摘要
针对SSD目标检测算法在检测目标过程中存在漏检的现象,提出一种特征增强的SSD目标检测算法。该算法通过将特定连续特征层进行特征融合,获取更为丰富的目标细节特征信息,以此来改善目标的特征表达效果,提升目标检测正确率。经仿真测试,该算法对车辆、自行车和行人等道路参与者检测效果均有提高,在PASCALVOC2007数据集上的测试结果与原有SSD检测算法相比,mAP提高0.78%,适合于车载与移动机器人等场景的目标检测应用环境。
Aiming at the problem of missed detection of partial object failure in the detection process of SSD object detection algorithm, this paper proposes an enhanced feature SSD (E-SSD) object detection algorithm. The algorithm combines the features of specific continuous feature layers to obtain more abundant detailed object feature information, so as to improve the feature expression effect of the object and raise the object detection accuracy. The simulation test shows that the algorithm improves the detection effect of road participants such as vehicles, bicycles and pedestrians. Test results on the PASCAL VOC2007 dataset show mAP is increased by 0.78% compared to the original SSD detection algorithm, which is suitable for object detection application environments such as vehicles and mobile robots.
作者
毕程
杨大伟
毛琳
BI Cheng;YANG Da-wei;MAO Lin(School of Electromechanical Engineering, Dalian Minzu University, Dalian Liaoning 116605, China)
出处
《大连民族大学学报》
2019年第3期209-213,234,共6页
Journal of Dalian Minzu University
基金
辽宁省自然科学基金资助项目(20170540192,20180550866)
关键词
目标检测
SSD
特征增强
object detection
SSD
enhanced feature
