自动驾驶中基于卷积神经网络的行人检测研究  

Research on Pedestrian Detection Based on Convolutional Neural Network in Autonomous Driving

作　　者：杨康 陈丽[1] YANG Kang;CHEN Li(College of Information Science and Engineering,Shenyang University of Technology,Shenyang 110870,China)

机构地区：[1]沈阳工业大学信息科学与工程学院,辽宁沈阳110870

出　　处：《电脑知识与技术》2020年第25期22-24,30,共4页Computer Knowledge and Technology

摘　　要：针对行人检测中检测精度低和速度慢的问题,文章提出了一种改进的U-NET网络架构。为了提高系统的检测精度,采用了多尺度融合技术来构建多层卷积神经网络(CNN)。为了提高检测速度,改善了网络结构的特征通道数量,减少了大量的计算时间,以满足自动驾驶领域数据处理的实时性。在训练阶段,使用批归一化(BN)算法对每一层的输入数据进行归一化,以加快模型的收敛速度。实验结果表明,改进的网络模型可以在保证一定的检测精度的前提下,提高系统的检测速度,并满足实时性要求。In this paper,an improved U-NET network architecture is proposed to solve the problems of low detection accuracy and slow detection speed in pedestrian detection.To improve the detection accuracy of the system,a multi-scale fusion technology is used to build a multi-layer convolutional neural network(CNN).To improve the detection speed,the number of feature channels in the network framework was modified,and a large amount of calculation time was reduced to meet the real-time nature of processing data in the field of autonomous driving.In the training phase,a batch normalization(BN)algorithm is used to normalize the input data of each layer to accelerate the model's convergence speed.The experimental results show that the improved network model can improve the detection speed of the system under the premise of ensuring a certain detection accuracy,and meet the real-time requirements.

关 键 词：行人检测 多尺度融合 卷积神经网络 批量归一化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]