Intelligent Detection Algorithm Research Articles

Deep Learning-Based Intelligent Detection Algorithm for Surface Disease in Concrete Buildings

In this study, the extent of concrete building distress is used to determine whether a building needs to be demolished and maintained, and the study focuses on accurately identifying target distress in different complex contexts and accurately distinguishing between their categories. To solve the problem of insufficient feature extraction of small targets in bridge disease images under complex backgrounds and noise, we propose the YOLOv8 Dynamic Plus model. First, we enhanced attention on multi-scale disease features by implementing structural reparameterization with parallel small-kernel expansion convolution. Next, we reconstructed the relationship between localization and classification tasks in the detection head and implemented dynamic selection of interactive features using a feature extractor to improve the accuracy of classification and recognition. Finally, to address problems of missed detection, such as inadequate extraction of small targets, we extended the original YOLOv8 architecture by adding a layer in the feature extraction phase dedicated to small-target detection. This modification integrated the neck part more effectively with the shallow features of the original three-layer YOLOv8 feature extraction stage. The improved YOLOv8 Dynamic Plus model demonstrated a 7.4 percentage-point increase in performance compared to the original model, validating the feasibility of our approach and enhancing its capability for building disease detection. In practice, this improvement has led to more accurate maintenance and safety assessments of concrete buildings and earlier detection of potential structural problems, resulting in lower maintenance costs and longer building life. This not only improves the safety of buildings but also brings significant economic benefits and social value to the industries involved.

EFS-YOLO: a lightweight network based on steel strip surface defect detection

Abstract With the advancement of deep learning technologies, industrial intelligent detection algorithms are gradually being applied in practical steel surface defect detection. Addressing the issues of high computational resource consumption and poor detection performance faced by existing models in large-scale industrial production lines, this paper proposes an EFS-YOLO (Efficient-Fast-Shared-YOLO) model based on improved YOLOv8s architecture. Firstly, the EfficientViT is employed as the feature extraction network, effectively reducing the model’s parameters and calculations while enhancing its capability to represent defect features. Secondly, the designed lightweight C2f-Faster-EffectiveSE Block (CFE-Block) was integrated into the model neck, accelerating feature fusion and better preserving detailed defect features at lower levels. Finally, the model detection head was reconstructed using the concept of shared parameters to address the high computational cost of the original detection head. Experimental results on the NEU-DET and GC10-DET datasets demonstrate that compared to the baseline model, the proposed model achieves a reduction in parameters, calculations and size by 49.5%, 62.7% and 46.9% respectively. It also exhibits an improvement in accuracy by 2.4% and 2.3% on the two datasets. The EFS-YOLO model effectively enhances precision in steel surface defect detection while maintaining lightweight characteristics, offering superior performance.

Dentists who can auscultate: Microphone-based toothbrushing quality monitoring system for electronic toothbrush

Tooth brushing is a widely recognized and effective method for daily oral care. However, the lack of real-time feedback for monitoring the brushing process can hinder individuals in maintaining their oral hygiene. Currently, only high-end electric toothbrushes offer evaluation functions based on built-in sensors, making them unaffordable for the majority of people. In this paper, we propose a novel and cost-effective wearable acoustic-based system (both hardware and software) for monitoring the quality of daily tooth brushing. Our system combines a throat-microphone and a Bluetooth earphone to capture brushing noises, which are unique acoustical signals generated from the air and the human body during tooth brushing, respectively. By leveraging the distinctive characteristics of brushing sounds in these two different conductive media, we introduce an intelligent detection algorithm based on Single-Input Multiple-Output (SIMO) modeling and frequency response functions (FRFs) to detect the corresponding brushing area. Through experimentation, we have found that the divergence of FRFs provides more stable features compared to general statistical features used in previous studies. After comparing various machine learning algorithms, we have selected XGBoost as the final detection model. Additionally, we utilize the Symmetric Nearest Neighbor (SNN) filter in the post-processing stage to further reduce the jitter noise in continuous detection. The final system achieves an average precision of 98.46% when tested on tooth brushing data collected from different volunteers. To enhance user experience, we have developed an Android App with a user-friendly interface that incorporates the detection system. With a total wearable hardware cost of less than 10 dollars, our system shows great potential for integration into future smart wearable devices.

