一种受电弓电火花故障检测方法、装置、设备及存储介质
By using a fully convolutional single-stage pantograph detection model for hierarchical feature extraction and detection, the problems of insufficient robustness and poor performance in detecting minor faults in existing technologies are solved, and more efficient electric spark fault detection is achieved.
CN115797699BActive Publication Date: 2026-07-17CRRC INDUSTRAIL ACADEMY (QINGDAO) CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CRRC INDUSTRAIL ACADEMY (QINGDAO) CO LTD
- Filing Date
- 2022-12-15
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing pantograph spark fault detection technologies are not robust enough in complex environments and are prone to missed or false detections, especially for minor faults.
Method used
A fully convolutional single-stage pantograph detection model is adopted. Hierarchical feature extraction is performed through a fully convolutional backbone network and class feature pyramid. Electric spark fault detection is performed by combining dilated convolution and a detection head. The target detection model is selected by using the average accuracy of all classes.
Benefits of technology
It improves robustness in complex environments and performance in detecting minor electrical spark faults, thereby enhancing the accuracy and reliability of detection.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN115797699B_ABST
Abstract
本申请公开了一种受电弓电火花故障检测方法、装置、设备及存储介质，涉及受电弓检测领域，包括：获取受电弓摄像机采集的原始彩色图像，利用原始彩色图像对全卷积单阶段受电弓检测模型进行训练，得到训练好的检测模型；利用全类平均正确率对训练好的检测模型进行评估并筛选满足预设评估条件的目标检测模型；将待检测受电弓图像输入目标检测模型，以便利用全卷积主干网络提取待检测受电弓图像的核心特征图，基于类特征金字塔对核心特征图进行特征融合，得到不同尺度特征图，根据检测头对不同尺度特征图进行电火花故障检测以及位置定位检测。本申请采用全卷积单阶段受电弓检测模型克服了复杂环境下鲁棒性低的问题，提高对轻微电火花故障的检测性能。
Need to check novelty before this filing date? Find Prior Art