A modified YOLOv5 helmet detection algorithm based on Swin Transformer

Prohibited item detection is crucial for the safety of public places. Deep learning, one of the mainstream methods in prohibited item detection tasks, has shown superior performance far beyond traditional prohibited item detection methods. However, most neural network architectures in deep learning still lack sufficient local feature representation ability for overlapping and small targets, and ignore the problem of semantic conflicts caused by direct feature fusion. In this paper, we propose MFPIDet, a novel prohibited item detection neural network architecture based on improved YOLOV7 to achieve reliable prohibited item detection in complex environments. Specifically, a multi-scale attention module (MAM) backbone is proposed to filter the redundant information of target regions and further applied to enhance the local feature representation ability of overlapping objects. Here, to reduce the redundant information of target regions, a squeeze-excitation (SE) block is used to filter the background. Then, aiming at enhancing the feature expression ability of overlapping objects, a multi-scale feature extraction module (MFEM) is designed for local feature representation. In addition, to obtain richer context information, We design an adaptive fusion feature pyramid network (AF-FPN) to combine the adaptive context information fusion module (ACIFM) with the feature fusion module (FFM) to improve the neck structure of YOLOV7. The proposed method is validated on the PIDray dataset, and the tested results showed that our method obtained the highest mAP (68.7%), which is improved by 3.5% than YOLOV7 methods. Our approach provides a new design pattern for prohibited item detection in complex environments and shows the development potential of deep learning in related fields.

In dynamic and complex real-world environments, artificial intelligence (AI) vision systems continue to face significant challenges in accurately detecting and tracking small objects. The core difficulty lies in the fact that small targets usually exhibit limited spatial and textural features, while dynamic backgrounds often generate numerous misleading motion cues, thereby interfering with reliable discrimination between targets and backgrounds. Inspired by the remarkable capability of the insect brain in detecting small moving objects, this study proposes a visual neural network model enhanced by a feedback mechanism. By adaptively responding to temporal variations, the proposed model is able to more precisely distinguish small targets from background-induced false targets. The network architecture consists of two main pathways: a motion detection pathway that extracts motion-related features from minute targets, and a feedback attention pathway that enhances the focus on true targets by leveraging the feature differences between real and false motion signals. In addition, a global inhibition module is incorporated to reduce the false alarm rate by filtering out background-induced false positives, thereby improving the overall detection performance of the model. Experimental results demonstrate that the proposed model achieves a detection rate of 0.81 in complex visual scenarios, whereas the compared models all achieve detection rates below 0.59, indicating a significant improvement in detection performance. Meanwhile, in terms of Precision and F1-score, the proposed model achieves values of 0.0648 and 0.12, respectively, while the compared models obtain values lower than 0.0077 and 0.015, further validating the superiority of the proposed method in detection accuracy and robustness.

Addressing the challenges associated with detecting thin clouds in domestically produced mediumhigh resolution satellite imagery—such as low detection accuracy, strong threshold dependence, high thin cloud omission rates, and the insufficient adaptability and computational efficiency of existing algorithms— which primarily stem from the lack of infrared bands, this paper proposes a Haze-Optimized Adaptive Thresholding (HOAT) method based on IDL (Interactive Data Language) . The HOAT method integrates an improved Haze Optimization Transformation (HOT) index with a dynamically weighted OTSU algorithm. Through multi-band dynamic weight optimization, accelerated matrix operations, and an adaptive processing mechanism for marine scenarios, it achieves efficient and precise cloud detection in complex environments. Experimental results demonstrate that HOAT significantly outperforms traditional methods in thin cloud detection accuracy, suppression of sun glint false detection, and processing efficiency, providing reliable technical support for disaster emergency remote sensing monitoring.

