| 改进的SSD行人检测算法 |
| |
| 引用本文： | 姜敏,王力,王冬冬.改进的SSD行人检测算法[J].软件,2020(2):57-61,74. |
| |
| 作者姓名： | 姜敏 王力 王冬冬 |
| |
| 作者单位： | ;1.贵州大学大数据与信息工程学院;2.贵州工程应用技术学院信息工程学院 |
| |
| 基金项目： | 贵州省教育厅创新群体重大研究项目，黔教合KY字[2016]057;国家新工科实践项目，黔教高涵[2018]209号 |
| |
| 摘 要： | 针对行人检测中检测速度慢,不能实现实时性检测的问题,提出一种改进的SSD(Single Shot MultiBox Detector)行人检测算法。改进网络通过调整基础网络中卷积层的数量,去除冗余的卷积层,降低模型复杂度,提高检测速度;不同尺度特征图进行预测之前加入残差块,进一步提取特征,提高准确率。提取PASCAL VOC数据集中的行人图像和INRIA数据集形成混合数据集进行训练,增加模型泛化性,实验证明本方法拥有较高的精度和较快的速度,具有良好的泛化性,满足实时性要求。
|
| 关 键 词： | 行人检测 单发多框检测器 卷积神经网络 深度残差网络 深度学习 |
Improved SSD Pedestrian Detection Algorithms |
| |
| JIANG Min,WANG Li,WANG Dong-dong.Improved SSD Pedestrian Detection Algorithms[J].Software,2020(2):57-61,74. |
| |
| Authors: | JIANG Min WANG Li WANG Dong-dong |
| |
| Affiliation: | (College of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China;School of Information Engineering,Guizhou University of Engineering Science,Bijie 551700,China) |
| |
| Abstract: | To solve the problem that the detection speed is too slow to realize real-time detection in pedestrian detection,an improved SSD(Single Shot Multibox Detector)pedestrian detection algorithm is proposed.By adjusting the number of convolution layers in the basic network and removing the redundant convolution layers,the complexity of the model can be reduced and the detection speed can be improved.Residual blocks are added before different scale feature maps are predicted to further extract features and improve the accuracy.The person images of PASCAL VOC dataset are combined with INRIA dataset to form a mixed dataset for training,and the generalization of the model is increased.Experiments show that the method has high accuracy and speed,good generalization and real-time requirements. |
| |
| Keywords: | Pedestrian detection SSD Convolutional neural network Deep residual network Deep learning |
| 本文献已被 维普 等数据库收录！ |
|
