摘要
近年来,计算机视觉方向深度学习的卷积神经网络发展迅速并逐步应用于日常生活的检测之中。针对于摩托车手是否佩戴头盔的检测问题,论文引入RFB网络,采用VGG-16作为基础网络提取网络结构特征,改进使用了余弦衰减学习率更好地训练样本,增加了网络泛化能力。从而完成对摩托车头盔的检测。实验结果表明改进后,RFB-Net模型在摩托车驾驶人头盔检测中精度较高、速度较快,且具有较好的推广性。
In recent years, the convolutional neural network of computer vision direction deep learning develops rapidly and is gradually applied to the detection of daily life.In order to detect whether motorcycle drivers wear helmets or not, this paper introduces RFB network, uses vgg-16 as the basic network to extract network structure features, improves the use of cosine decay learning rate to better train samples, and increases the network generalization ability.Ultimately the test of motorcycle helmet was completed.The experimental results show that the improved RFB net model has higher accuracy, faster speed and better generalization in motorcycle driver helmet detection.
作者
刘琛
王江涛
LIU Chen;WANG Jiang-tao(College of Physics and Electronic Information,Huaibei Normal University,Anhui Huaibei 235000,China)
出处
《太原科技大学学报》
2021年第6期496-500,共5页
Journal of Taiyuan University of Science and Technology
基金
国家自然科学基金(61203272)
安徽省高校自然科学研究重大项目(KJ2018ZD038)
安徽省高等学校省级质量工程项目(2017kfk043,2019jxtd142)。
