The task of multiple-tiny-object detection from diverse perspectives in unmanned aerial vehicles (UAVs) using onboard edge devices is a significant and complex challenge within computer vision. In order to address this challenge, we propose a lightweight and efficient tiny-object-detection algorithm named LE-YOLO, based on the YOLOv8n architecture. To improve the detection performance and optimize the model efficiency, we present the LHGNet backbone, a more extensive feature extraction network, integrating depth-wise separable convolution and channel shuffle modules. This integration facilitates a thorough exploration of the inherent features within the network at deeper layers, promoting the fusion of local detail information and channel characteristics. Furthermore, we introduce the LGS bottleneck and LGSCSP fusion module incorporated into the neck, aiming to decrease the computational complexity while preserving the detector’s accuracy. Additionally, we enhance the detection accuracy by modifying its structure and the size of the feature maps. These improvements significantly enhance the model’s capability to capture tiny objects. The proposed LE-YOLO detector is examined in ablation and comparative experiments on the VisDrone2019 dataset. In contrast to YOLOv8n, the proposed LE-YOLO model achieved a 30.0% reduction in the parameter count, accompanied by a 15.9% increase in the mAP(0.5). These comprehensive experiments indicate that our approach can significantly enhance the detection accuracy and optimize the model efficiency through the organic combination of our suggested enhancements.
The rapid advancement of industrial automation in the automotive manufacturing sector has heightened demand for welding quality, particularly in critical component welding, where traditional manual inspection methods are inefficient and prone to human error, leading to low defect recognition rates that fail to meet modern manufacturing standards. To address these challenges, an enhanced YOLOv8-based algorithm for steel defect detection, termed YOLOv8-WD (weld detection), was developed to improve accuracy and efficiency in identifying defects in steel. We implemented a novel data augmentation strategy with various image transformation techniques to enhance the model’s generalization across different welding scenarios. The Efficient Vision Transformer (EfficientViT) architecture was adopted to optimize feature representation and contextual understanding, improving detection accuracy. Additionally, we integrated the Convolution and Attention Fusion Module (CAFM) to effectively combine local and global features, enhancing the model’s ability to capture diverse feature scales. Dynamic convolution (DyConv) techniques were also employed to generate convolutional kernels based on input images, increasing model flexibility and efficiency. Through comprehensive optimization and tuning, our research achieved a mean average precision (map) at IoU 0.5 of 90.5% across multiple datasets, contributing to improved weld defect detection and offering a reliable automated inspection solution for the industry.
Target detection via remote sensing is extensively utilized across diverse domains because of its inherent potential value in applications. However, most objects within remote sensing images consist of multi-scale and dense small objects, observed from diverse angles against complex backgrounds, resulting in insufficient detection performance. To enhance the detection accuracy and robustness in detecting multi-scale objects, we present the YOLO-GE algorithm based on you only look once (YOLO). We introduce the ghost convolution hierarchical graph (G-HG) block that combines ghost convolutions and the cross-stage partial (CSP) strategy. This enhancement can efficiently utilize redundant feature maps, broaden the receptive field, and accurately extract multi-scale objects and advanced semantic features in complex backgrounds. By incorporating the G-HG block, we establish the ghost-convolution enhanced hierarchical graph (GE-HGNet) feature extraction backbone, thereby enhancing its ability to capture multi-scale object features and advanced semantic information. Additionally, we develop the E-SimAM attention mechanism using residual techniques to address the low-resolution feature loss limitation, thereby enhancing the precision in identifying low-resolution features against intricate backgrounds. Furthermore, to improve the capability of detecting densely packed small objects, we reconstruct the structure of the neck and add a tiny detection head. This additional tiny detection head is specifically designed to better focus on densely packed small targets, fully leveraging the fine-grained information in shallow feature maps. Extensive experiments conducted on the DIOR, NWPU VHR-10, and VisDrone2019 datasets demonstrate the effectiveness and robustness of our YOLO-GE algorithm. Notably, compared to the state-of-the-art algorithm, our YOLO-GE-n achieves improvements of 20.1% and 22.2% in mAP0.5 and mAP 0.5:0.95 respectively on the VisDrone2019 dataset.
To tackle the challenges of poor stability during real-time random gait switching and precise trajectory control for hexapod robots under limited stride and steering conditions, a novel real-time replanning gait switching control strategy based on an omnidirectional gait and fuzzy inference is proposed, along with an attitude control method based on the single-neuron adaptive proportional–integral–derivative (PID). To start, a kinematic model of a hexapod robot was developed through the Denavit–Hartenberg (D-H) kinematics analysis, linking joint movement parameters to the end foot’s endpoint pose, which formed the foundation for designing various gaits, including omnidirectional and compound gaits. Incorporating an omnidirectional gait could effectively resolve the challenge of precise trajectory control for the hexapod robot under limited stride and steering conditions. Next, a real-time replanning gait switching strategy based on an omnidirectional gait and fuzzy inference was introduced to tackle the issue of significant impacts and low stability encountered during gait transitions. Finally, in view of further enhancing the stability of the hexapod robot, an attitude adjustment algorithm based on the single-neuron adaptive PID was presented. Extensive experiments confirmed the effectiveness of this approach. The results show that our approach enabled the robot to switch gaits seamlessly in real time, effectively addressing the challenge of precise trajectory control under limited stride and steering conditions; moreover, it significantly improved the hexapod robot’s dynamic stability during its motion, enabling it to adapt to complex and changing environments.
Low-voltage electrical appliances, represented by circuit breakers, contactors, and proximity switches, are widely used in various electrical control systems in transportation vehicles. During vehicle operation, engine vibrations, poor workplace balance, constantly changing operating directions, unbalanced transmission systems, emergency braking, and other factors can all cause variable loads. These variable loads may decrease the effectiveness of low-voltage electrical contacts; the failure rate of the low-voltage electrical appliances used in transportation vehicles is three times that of normal indoor low-voltage electrical appliances. This study analyzes the failure mechanism of low-voltage electrical appliances used in transportation vehicles and establishes a model for their reliability evaluation and prediction based on a variable-load data-driven approach. This variable-load data comes from the constructed simulated vehicle operation test platform. Nonlinear variable-load test data generated by simulation is collected through the test platform and is then processed. An evaluation feature dataset is constructed and input into the reliability evaluation and prediction model to obtain the remaining life of low-voltage electrical appliances. The analysis and verification of the predicted evaluation values and the real values detected through platform equipment showed that the accuracy of this model for these appliances based on the variable-load data-driven approach reached 94%, meeting the requirements of practical applications. This method used in this study to derive the model can provide a theoretical basis for online evaluation and prediction of low-voltage electrical appliance reliability for transportation vehicles. This can not only prevent vehicle failures and avoid sudden accidents, but also fully utilize the remaining life of low-voltage electrical appliances and reduce the cost of replacing them.