One of the fundamental operations in computer vision applications is change detection, in which moving foreground objects are segmented from a static background. A common approach for change detection is the comparison of an image frame with the stored background model using a matching algorithm, a process known as background subtraction. However, such techniques fail in environments with dynamic backgrounds, illumination changes, or shadow and camera jitters. This study focuses on effectively detecting scene changes in complex environments. To this end, we proposed a new colour descriptor named local colour difference pattern (LCDP) that is insusceptible to shadow and is able to capture both colour and texture features at a pixel location. Furthermore, a scene change detection framework was proposed to handle dynamic scenes based on sample consensus that integrates LCDP and a novel spatial model fusion mechanism. Experiments using the CDnet benchmark dataset demonstrated the effectiveness of the proposed approach to change detection in complex environments.

Insulator defect detection is important in ensuring power systems’ safety and stable operation. To solve the problems of its low accuracy, high delay, and large model size in complex environments, following the principle of progressive extraction from high-entropy details to low-entropy semantics, an improved YOLOv8 target detection network for insulator defects based on bidirectional weighted feature fusion was proposed. A C2f_DSC feature extraction module was designed to identify more insulator tube features, an EMA (encoder–modulator–attention) mechanism and a BiFPN (bidirectional weighted feature pyramid network) fusion layer in the backbone network were introduced to extract different features in complex environments, and EIOU (efficient intersection over union) as the model’s loss function was used to accelerate model convergence. The CPLID (China Power Line Insulator Dataset) was tested to verify the effectiveness of the proposed algorithm. The results show its model size is only 6.40 M, and the mean accuracy on the CPLID dataset reaches 98.6%, 0.8% higher than that of the YOLOv8n. Compared with other lightweight models, such as YOLOv8s, YOLOv6, YOLOv5s, and YOLOv3Tiny, not only is the model size reduced, but also the accuracy is effectively improved with the proposed algorithm, demonstrating excellent practicality and feasibility for edge devices.

Face mask-wearing detection is of great significance for safety protection during the epidemic. Aiming at the problem of low detection accuracy due to the problems of occlusion, complex illumination, and density in mask-wearing detection, this paper proposes a neural network model based on the loss function and attention mechanism for mask-wearing detection in complex environments. Based on YOLOv5s, we first introduce an attention mechanism in the feature fusion process to improve feature utilization, study the effect of different attention mechanisms (CBAM, SE, and CA) on improving deep network models, and then explore the influence of different bounding box loss functions (GIoU, CIoU, and DIoU) on mask-wearing recognition. CIoU is used as the frame regression loss function to improve the positioning accuracy. By collecting 7,958 mask-wearing images and a large number of images of people without masks as a dataset and using YOLOv5s as the benchmark model, the mAP of the model proposed in the paper reached 90.96% on the validation set, which is significantly better than the traditional deep learning method. Mask-wearing detection is carried out in a real environment, and the experimental results of the proposed method can meet the daily detection requirements.

The continuous advancement of intelligent coal mining technology has made precise detection of coal-rock interfaces a core factor in enhancing mining efficiency and safety. Deep learning-based detection methods are constrained by the difficulty in acquiring high-quality coal-rock interface images. Ground penetrating radar (GPR), with its high real-time capability, high accuracy, and non-destructive detection, is employed as the tool for acquiring B-Scan image datasets of the coal-rock interface. Unlike hyperbolic target recognition, the coal-rock interface poses challenges for efficient and accurate detection in complex environments due to its continuous nature and the presence of extensive clutter signals. To address these challenges, this paper proposes the GPR Coal-Rock Interface Detection Method (G-FSCNet), which integrates an improved Faster R-CNN with cubic spline interpolation anomaly midpoint filtering. This approach combines an enhanced object detection network with the proposed cubic spline interpolation anomaly midpoint filtering to achieve accurate detection and fitting of coal-rock interface regions and curves. Experimental results demonstrate high detection accuracy and robustness under complex backgrounds and clutter conditions, providing technical support for achieving transparent and intelligent coal mining operations.

