计算机技术与发展2025，Vol.35Issue(5)：205-213,9.DOI:10.20165/j.cnki.ISSN1673-629X.2024.0381
基于目标检测和易混淆分类的输电线路山火检测
Wildfire Detection of Transmission Lines Based on Object Detection and Confusion-aware Classification
摘要
Abstract
Wildfire monitoring and localization are crucial for enhancing the reliability of power transmission lines.However,image data of wildfires along high-voltage power lines possesses a certain degree of complexity,and the detected targets are usually confused.Traditional object detection algorithms cannot meet the requirements of wildfire detection in terms of both recognition accuracy and speed.Therefore,we propose a two-stage efficient wildfire detection algorithm with low false alarm rate to achieve a real-time wildfire detection system with high recall rate and high accuracy rate.In the algorithm architecture,the object detection algorithm based on positive and negative wildfire image samples is adopted in the first stage,and the neural network model YOLOv7 is used as the object de-tector.In the training,sufficient and positively and negatively balanced sample images are used to improve the reliability of the model.In the second stage,a classification algorithm for confused targets is adopted,with the lightweight neural network MobileNetV3 serving as the feature extraction network.Different classification weight parameters are selected based on the confused categories obtained in the first stage to enhance the distinguishability of these targets.In addition,the influence of training data on the wildfire object detection algorithm is analyzed,and the comparison experiment between the proposed algorithm and the mainstream object detector is carried out on the wildfire detection datasets.The effectiveness of the proposed algorithm is validated by combining the model's multi-process parallel inference frame rate.关键词
输电线路/山火检测/烟雾检测/图像处理/目标检测Key words
transmission line/wildfire detection/smoke detection/image processing/object detection分类
信息技术与安全科学引用本文复制引用
苏仁斌,熊卫红,夏闻远,周孟霜,周庆,刘先珊..基于目标检测和易混淆分类的输电线路山火检测[J].计算机技术与发展,2025,35(5):205-213,9.基金项目
国家电网公司华中分部科技项目(52140023000A) （52140023000A）
