Ensuring the safety of transmission lines necessitates effective insulator defect detection. Traditional methods often need more efficiency and accuracy, particularly for tiny defects. This paper proposes an innovative insulator defect recognition method leveraging YOLOv8s-SwinT. Combining Swin Transformer and Convolutional Neural Network (CNN) enhances the model’s understanding of multi-scale global semantic information through cross-layer interactions. The improved BiFPN structure in the neck achieves bidirectional cross-scale connections and weighted feature fusion during feature extraction. Additionally, a new small-target detection layer enhances the capability to detect tiny defects. The experimental results showcase outstanding performance, with precision, recall, and mAP reaching 95.6%, 95.3%, and 97.7%, respectively. This boosts detection efficiency and ensures high accuracy, providing robust support for real-time detection of tiny insulator defects.

Video object segmentation (VOS) plays an important role in video analysis and understanding, which in turn facilitates a number of diverse applications, including video editing, video rendering, and augmented reality / virtual reality. However, existing deep learning-based approaches rely heavily on a large number of pixel-wise annotated video frames to achieve promising results, which is notoriously laborious and costly. To address this, in this paper, we formulate unsupervised video object detection by exploring simulated dense labels and explicit motion clues. Specifically, we first propose an effective video label generator network based on the sparsely annotated frames and the flow motion between them. It can largely alleviate our dependence and limitation on the sparse labels. Furthermore, we propose a transformer-based architecture to model the appearance and motion clues simultaneously with the cross-attention module, in order to maximally overcome non-linear motion with potential occlusions. Extensive experiments show that the proposed method outperforms recent VOS methods on four popular benchmarks (i.e., DAVIS-16, FBMS, Youtube-VOS and SegTrack-v2). Moreover, the proposed method can be further applied to a wide range of wild scenes such as wild forests and animals. Because of its effectiveness and generalization, we believe that our method could serve as a useful basis for alleviating the dependence on dense annotation in video data.

Aiming at the problem that conventional tracking algorithms are difficult to deal with abrupt motion efficiently, an optimization algorithm called hybrid Teaching-learning-based optimization with Adaptive Grasshopper Optimization Algorithm (TLGOA) is proposed in this paper. Firstly, the non-linear strategy based on tangent function is used to replace the linear mechanism in the standard Grasshopper Optimization Algorithm (GOA). The improved adaptive GOA (AGOA) can avoid the local trapping problem and enhance the global optimization ability, which can handle the problem of abrupt motion. Secondly, considering that Teaching-learning-based optimization (TLBO) has obviously local exploitation operator and fast convergence, a hybrid TLGOA tracker is designed by combining the advantages of both AGOA and TLBO. The approach can enable better tracking accuracy and efficiency. Finally, extensive experimental results show that the proposed algorithm has obvious advantages over other algorithms, and also prove that TLGOA tracker is very competitive compared to other state-of-the-art trackers, especially for abrupt motion tracking.

In the field of instrumentation and measurement science, quantized system identification based on sophisticated sensors has greatly reduced the cost of regular sensors. Although existing identification techniques are available, an identification algorithm with a novel-framework and high estimation performance is required for new applications. This report is concerned with the system identification of a nonlinear Wiener-Hammerstein system with binary-valued measurements. In quantized system identification communities, stochastic approximation type scheme is a main direction of research by directly constructing an effective identification algorithm based on the error learning feedback principle. To overcome the difficulty in constructing the estimator by using the data directly related to parameter estimation (e.g., estimation error information, initial error information), this report aims to introduce a method to utilize the estimation error information, and to establish an adaptive estimator by combining the parameter initial error information. A novel-structured adaptive filter is introduced to improve the estimation bias phenomenon. By use of auxiliary vectors and matrices, an estimation error representation is established. Then, the estimation error data with the conversion operator and initial error data with a smoothing factor are merged to derive the identifier, in which the time-varying gain is also provided. Theoretical analysis shows that the estimate reaches the true value of the parameter in the sense of almost surely. Numerical results and practical applications are supplied to clarify and verify the theoretical findings.