基于DA-SSD的有轨电车轨道小目标障碍物检测算法  被引量:1

Detection of Small Target Obstacles on Tram Tracks Based on DA-SSD Algorithm

在线阅读下载全文

作  者:王运明 彭国都 周奕昂 李卫东[1] WANG Yunming;PENG Guodu;ZHOU Yiang;LI Weidong(School of Automation and Electrical Engineering,Dalian Jiaotong University,Dalian 116028,China;State Engineering Technology Center of CRRC Changchun Railway Vehicle Co.,Ltd,Changchun 130000,China)

机构地区:[1]大连交通大学自动化与电气工程学院,辽宁大连116028 [2]中车长春轨道客车股份有限公司、国家工程技术中心,吉林长春130000

出  处:《大连交通大学学报》2023年第2期108-114,共7页Journal of Dalian Jiaotong University

基  金:辽宁省教育厅基本科研资助项目(LJKMZ20220857);辽宁省科学技术计划资助项目(2021-BS-219);大连市科技创新基金项目(2021JJ13SN82)。

摘  要:城市有轨电车轨道障碍物的高精度、快速检测对保障城市有轨电车安全行驶具有重要意义。针对SSD算法检测轨道小目标障碍物精度较低的问题,提出了基于DA-SSD的城市有轨电车轨道小目标障碍物检测算法。在SSD目标检测算法的基础上,设计低层双段反卷积模块,丰富低层特征层的语义信息,增加自适应注意力机制模块,生成具有更强语义信息和精确位置信息的低层特征预测层,修正先验框生成方式,缩小各个特征层先验框的大小,增强轨道小目标障碍物检测的适应性。通过自制有轨电车轨道障碍物数据集进行训练与测试。结果表明:当Riou=0.6时,DA-SSD算法的MAP达到78.17%,检测速度为23.4 f/s,相比SSD算法,该算法在保持高速检测的前提下,提高了有轨电车小目标障碍物的检测精度。The high-precision and rapid detection of urban tram track obstacles is of great significance to ensuring the safe driving of urban trams.Aiming at the problem that the SSD algorithm has low accuracy in detecting small target obstacles on tracks,a detection algorithm for small target obstacles on urban tram tracks is proposed based on DA-SSD.A low-level two-segment deconvolution module is designed to enrich the semantic information of the low-level feature layer based on the SSD object detection algorithm.In order to generate a low-level feature prediction layer with stronger semantic information and precise location information,the a-daptive attention mechanism module is added.Finally,a priori frame generation method is modified in order to enhance the adaptability of track small target obstacle detection,and the size of the prior frame of each feature layer is reduced.Through the training and testing of the self-made tram track obstacle dataset,the results show that the mAP of the DA-SSD algorithm reaches 7&17%,and the detection speed is 23.4 f/s.Under the premise of maintaining high-speed detection,the algorithm improves the detection accuracy of small target obstacles on trams.

关 键 词:城市有轨电车 小目标检测 SSD 双段反卷积 自注意力机制 

分 类 号:U482.1[交通运输工程—载运工具运用工程] TP391.41[交通运输工程—道路与铁道工程]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象