BS-YOLOV8: an intelligent detection model for bearing pin support-piece states of high-rise building machine

Abstract As the main support part of the working platform of a high-rise building machine, the bearing pin support (BPS) plays a crucial role in the safety and stability of the platform, the conventional method has the problems of low detection efficiency, low accuracy, and high cost. To improve the accuracy and robustness of the detection algorithm under weak light, this paper proposes an intelligent detection algorithm for the BPS-piece states of the BS-YOLOV8, to improve the feature map utilization and reduce the model leakage detection error detection rate, Swin transformer is used to improve the YOLOV8 backbone network. In addition, the BiFormer attention mechanism is used to weigh the feature map to solve the problem of feature information loss in different feature layers and weak lighting conditions, and then the Scylla-IOU loss function is used instead of the original localization loss function to guide the model to learn to generate a predicted bounding box closer to the real target bounding box. Finally, the BS-YOLOV8 algorithm is used to compare with its classical algorithm on the self-constructed dataset of this study, The results show that the mAP0.5, mAP0.5:0.95, and FPS values of the BS-YOLOV8 algorithm reach 97.9%, 96.3% and 40 under normal lighting. The mAP0.5 value reaches 87.6% under low light conditions, which effectively solves the problems of low detection efficiency and poor detection under low light conditions, and is superior compared to other algorithms.

A YOLO-based intelligent detection algorithm for risk assessment of construction sites

A YOLO-based intelligent detection algorithm for risk assessment of construction sites

Lightweight Saliency Target Intelligent Detection based on Multi-scale Feature Adaptive Fusion

In order to solve the problems of small targets, variable shooting angles, and heights in drone images, the author proposes an adaptive drone target intelligent detection algorithm based on multi-scale feature fusion. The results show that after adding a deconvolution cascade structure to the network, mAP increased by about 2.5 percentage points and AP$^{50}$ increased by about 3 percentage points. Compared with Method 3, Method 4 uses GA-RPN instead of RPN, and when the IOU is 75, the AP increases by 3.5 percentage points, reflecting that the target prediction candidate boxes generated using semantic features adaptively match better than the manually designed target candidate boxes. This indicates that the proposed target detection framework has better classification ability and higher frame regression accuracy. Multi scale adaptive candidate regions are used to generate fused features of different scales generated by the network, weighted fused multi-scale features are used for target prediction, and semantic features are used to guide the network to adaptively generate target candidate frames, greatly enhancing the feature expression ability of various targets and improving the detection accuracy of aerial targets.

STAE‐YOLO: Intelligent detection algorithm for risk management of construction machinery intrusion on transmission lines based on visual perception

AbstractThe distribution of transmission lines is complex, influenced by terrain and topography, making their maintenance and management challenging. With the advancement of global mechanization, various types of construction machinery and large‐scale mechanical equipment are continually increasing. Accidental contact of construction machinery near transmission lines poses a serious threat to the stability of the power system. Traditional drones and inspection robots face difficulties in achieving real‐time monitoring of construction machinery and equipment around transmission lines, quickly identifying potential risks. Conventional image processing techniques and convolutional neural networks struggle with effective handling of small targets and densely packed detection tasks involving multiple targets. To address these challenges, this paper proposes an intelligent detection algorithm Swin Transformer attention efficient algorithm‐you only look once (YOLO) (STAE‐YOLO). It is based on the Swin Transformer global self‐attention mechanism, cross‐channel fusion attention mechanism, enhanced small object detection framework, and a focused and efficient regression loss function. The experimental results show that the STAE‐YOLO algorithm model improved by 6.3% in mean average precision, 3.7% in precision, and 3.1% in recall compared to the baseline model. Meanwhile, deploying the window multi‐head self‐attention global self‐attention mechanism in the model can strengthen the global multi‐scale semantic information in detection.

