9 January 2023 in IEEE Transactions on Instrumentation and Measurement
As a promising and non-contact detection technique, machine vision has been widely used in fault diagnosis of substation equipment. The rapid and accurate detection of substation equipment in infrared images is one of the key steps for automatic fault diagnosis. However, the complexity of image background, the low contrast of infrared images and the rotational targets in infrared images pose a great challenge to detection task. This study aims to improve the detection accuracy of the model while having real-time detection speed, and propose a lightweight power equipment detection network (PEDNet) based on YOLOv4-tiny. First, a novel global information aggregation module is constructed to guide the network to focus on the salient regions where the target equipment is located. Second, an improved spatial transformer network is introduced to reduce the impact of rotational targets on detection accuracy. Finally, a feature enhanced fusion network is designed through the use of a multi-scale feature cross-fusion structure. It can fully fuse the feature information of the salient region, the rotational targets and the strong semantic information. The experimental results show that the proposed PEDNet can reach 92.66% detection accuracy and 107.07 FPS real-time detection speed on testing datasets. Compared with YOLOv4-tiny, there is a small sacrifice in detection speed, but the detection accuracy is improved and significantly higher than the existing state of the art (SOTA) object detection models.
9 June 2025 in IEEE Transactions on Instrumentation and Measurement
Detecting substation equipment in aerial infrared images is a critical task in automatic visual fault inspection for power systems. However, the uneven spatial distribution of objects, unmanned aerial vehicle (UAV) viewpoint variations, object scale variations, and rare computational resources pose significant challenges for substation equipment detection. This paper proposes a substation equipment detection network (SEDNet) for UAV automatic power inspection. First, a long-distance feature capture attention is proposed to guide the network to focus on learning features in important regions and to capture the information that it is prone to miss at the border of an image. Moreover, a lightweight, GS cross-stage partial structure is proposed to improve the speed of feature processing. Second, a multilayer receptive field feature enhancement module is constructed to extract diverse, fine-grained features of objects using multiple convolutional branches. Last, to effectively detect multiscale objects, we propose a concatenation and reorganization enhanced feature pyramid module. The experimental results demonstrate the effectiveness of SEDNet on a multiclass, substation equipment, infrared image dataset, achieving a detection accuracy of 99.2% and a real-time detection speed of 105.3 FPS, which outperforms state-of-the-art models in terms of mAP. SEDNet successfully meets the requirements for accurate and real-time substation equipment inspection in complex scenarios.
20 January 2024 in Journal of Advanced Computational Intelligence and Intelligent Informatics
Title: Reading Recognition Method of Mechanical Water Meter Based on Convolutional Neural Network in Natural Scenes | Keywords: water meter reading, convolution neural network, automatic meter reading, object detection | Author: Jianqi Li, Jinfei Shen, Keheng Nie, Rui Du, Jiang Zhu, and Hongyu Long