摘要
针对传统目标检测算法无法自适应提取目标相应特征并完成识别的现象,提出一种基于快速区域卷积神经网络(Faster R-CNN)模型的电器识别方法,其优势在于可以自适应获取不同场景下目标的特征,避免由于人为设计目标的特征而带来的主观因素影响,具有良好的鲁棒性与准确性。Faster R-CNN中首先通过建立区域建议网络RPN(Region Proposal Network)代替Fast R-CNN中的Selective Search方法,得到建议位置后再进行检测。为了解决训练过程当中正负样本失衡问题,在Faster R-CNN中引入了难负样本挖掘策略,增强了模型的判别能力,提高检测的精度。
Aiming at solving the problem that the traditional target detection algorithm cannot adaptively extract the corresponding features of the target,this paper proposes an electrical detection method based on Faster Region Convolutional Neural Network(Faster R-CNN)model.The advantage of the target detection and recognition method based on deep learning is that it can adaptively acquire the features of the target in different scenarios,avoiding the influence of subjective factors due to the characteristics of the artificial design target.This method has good robustness and accuracy.The method used in this paper establishes the Regional Proposal Network(RPN)to replace the selective search method in the original Fast R-CNN,which can obtains the recommended location before detecting it.In order to solve the problem of positive and negative sample imbalance during training,this paper introduces a hard example mining strategy in Faster R-CNN,which enhances the discriminative ability of the model and improves the accuracy of detection.
作者
陈从平
李游
徐道猛
邓扬
何枝蔚
CHEN Congping;LI You;XU Daomeng;DENG Yang;HE Zhiwei(College of Mechanical&Power Engineering,China Three Gorges University,Yichang 443002,China;School of Mechanical Engineering,Changzhou University,Changzhou 213164,China)
出处
《机械》
2020年第1期1-8,共8页
Machinery
基金
国家重点研发计划课题(2018YFC1903101,废线路板器件智能拆解和分选技术研究与示范)
国家自然科学基金项目(51475266,流体微挤出/堆积制备组织工程支架过程形态调控机理研究)
关键词
目标检测
识别
深度学习
神经网络
target detection
identify
deep learning
the neural network