s most important crops, and maize leaf diseases can have a direct impact on maize yields. Although deep learning-based detection methods have been applied to maize leaf disease detection, it is difficult to guarantee detection accuracy when using a lightweight detection model. Considering the above problems, we propose a lightweight detection algorithm based on improved YOLOv5s. First, the Faster-C3 module is proposed to replace the original CSP module in YOLOv5s, to significantly reduce the number of parameters in the feature extraction process. Second, CoordConv and improved CARAFE are introduced into the neck network, to improve the refinement of location information during feature fusion and to refine richer semantic information in the downsampling process. Finally, the channel-wise knowledge distillation method is used in model training to improve the detection accuracy without increasing the number of model parameters. In a maize leaf disease detection dataset (containing five leaf diseases and a total of 12,957 images), our proposed algorithm had 15.5% less parameters than YOLOv5s, while the mAP(0.5) and mAP(0.5:0.95) were 3.8% and 1.5% higher, respectively. The experiments demonstrated the effectiveness of the method proposed in this study and provided theoretical and technical support for the automated detection of maize leaf diseases.

Recommendation attack detection based on improved Meta Pseudo Labels

Attackers attempt to bias the outputs of collaborative recommender systems by maliciously rating goods or services. To detect such attacks, many deep learning-based detection methods have been proposed and shown to be feasible. However, most methods require a large number of labeled user profiles for training to ensure good detection performance. To address this issue, in this paper, we propose a deep semisupervised detection approach based on the improved Meta Pseudo Labels, named DSSD-ImMPL. DSSD-ImMPL can achieve high detection performance given a small number of labeled training samples and a certain number of unlabeled training samples. We first improve the Meta Pseudo Labels method by generating a group of student networks by an experienced teacher network instead of only one student network in the original Meta Pseudo Labels method to improve the classification performance. Then, we use the group of student networks to detect the recommendation attack. The detection performance is verified with classical, mixed, GSA-GANs, and real attacks on three benchmark datasets by comparing DSSD-ImMPL with the state-of-the-art detection methods.

Real-Time Detection System of Broken Corn Kernels Based on BCK-YOLOv7

Accurately and effectively measuring the breaking quality of harvested corn kernels is a critical step in the intelligent development of corn harvesters. The detection of broken corn kernels is complicated during the harvesting process due to turbulent corn kernel movement, uneven lighting, and interference from numerous external factors. This paper develops a deep learning-based detection method in real time for broken corn kernels in response to these issues. The system uses an image acquisition device to continuously acquire high-quality corn kernel image data and cooperates with a deep learning model to realize the rapid and accurate recognition of broken corn kernels. First, we defined the range of broken corn kernels based on image characteristics captured by the acquisition device and prepared the corn kernel datasets. The corn kernels in the acquired image were densely distributed, and the highly similar features of broken and whole corn kernels brough challenges to the system for visual recognition. To address this problem, we propose an improved model called BCK-YOLOv7, which is based on YOLOv7. We fine-tuned the model’s positive sample matching strategy and added a transformer encoder block module and coordinate attention mechanism, among other strategies. Ablation experiments demonstrate that our approach improves the BCK-YOLOv7 model’s ability to learn effectively broken corn kernel features, even when high-density features are similar. The improved model achieved a precision rate of 96.9%, a recall rate of 97.5%, and a mAP of 99.1%, representing respective improvements of 3.7%, 4.3%, and 2.8% over the original YOLOv7 model. To optimize and deploy the BCK-YOLOv7 model to the edge device (NVIDIA Jetson Nano), TensorRT was utilized, resulting in an impressive inference speed of 33 FPS. Finally, the simulation system experiment for corn kernel broken rate detection was performed. The results demonstrate that the system’s mean absolute deviation is merely 0.35 percent compared to that of manual statistical results. The main contribution of this work is the fact that this is the first time that a set of deep learning model improvement strategies and methods are proposed to deal with the problem of rapid and accurate corn kernel detection under the conditions of high density and similar features.

Deep Learning-Based Cow Tail Detection and Tracking for Precision Livestock Farming

Cow tail detection and tracking in videos provide valuable information for individual identification, calving process, behavior analysis, and body condition monitoring. Although deep learning-based detection methods have demonstrated good performance, many of them have high complexity and still require an improvement in video object detection by reducing the computation time and false positives. To fully exploit the interframe information and achieve a real-time online detection, the optimized cow tail detection and tracking method is proposed based on an improved single shot multibox detector (SSD) and Kalman filter. Here, our improved SSD-integrated DenseNet and Inception-v4 reduce detection information loss and network parameters. Then, an improved window function-based Kalman filter and Hungarian are adopted to remove error detections and enhance the cow tail tracking accuracy. Experiments on our acquired rear-view videos show that the proposed approach achieved fast tail detection with an accuracy of 96.97%, a speed of 96 fps, a smaller model size of 25 MB, and higher position accuracy with an average 6.45 pixels deviation. The proposed approach outperformed the region-based R-CNN models and other tracking methods (e.g., particle filter), which provides a new solution to automatic cow detection and tracking in smart livestock farming.

Few-Shot Training GAN for Face Forgery Classification and Segmentation Based on the Fine-Tune Approach

There are many techniques for faking videos that can alter the face in a video to look like another person. This type of fake video has caused a number of information security crises. Many deep learning-based detection methods have been developed for these forgery methods. These detection methods require a large amount of training data and thus cannot develop detectors quickly when new forgery methods emerge. In addition, traditional forgery detection refers to a classifier that outputs real or fake versions of the input images. If the detector can output a prediction of the fake area, i.e., a segmentation version of forgery detection, it will be a great help for forensic work. Thus, in this paper, we propose a GAN-based deep learning approach that allows detection of forged regions using a smaller number of training samples. The generator part of the proposed architecture is used to synthesize predicted segmentation which indicates the fakeness of each pixel. To solve the classification problem, a threshold on the percentage of fake pixels is used to decide whether the input image is fake. For detecting fake videos, frames of the video are extracted and it is detected whether they are fake. If the percentage of fake frames is higher than a given threshold, the video is classified as fake. Compared with other papers, the experimental results show that our method has better classification and segmentation.