Lightweight DB-YOLO Facemask Intelligent Detection and Android Application Based on Bidirectional Weighted Feature Fusion

Conventional facemask detection algorithms face challenges of insufficient accuracy, large model size, and slow computation speed, limiting their deployment in real-world scenarios, especially on edge devices. Aiming at addressing these issues, we proposed a DB-YOLO facemask intelligent detection algorithm, which is a lightweight solution that leverages bidirectional weighted feature fusion. Our method is built on the YOLOv5 algorithm model, replacing the original YOLOv5 backbone network with the lightweight ShuffleNetv2 to reduce parameters and computational requirements. Additionally, we integrated BiFPN as the feature fusion layer, enhancing the model’s detection capability for objects of various scales. Furthermore, we employed a CARAFE lightweight upsampling factor to improve the model’s perception of details and small-sized objects and the EIOU loss function to expedite model convergence. We validated the effectiveness of our proposed method through experiments conducted on the Pascal VOC2007+2012 and Face_Mask datasets. Our experimental results demonstrate that the DB-YOLO model boasts a compact size of approximately 1.92 M. It achieves average precision values of 70.1% and 93.5% on the Pascal VOC2007+2012 and Face_Mask datasets, respectively, showcasing a 2.3% improvement in average precision compared to the original YOLOv5s. Furthermore, the model’s size is reduced by 85.8%. We also successfully deployed the model on Android devices using the NCNN framework, achieving a detection speed of up to 33 frames per second. Compared to lightweight algorithm models like YOLOv5n, YOLOv4-Tiny, and YOLOv3-Tiny, DB-YOLO not only reduces the model’s size but also effectively improves detection accuracy, exhibiting excellent practicality and promotional value on edge devices.

Technologies and Equipment of Mechanized Blossom Thinning in Orchards: A Review

Orchard thinning can avoid biennial bearing and improve fruit quality, which is a necessary agronomic section in orchard management. The existing methods of artificial fruit thinning and chemical spraying are no longer suitable for the development of modern agriculture. With the continuous acceleration of the construction process of modern orchards, blossom thinning mechanization has become an inevitable trend in the development of the orchard flower and fruit management. Based on relevant reports in the past 20 years, the paper discusses the current level of development of mechanized blossom thinning technologies and equipment in orchards from three aspects: mechanism research, machine development, and intelligent upgrading. Firstly, for thinning mechanism research, three directions were investigated: the rope flexible hitting force, thinning agronomic requirements, and the fruit tree growth model between thinning and fruit yields. Secondly, for marketable machine developments, two types of machines were investigated: the hand-held thinner and tractor-mounted thinner. The hand-held thinner is mainly suitable for traditional old orchards with a messy canopy structure, especially in the interior and top of the canopy. The tractor-mounted thinner is mainly suitable for orchards with the same crown structure, such as the hedge type, trunk type, and V-type. Thirdly, for equipment intelligent upgrading, the research of the intelligent detection algorithm for inflorescence on the fruit tree was investigated, for species including the apple, pear, citrus, grape, litchi, mango, and apricot. Finally, combining the advantages and disadvantages of the research, the authors propose thoughts and prospects, which can provide a reference for the design and applications of orchard mechanized blossom thinning.

Intelligent Detection Algorithm Based on 2D/3D-UNet for Internal Defects of Carbon Fiber Composites

ABSTRACT Infrared thermography testing (IRT) has been widely used in the defect detection of composite materials. However, the identification of defects characteristics is unsatisfying due to the interference of factors such as uneven background and noise in the original thermal image sequence. A novel thermography-based defect detection method with the semantic segmentation network is proposed to enhance the defect contrast and extract perfect features. To Figure out the abnormal distribution of temperature field in thermal images, AG-UNet was used with a spatial self-attention gate module to extract spatial features of thermal images. The 3D-UNet network was obtained by the 3D convolution module and the temporal convolution module to extract thermal temporal and spatial features simultaneously which could help the defect segmentation in thermal videos. Compared with traditional algorithms such as principal component analysis (PCA), thermographic signal reconstruction (TSR), and fast Fourier transform (FFT), defects detection results were significantly enhanced with the proposed method, and defects of smaller diameter-to-depth ratio can be detected by deep learning models.

