Yaru Wang

Insulator defect detection is a pivotal task in the intelligent inspection of power transmission lines. Currently, challenges such as the scarcity of real-world image samples and high annotation costs persist. While utilizing synthetic samples offers a viable solution, significant distributional shifts exist between synthetic and real-world domains. To address these issues, this paper proposes an Unsupervised Teacher-Student Collaborative Optimization (TSCO) framework for cross-domain insulator defect detection. The proposed framework performs joint optimization across three dimensions: representation learning in the student network, cross-domain knowledge transfer, and instance-level domain alignment. Specifically, a Cross-View Invariance Enhancement module is incorporated into the student network to enhance feature representation stability via category-wise and spatial consistency constraints. Concurrently, a Low Confidence Driven Self-Refinement module is constructed within the teacher network. This module employs a self-optimization mechanism driven by low-confidence instances to guide the teacher model in progressively rectifying prediction biases, thereby improving the reliability of pseudo-labels. Furthermore, a Curriculum-guided Hard Instance Alignment module is designed to achieve robust instance-level cross-domain feature alignment by progressively filtering reliable foreground instances and intensifying the learning of hard examples. Extensive experiments on the self-constructed dataset verify the effectiveness and superiority of the proposed approach. Moreover, evaluations on public datasets (IDID and TLID) demonstrate that TSCO outperforms mainstream domain adaptation and state-of-the-art methods. Specifically, compared with the base detection model trained on manually annotated samples, AP50 for insulator damage and drop improves by 10.8% and 7.9%, respectively, while gains of 5.5% and 5.1% are achieved over the baseline model. The code implementation of the proposed framework is available at https://github.com/ssk86/UTSCO.

Insulator defect detection is a crucial task in the intelligent inspection of transmission lines. Currently, there are challenges such as insufficient image samples and difficulties in annotation. Artificially generating samples is a feasible solution, but there are discrepancies between artificial and real sample distributions. This paper proposes a Cross-domain Multi-Level Feature Alignment R-CNN network. It uses a large number of artificially labeled images and a small number of unlabeled real images as the source and target domains, respectively. Based on Faster R-CNN, an instance-level feature adaptive alignment module is constructed. Different gradient adaptive training strategies are employed for the source and target domains to better achieve cross-domain instance-level feature alignment. An image-level multi-scale local feature aggregation module is built to achieve cross-domain image-level local feature alignment. A global feature alignment module is also constructed to achieve cross-domain image-level global feature alignment. In the insulator defect detection experiment, the Average Precision at Intersection over Union of 0.5(AP50) of the proposed method is 6.8% higher than that of the baseline model, 8.6% higher than that of Faster-RCNN trained with artificial samples alone, and 1.1% higher than that of Faster-RCNN trained with real samples alone. Moreover, the proposed method achieves the highest accuracy on both self-built and public datasets, exhibiting an average improvement of 7.7% and 8.6% in AP50. respectively, compared to multiple comparable methods. The code for this paper can be found at https://github.com/n-tong/CMFAA-R-CNN.

In the detection of insulator defects on transmission lines, the detection precision is still not ideal, primarily attributed to the significant variation in target scale and complex image backgrounds. We propose the multiscale channel information (MCI)-global-local attention (GLA), a plug-in designed for YOLO series models, featuring two modules: the MCI extraction module and the GLA based on context information module (GLA-CI). MCI comprehensively extracts and utilizes multiscale feature map information, while GLA-CI captures both global context information and local spatial details, thereby augmenting the learning capability of networks. Experimental results indicate that the MCI-GLA plug-in improves the average precision (AP) of YOLOv4 to YOLOv8 models in detecting insulator breakage defects by 7.3%, 4.6%, 4.5%, 4.0%, and 5.3%, respectively. In particular, YOLOv7+MCI-GLA exhibits superior precision and inference time compared to other methods on self-constructed and public datasets. The code for this article can be found at https://github.com/falian0527/MCI-GLA .

The zero-shot image classification (ZSIC) is designed to solve the classification problem when the sample is very small, or the category is missing. A common method is to use attribute or word vectors as a priori category features (auxiliary information) and complete the domain transfer from training of seen classes to recognition of unseen classes by building a mapping between image features and a priori category features. However, feature extraction of the whole image lacks discrimination, and the amount of information of single attribute features or word vector features of categories is insufficient, which makes the matching degree between image features and prior class features not high and affects the accuracy of the ZSIC model. To this end, a spatial attention mechanism is designed, and an image feature extraction module based on this attention mechanism is constructed to screen critical features with discrimination. A semantic information fusion method based on matrix decomposition is proposed, which first decomposes the attribute features and then fuses them with the extracted word vector features of a dataset to achieve information expansion. Through the above two improvement measures, the classification accuracy of the ZSIC model for unseen images is improved. The experimental results on public datasets verify the effect and superiority of the proposed methods.

The evaluation on clustering results is an important component of clustering analysis, which can be conducted by the cluster validity index. However, the performances of most existing indices depend on not only the specific clustering algorithms but also the measurements of within-and between-cluster distances and data structures, resulting in limited applications in practice. In this paper, a new within-cluster distance under a general assumption is defined first. After adjusting within-cluster distances of each point according to the adjustment rule, a novel cluster validity index is proposed. Moreover, the notion of chain is introduced to eliminate the effects of sizes, densities, and shapes of clusters. This index does not need any prior information about clustering algorithms and is independent of data structures. Two groups of synthetic datasets with various characteristics and real-world datasets are used to validate this proposed validity index. Experimental results demonstrate that the evaluation accuracy of this index is higher than that of the existing typical indices and performs well on datasets with irregular-shaped clusters.