基于改进SSD的轻量化小目标检测算法
A lightweight small object detection algorithm based on improved SSD
查看参考文献12篇
文摘
|
为提高SSD目标检测算法的小目标检测能力,提出在SSD算法中引入转置卷积结构,采用转置卷积将低分辨率高语义信息特征图与高分辨率低语义信息特征图相融合,增加低层特征提取能力,提高SSD算法的平均精准度。同时针对SSD算法存在模型过大,运行内存占用量过高,无法在嵌入式ARM设备上运行的问题,以DenseNet为基础,结合深度可分离卷积,逐点分组卷积与通道重排提出轻量化特征提取最小单元,将SSD算法特征提取部分替换为轻量化特征提取最小单元的组合后,可在嵌入式ARM设备上运行。在PASCAL VOC数据集和KITTI自动驾驶数据集上进行对比实验,结果表明改进后的网络结构在平均精准度上得到明显提升,模型参数数量得到有效降低。 |
其他语种文摘
|
In order to improve the small object detection ability of SSD object detection algorithm,the transposed convolution structure in SSD algorithm was proposed,the low resolution high semantic information feature map was integrated with high resolution low semantic information feature map using transposed convolution,which increased the ability of low level feature extraction and improved the average accuracy of SSD algorithm.At the same time for the problem that SSD algorithm model being large,running memory consumption high,without running on the embedded equipment ARM,a lightweight feature extraction minimum unit was proposed based on DenseNet,combining depthwise separable convolutions,pointwise group convolution and channel shuffle,running on the embedded equipment ARM cloud be realized.The comparative experiments on PASCAL VOC data set and KITTI autopilot data set show that the mean average is significantly improved by improved network structure,and the number of model parameters is effectively reduced. |
来源
|
红外与激光工程
,2018,47(7):0703005-1-0703005-7 【核心库】
|
DOI
|
10.3788/IRLA201847.0703005
|
关键词
|
目标检测
;
转置卷积
;
深度可分离卷积
;
嵌入式
;
PASCAL VOC数据集
;
KITTI数据集
|
地址
|
1.
沈阳工业大学软件学院, 辽宁, 沈阳, 110870
2.
中国科学院沈阳自动化研究所, 辽宁, 沈阳, 110016
|
语种
|
中文 |
文献类型
|
研究性论文 |
ISSN
|
1007-2276 |
学科
|
自动化技术、计算机技术 |
基金
|
国家自然科学基金
;
装发部共用技术课题项目
|
文献收藏号
|
CSCD:6373938
|
参考文献 共
12
共1页
|
1.
张迪飞. 基于SVM分类的红外舰船目标识别.
红外与激光工程,2016,45(1):0104004
|
被引
5
次
|
|
|
|
2.
罗海波. 基于深度学习的目标跟踪方法研究现状与展望.
红外与激光工程,2017,46(5):0502002
|
被引
18
次
|
|
|
|
3.
Ren S. Faster R-CNN: towards real-time object detection with region proposal networks.
IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(6):1137-1149
|
被引
1012
次
|
|
|
|
4.
Redmon J. You only look once: unified, real-time object detection.
2016 IEEE Conference on Computer Vision and Pattern Recognition,2016:779-788
|
被引
172
次
|
|
|
|
5.
Liu W. SSD: single shot multiBox detector.
European Conference on Computer Vision,2016:21-37
|
被引
813
次
|
|
|
|
6.
熊昌镇. 结合主体检测的图像检索方法.
光学精密工程,2017,25(3):792-798
|
被引
11
次
|
|
|
|
7.
吴鑫. Jetson TK1平台实现快速红外图像背景预测算法.
红外与激光工程,2015,44(9):2615-2621
|
被引
4
次
|
|
|
|
8.
Lin T Y. Feature pyramid networks for object detection.
arXiv: 1612.03144[cs.CV],2016
|
被引
1
次
|
|
|
|
9.
Zhang X. Shufflenet: An extremely efficient convolutional neural network for mobile devices.
arXiv: 1707.01083[cs.CV],2017
|
被引
1
次
|
|
|
|
10.
Howard A G. Mobilenets: Efficient convolutional neural networks for mobile vision applications.
arXiv:1704.04861[cs.CV],2017
|
被引
3
次
|
|
|
|
11.
刘智. 构建多尺度深度卷积神经网络行为识别模型.
光学精密工程,2017,25(3):799-805
|
被引
21
次
|
|
|
|
12.
Huang G. Densely connected convolutional networks.
arXiv: 1608.06993[cs.CV],2016
|
被引
1
次
|
|
|
|
|