DFF-SC4N: A Deep Federated Defence Framework for Protecting Supply Chain 4.0 Networks

The management of contemporary communication networks of supply chain (SC) 4.0 is becoming more complex due to the heterogeneity requirements of new devices concerning the integration of the Internet of Things in the legacy industry networks. Hence, it becomes a challenging task to secure networks of SC 4.0 from cyber-attacks and provide a robust and efficient defence framework that can resist sophisticated attacks. Machine learning-based intelligent detection algorithms are often trained at either a centralized or single server, which makes it difficult to train an effective model and also it violates privacy concerns if gathering data from other servers at the edge. Classical machine learning approaches function on the legacy group of data placed on a central or single server, which brands it the least favored choice for supply chain networks, with data privacy issues. To address these problems, this article proposes a federated learning-based efficient detection model named, DFF-SC4N, to proactively identify intrusions from SC 4.0 networks using distributed local data training. DFF-SC4N uses communication rounds in a federated learning manner having gated recurrent units by only sharing the learned parameters and keeps the data intact on local servers. The accuracy of the global model is optimized by an aggregating model, which updates from multiple servers and multiple SC 4.0 networks. Extensive experiments on real industrial network data demonstrate that the DFF-SC4N outperforms both centralized training models and state-of-the-art peer methods in protecting SC 4.0 networks.

A Novel Intelligent Detection Algorithm of Aids to Navigation Based on Improved YOLOv4

Aiming at the problem of high-precision detection of AtoN (Aids to Navigation, AtoN) in the complex inland river environment, in the absence of sufficient AtoN image types to train classifiers, this paper proposes an automatic AtoN detection algorithm Aids-to-Navigation-YOLOv4 (AN-YOLOv4) based on improved YOLOv4 (You Only Look Once, Yolo). Firstly, aiming at the problem of an insufficient number of existing AtoN datasets, the Deep Convolutional Generative Adversarial Networks (DCGAN) is used to expand and enhance the AtoN image dataset. Then, aiming at the problem of small target recognition accuracy, the image pyramid is used to multi-scale zoom the dataset. Finally, the K-means clustering algorithm is used to correct the candidate box of AN-YOLOv4. The test on the test dataset shows that the improvement effect of AN-YOLOv4 is obvious. The accuracy rate of small targets is 92%, and the average accuracy (mAP) of eight different types of AtoN is 92%, which is 14% and 13% higher than the original YOLOv4, respectively. This research has important theoretical significance and reference value for the intelligent perception of the navigation environment under the intelligent shipping system.

Overlapping Pellet Size Detection Method Based on Marker Watershed and GMM Image Segmentation

The particle size of pellets is an important parameter in steel big data, and the high density and high overlap rate of pellets bring a great challenge to particle size detection. To address this problem, a particle size intelligent detection algorithm with an improved watershed and a Gaussian mixture model (GMM) is proposed. First, the initial segmentation of the pellets and background is achieved by using adaptive binary segmentation, and then the secondary fine segmentation of the pellets and background is achieved by combining morphological operations such as skeleton extraction and marked watershed segmentation; then, the contour of the connected domain of pellets is calculated, and the non-overlapping pellets in the foreground and the overlapping pellets are filtered according to the roundness of their contours. Finally, the number of overlapping pellets is predicted by Gaussian reconstruction of the grayscale image of the overlapping pellets, and the number and granularity of the overlapping pellets are predicted by the Gaussian reconstruction of the overlapping pellets. The experimental results showed that the algorithm achieved a 91.98% segmentation accuracy in the experimental images. Compared with other algorithms, the algorithm can also effectively suppress the over-segmentation and under-segmentation problems, and it can effectively realize the pellet size detection of dense, overlapping pellets such as those on a pelletizing disk, which provides an effective technical means for the metallurgical performance analysis of pellet ore and intelligent pellet-making driven by big data.

Application Research of an Intelligent Detection Algorithm for Vehicle Trajectory Route Deviation

Application Research of an Intelligent Detection Algorithm for Vehicle Trajectory Route Deviation

Intelligent Detection Algorithm for Small Targets Based on Super-Resolution Reconstruction

Intelligent Detection Algorithm for Small Targets Based on Super-Resolution Reconstruction

Intelligent Detection Algorithm of Medium and Low Voltage Switch Performance Based on the Improved K-means Clustering

The performance of medium and low voltage switches will directly affect the reliability of medium and low voltage distribution networks. The medium and low voltage switch algorithm’s failure form is mainly mechanical. In order to realize its intelligent detection, a medium and low voltage switch detection algorithm based on the improved K-means clustering algorithm is proposed. Optimizing the algorithm’s clustering center selection establishes a model of the switch working state under different vibration signals and an intelligent detection algorithm for medium and low voltage switches with intelligent judgment, adaptive adjustment, and remote fault diagnosis. Experiments show that: the detection algorithm of the medium and low voltage switch controller based on the improved K-means algorithm can more accurately identify the fault of the switch algorithm according to the vibration signals, which provides a new idea for the fault identification and monitoring of the medium and low voltage switch.

