首页|面向变电站鸟害防治的轻量化鸟类目标检测方法

面向变电站鸟害防治的轻量化鸟类目标检测方法

扫码查看
近年来,随着我国生态环境的改善以及电网建设的推进,鸟类活动范围与变电站、输电线路的交集越来越大,带来的安全风险越发突出.针对传统的驱鸟方式耗时费力、效果不佳等问题,首先,通过收集鸟类图像构建了模型训练所需数据集;其次,将PicoDet模型应用于鸟类检测并提出了目标检测模型评价指标;最后,将优化后的模型进行推理验证,并对比了不同算法的性能参数.通过与常用的YOLO系列目标检测模型对比,验证了所提PicoDet模型在实现较高检测精度的同时,具有占用内存小、检测速度快的优势,适用于变电站鸟类高精度实时目标检测场景.
Lightweight Bird Target Detection Method for Bird Infestation Prevention and Control in Substation
In recent years,with the improvement of China's ecological environment and the advancement of power grid construc-tion,the intersection of bird activity ranges with substations and transmission lines has increased,leading to more prominent safety risks.To address the time-consuming and ineffective traditional bird repellent methods,we firstly collected bird images to construct a dataset for model training.Secondly,we applied the PicoDet model for bird detection and proposed evaluation metrics for the object detection model.Finally,the optimized model was inferred and validated,and its performance parameters were compared with different algorithms.By comparing with the commonly used YOLO series object detection models,the proposed PicoDet model has the advantages of high detection accuracy,small memory consumption and fast detection speed,making it suitable for high-precision real-time target detection of birds in substations.

embedded systempower visionobject detectionbird identificationPicoDet

吴艺明、张秋阳、宗成琦、张雨、马建鹏

展开 >

国网山东省电力公司青岛市即墨区供电公司,山东 青岛 266000

嵌入式系统 电力视觉 目标检测 鸟类识别 PicoDet

国网山东省电力公司科技项目

520602230008

2024

河北电力技术
河北省电机工程学会,河北省电力研究院

河北电力技术

影响因子:0.306
ISSN:1001-9898
年,卷(期):2024.43(5)