In complex environments, the clutter statistical characteristics of synthetic aperture radar (SAR) are inconstant, and the constant detection performance of a false alarm rate (CFAR) detector based on a clutter statistical model is also hard to achieve. As a result, the overestimated threshold leads to a degradation in detection probability. To this end, this paper proposes a SAR ship detector different from CFAR detectors, which is independent of traditional clutter statistical distribution models and the probability of a false alarm (PFA). The proposed detector aims to raise the ship detection probability and alleviate interference from complex environments such as multi-target areas, shores, and breakwaters. It estimates clutter-truncated thresholds based on clutter intensity statistics (CIS). Firstly, three statistical parameters, including the mean, standard deviation, and maximum intensity of background clutter contaminated by outliers, are calculated; secondly, these parameters are utilized to estimate the clutter-truncated threshold using the novel CIS; and finally, the pixel under test is determined according to the CIS detection rule. Compared with CFAR-based algorithms, CIS obtains a high probability of detection in complex environments. As for other aspects, the CIS detector is insensitive to the structure of the detection window, as well as the size. It is also computationally efficient due to its simple calculations. The superiority of the CIS detector is validated on scene-differed SAR images from the DSSDD dataset.

In the current agricultural landscape, a significant portion of tomato plants suffer from leaf diseases, posing a major challenge to manual detection due to the task's extensive scope. Existing detection algorithms struggle to balance speed with accuracy, especially when identifying small-scale leaf diseases across diverse settings. Addressing this need, this study presents FCHF-DETR (Faster-Cascaded-attention-High-feature-fusion-Focaler Detection-Transformer), an innovative, high-precision, and lightweight detection algorithm based on RT-DETR-R18 (Real-Time-Detection-Transformer-ResNet18). The algorithm was developed using a carefully curated dataset of 3147 RGB images, showcasing tomato leaf diseases across a range of scenes and resolutions. FasterNet replaces ResNet18 in the algorithm's backbone network, aimed at reducing the model's size and improving memory efficiency. Additionally, replacing the conventional AIFI (Attention-based Intra-scale Feature Interaction) module with Cascaded Group Attention and the original CCFM (CNN-based Cross-scale Feature-fusion Module) module with HSFPN (High-Level Screening-feature Fusion Pyramid Networks) in the Efficient Hybrid Encoder significantly enhanced detection accuracy without greatly affecting efficiency. To tackle the challenge of identifying challenging samples, the Focaler-CIoU loss function was incorporated, refining the model's performance throughout the dataset. Empirical results show that FCHF-DETR achieved 96.4% Precision, 96.7% Recall, 89.1% mAP (Mean Average Precision) 50-95 and 97.2% mAP50 on the test set, with a reduction of 9.2G in FLOPs (floating point of operations) and 3.6M in parameters. These findings clearly demonstrate that the proposed method improves detection accuracy and reduces computational complexity, addressing the dual challenges of precision and efficiency in tomato leaf disease detection.

Rail line detection is a crucial technology for ensuring the safety and efficiency of rail transit systems. Despite significant advancements in computer vision and deep learning, existing methods struggle to accurately detect rail lines in distant and curved sections, particularly under complex environmental conditions. To address these limitations, this study proposes a novel dynamic regression loss function, termed Dynamic Radius-Based intersection-over-union Loss (Dynamic RIoU Loss), designed to optimize detection accuracy in challenging scenarios. The proposed Dynamic RIoU (DR) loss function integrates a radius-based intersection-over-union metric with a width-based dynamic weighting mechanism, enabling the model to prioritize distant curved regions and adapt to diverse track geometries. Extensive experiments were conducted on the RailSem19 and DL-Rail datasets, where the DR loss function consistently outperformed existing methods, achieving improvements of 2-4% in detection accuracy and reductions of 5 to 8 points in regression errors. This study not only advances rail line detection in complex environments but also provides a foundation for enhancing intelligent rail transit systems, contributing to safer and more reliable rail operations.