Research on Transformer Oil Level Monitoring System Based on Inspection Robot

Because the traditional oil level gauge transformer relies on manual regular patrol detection, it is not easy to know the state change of the oil level in real time. In recent years, the rapid development of AR/VR technology, the positioning, and the construction of maps of robots have also received widespread attention, and the oil level inspection task has gradually evolved in the direction of intelligent machine inspection. Based on this background, this paper studies the Cartographer algorithm, according to the mapping results for path planning, to achieve robot autonomous navigation of the designated location. After arriving at the destination, the user obtains the oil level by the depth camera, and uses the YOLOv4 to design an intelligent detection algorithm for oil level detection, using neural networks to learn the oil level state in various weather, and train a dedicated model. Through experimental verification, the inspection robot can complete the construction of the map and realize autonomous path navigation according to the target point, and the oil level detection system based on the YOLOv4 algorithm can achieve an accuracy of more than 90% for the recognition of oil level information, which has good applicability.

Lightweight CNN-Based Method for Spacecraft Component Detection

Spacecraft component detection is essential for space missions, such as for rendezvous and on-orbit assembly. Traditional intelligent detection algorithms suffer from drawbacks related to high computational burden, and are not applicable for on-board use. This paper proposes a convolutional neural network (CNN)-based lightweight algorithm for spacecraft component detection. A lightweight approach based on the Ghost module and channel compression is first presented to decrease the amount of processing and data storage required by the detection algorithm. To improve feature extraction, we analyze the characteristics of spacecraft imagery, and multi-head self-attention is used. In addition, a weighted bidirectional feature pyramid network is incorporated into the algorithm to increase precision. Numerical simulations show that the proposed method can drastically reduce the computational overhead while still guaranteeing good detection precision.

Intelligent DC Arc-Fault Detection of Solar PV Power Generation System via Optimized VMD-Based Signal Processing and PSO–SVM Classifier

In a solar photovoltaic (PV) power generation system, arc faults including series arc fault (SAF) and parallel arc fault (PAF) may occur due to aging of joints or other reasons. It may lead to a major safety accident, such as fire, if the high temperature caused by the continuous arc fault is not identified and solved in time. Because the SAF without drastic current change is difficult to detect, an intelligent detection algorithm based on the optimized variational mode decomposition and the support vector machine (SVM) is investigated in this article. The proposed algorithm uses the variational mode decomposition to extract the fault information from current signals, and then screens the statistical information of the signals in each frequency band by the proposed adaptive feature screening. The features to be strongly correlated with classification are taken as inputs into the SVM optimized by the particle swarm optimization for classification finally. The proposed intelligent framework not only can accurately identify the SAF occurring at different locations, but also identify the PAF. Moreover, it can also maintain good diagnosing results under the occurrence of dynamic shading, inverter startup, and SAF under wind blowing. In addition, single-series PV string and solar PV power generation systems in different countries are also used to examine the universal ability of the proposed algorithm. As for experimental results, the detection accuracy is more than 98.21% under all examined conditions.

Application of Recurrent Neural Network Algorithm in Intelligent Detection of Clinical Ultrasound Images of Human Lungs.

Lung ultrasound has great application value in the differential diagnosis of pulmonary exudative lesions. It has good sensitivity and specificity for the diagnosis of various pulmonary diseases in neonates and children. It is believed that it can replace chest CT examination. It is routinely used for the diagnosis of pulmonary diseases in emergency critical care medicine. However, the interpretation of the impact of ultrasound on the human lungs relies heavily on experienced physicians, which greatly restricts the interpretation efficiency of the impact of ultrasound. In order to improve the efficiency of monitoring and interpretation of the impact of ultrasound, this paper proposes an intelligent detection algorithm for human lung clinical ultrasound images based on recurrent neural network. Transfer learning is used to replace the fully connected layer of the VGG16 model and improve the loss function, so that the same the Euclidean distance between category images can be reduced, and the Euclidean distance between different categories of images can be increased, enhancing the resolution of the entire model, thereby achieving better image feature extraction results. The experimental results show that the algorithm proposed in this paper can surpass the doctor's level in the identification of various lung diseases.