Embedding Weather Simulation in Auto-Labelling Pipelines Improves Vehicle Detection in Adverse Conditions.

The performance of deep learning-based detection methods has made them an attractive option for robotic perception. However, their training typically requires large volumes of data containing all the various situations the robots may potentially encounter during their routine operation. Thus, the workforce required for data collection and annotation is a significant bottleneck when deploying robots in the real world. This applies especially to outdoor deployments, where robots have to face various adverse weather conditions. We present a method that allows an independent car tansporter to train its neural networks for vehicle detection without human supervision or annotation. We provide the robot with a hand-coded algorithm for detecting cars in LiDAR scans in favourable weather conditions and complement this algorithm with a tracking method and a weather simulator. As the robot traverses its environment, it can collect data samples, which can be subsequently processed into training samples for the neural networks. As the tracking method is applied offline, it can exploit the detections made both before the currently processed scan and any subsequent future detections of the current scene, meaning the quality of annotations is in excess of those of the raw detections. Along with the acquisition of the labels, the weather simulator is able to alter the raw sensory data, which are then fed into the neural network together with the labels. We show how this pipeline, being run in an offline fashion, can exploit off-the-shelf weather simulation for the auto-labelling training scheme in a simulator-in-the-loop manner. We show how such a framework produces an effective detector and how the weather simulator-in-the-loop is beneficial for the robustness of the detector. Thus, our automatic data annotation pipeline significantly reduces not only the data annotation but also the data collection effort. This allows the integration of deep learning algorithms into existing robotic systems without the need for tedious data annotation and collection in all possible situations. Moreover, the method provides annotated datasets that can be used to develop other methods. To promote the reproducibility of our research, we provide our datasets, codes and models online.

Uncertainty-aware accurate insulator fault detection based on an improved YOLOX model

The surveillance and inspection of power line insulators, which act as essential components for the connection and insulation of power lines, play pivot roles in the daily maintenance of the power grid system since the failure of power line insulators could cause abrupt power cuts and accidents. Vision-assisted Unmanned Aerial Vehicle (UAV) technology has recently become an efficient and economical solution for automatic insulator fault inspection and demonstrated considerable accuracy with deep learning-based detection methods. However, current detection methods cannot predict and handle the uncertainty of insulator detection, and their capabilities are limited. This paper presents a novel deep learning-based methodology, namely YOLOD, to address the uncertainty issue by applying a Gaussian prior to the detection heads of YOLOX, the current state-of-the-art model of compact object detectors, for both bounding box regression and corresponding uncertainty estimation. Then, the estimated uncertainty scores are utilized to refine the bounding box prediction and further improve the robustness of the detection. Finally, in the comprehensive experiments, the proposed YOLOD model outperforms other benchmark models on a public insulator dataset and achieves the highest average precision (73.9%), which is 2.1% higher than that of YOLOX. Thus, the effectiveness and superiority of the proposed method for robust insulator defect inspection are validated.

Improving Deep Learning-Based Recommendation Attack Detection Using Harris Hawks Optimization

Recommendation attack attempts to bias the recommendation results of collaborative recommender systems by injecting malicious ratings into the rating database. A lot of methods have been proposed for detecting such attacks. Among these works, the deep learning-based detection methods get rid of the dependence on hand-designed features of recommendation attack besides having excellent detection performance. However, most of them optimize the key hyperparameters by manual analysis which relies too much on domain experts and their experience. To address this issue, in this paper we propose an approach based on the Harris Hawks Optimization (HHO) algorithm to improve the deep learning-based detection methods. Being different from the original detection methods which optimize the key hyperparameters manually, the improved deep learning-based detection methods can optimize the key hyperparameters automatically. We first convert the key hyperparameters of discrete type to continuous type according to the uniform distribution theory to expand the application scope of HHO algorithm. Then, we use the detection stability as an early stop condition to reduce the optimization iterations to improve the HHO algorithm. After that, we use the improved HHO algorithm to automatically optimize the key hyperparameters for the deep learning-based detection methods. Finally, we use the optimized key hyperparameters to train the deep learning-based detection methods to generate classifiers for detecting the recommendation attack. The experiments conducted on two benchmark datasets illustrate that the improved deep learning-based detection methods have effective performance.

CYBER SECURITY DETECTION USING ANN

One of the major challenges in cyber security is the provision of an automated and effective cyber-threats detection technique. We present an AI technique for cyber-threats detection, based on artificial neural networks. The proposed technique converts multitude of collected security events to individual event profiles and use a deep learning-based detection method for enhanced cyberthreat detection. For this work, we developed an AI-SIEM system based on a combination of event profiling for data preprocessing and different artificial neural network methods, including FCNN, CNN, and LSTM. The system focuses on discriminating between true positive and false positive alerts, thus helping security analysts to rapidly respond to cyber threats. All experiments in this study are performed by authors using two benchmark datasets (NSLKDD and CICIDS2017) and two datasets collected in the real world. To evaluate the performance comparison with existing methods, we conducted experiments using the five conventional machine-learning methods (SVM, k- NN, RF, NB, and DT). Consequently, the experimental results of this study ensure that our proposed methods are capable of being employed as learning-based models for network intrusion-detection, and show that although it is employed in the real world, the performance outperforms the conventional machine-learning methods.