There are a lot of flammable materials in the textile workshop, and once a fire occurs, it will cause property damage and casualties. At present, smoke detection in textile workshops mainly relies on temperature‐sensing smoke sensors with low detection rate and poor real‐time performance, which cannot meet the task of smoke detection in complex environments. Therefore, this paper proposes an improved mixed Gaussian and YOLOv5 smoke detection algorithm for textile workshops. In order to reduce the interference of static background in smoke detection, an improved gaussian mixture algorithm is used to extract suspected smoke areas in video by using the dynamic characteristics of smoke. Then, an adaptive attention module is added to the feature pyramid infrastructure of the YOLOv5 target detection network to improve the multi‐scale target recognition ability. In addition, the focal loss function is used to reduce the impact of background and foreground class imbalances on the detection results. The experimental results show that the detection accuracy of the proposed method is 94.7%, and the average detection speed is 66.7 FPS. By comparing with the existing state‐of‐the‐art algorithms, the detection capability of this method has been significantly improved. At the same time, it has high real‐time performance and detection accuracy in smoke detection in textile workshops.

With the rapid development of the low-altitude economy and smart city construction, the demand for small target detection has been growing significantly. This paper proposes a deep learning model, MAE-ResTransFPN, for radar target detection tasks in clutter-rich environments. First, real-world measured data is acquired using holographic staring radar, and the Range-Doppler plane is used as the input to the neural network to complete dataset preparation and annotation. Second, the proposed MAE-ResTransFPN model combines the pre-training capabilities of the masked autoencoder (MAE), the feature extraction capabilities of ResNet50, and the multi-scale feature fusion capabilities of Transformer-FPN, aiming to improve the accuracy and robustness of radar target detection in complex environments. Experimental results show that the proposed model achieves 78.2% in precision, 71.4% in recall, and 73.4% in mAP50, demonstrating its effectiveness in radar target detection tasks.

Ship detection plays an important role in the port monitor and management. The difficulty in ship detection is that with the change of ship traffic flow, limited with small ship imaging scales and face the foggy navigation. The traditional ship image detection method has insufficient detection performance, which cannot meet the needs of the port. In order to improve the performance of ship detection in the port, we detect ship by yolov3 detection algorithm to solve the above problems. This ship detection framework based on convolutional neural networks. The characteristics of the ship are extracted layer by layer through the DarkNet53 network, and multi-scale image pyramid features are formed to detect ships of different scales. Specifically, large ships, medium ships, small ships. And uses yolov3 algorithm to achieve ship detection. We have selected four typical port navigation scenes: (1) small traffic flow (2)foggy ship navigation (3)large traffic flow (4) small imaging scale. Experiments show that the ship detection based on yolov3 has high accuracy in the face of complex sea navigation conditions and can cope with the detection requirements of different scenes at sea.

Miniaturized photoelectrochemical sensing system for reusable detection of macromolecules and its applications for unattended environmental monitoring

Accurately detecting and segmenting grape cluster in the field is fundamental for precision viticulture. In this paper, a new backbone network, ResNet50-FPN-ED, was proposed to improve Mask R-CNN instance segmentation so that the detection and segmentation performance can be improved under complex environments, cluster shape variations, leaf shading, trunk occlusion, and grapes overlapping. An Efficient Channel Attention (ECA) mechanism was first introduced in the backbone network to correct the extracted features for better grape cluster detection. To obtain detailed feature map information, Dense Upsampling Convolution (DUC) was used in feature pyramid fusion to improve model segmentation accuracy. Moreover, model generalization performance was also improved by training the model on two different datasets. The developed algorithm was validated on a large dataset with 682 annotated images, where the experimental results indicate that the model achieves an Average Precision (AP) of 60.1% on object detection and 59.5% on instance segmentation. Particularly, on object detection task, the AP improved by 1.4% and 1.8% over the original Mask R-CNN (ResNet50-FPN) and Faster R-CNN (ResNet50-FPN). For the instance segmentation, the AP improved by 1.6% and 2.2% over the original Mask R-CNN and SOLOv2. When tested on different datasets, the improved model had high detection and segmentation accuracy and inter-varietal generalization performance in complex growth environments, which is able to provide technical support for intelligent vineyard management.