基于改进的SSD监理目标检测研究

doi:10.16180/j.cnki.issn1007-7820.2022.05.002

摘要/Abstract

摘要：

针对装饰装修工程中由人工验收带来的诸多问题,文中提出了一种改进的SSD算法并将其应用于监理工作来代替人工验收,推动智能监理的实现。由于SSD算法存在对同一目标复检以及小目标检测效果欠佳等问题,故文中利用DPN网络替换基础特征提取网络VGG16。DPN结合了Resnet和Densenet的优点,具有更好的特征提取能力。通过加权FPN融合特征图,突出不同层特征图的贡献,丰富用于预测的特征图语义。利用深度可分离卷积降低模型的参数量,提高算法的推理速度。实验对比发现,改进后模型的平均精度提升了3.47%,对小数目检测平均精度的提升可达15%,证明新模型在监理目标检测任务中效果良好。

关键词: 监理, SSD, VGG16, DPN, Resnet, Densenet, 深度可分离卷积, FPN

Abstract:

In view of many problems caused by manual acceptance in decoration projects, this study proposes an improved SSD algorithm and applies it to supervision work to replace manual acceptance and promote the realization of intelligent supervision. Because the SSD algorithm has problems such as rechecking the same target and poor detection of small targets, the DPN network is employed to replace the basic feature extraction network VGG16. DPN combines the advantages of Resnet and Densenet, and has better feature extraction capabilities. Feature maps are fused by weighted FPN to highlight the contributions of feature maps of different layers and enrich the semantics of feature maps for prediction. Using depth separable convolution can reduce the amount of model parameters and improve the inference speed of the algorithm. Experimental comparison shows that the average accuracy of the improved model is increased by 3.47%, and the average accuracy of small numbers of detection is increased by up to 15%, which proves that the new model is effective in the task of supervision target detection.

Key words: supervision, SSD, VGG16, DPN, Resnet, Densenet, depth separable convolution, FPN

中图分类号:

TP311.1

黄静,谢宣. 基于改进的SSD监理目标检测研究[J]. 电子科技, 2022, 35(5): 7-13.

HUANG Jing,XIE Xuan. Research on Supervision Object Detection Based on Improved SSD[J]. Electronic Science and Technology, 2022, 35(5): 7-13.

图/表 10

图1

图2

图3

图4

图5

表1

表2

图6

表3

图7

参考文献 21

[1]	杨晓丽, 刘博涛, 赵哲, 等. 基于视觉与激光雷达数据融合的前车识别研究[J]. 汽车实用技术, 2020, 45(18):22-24.
	Yang Xiaoli, Liu Botao, Zhao Zhe, et al. Front vehicle recognition based on fusion of vision and lidar data[J]. Automobile Applied Technology, 2020, 45(18):22-24.
[2]	姚启才, 汪地, 廖茂生. 基于机器视觉的机械臂智能抓取系统设计[J]. 计量与测试技术, 2020, 47(10):28-33.
	Yao Qicai, Wang Di, Liao Maosheng. Intelligent grasping system design of robot arm based on machine vision[J]. Metrology & Measurement Technique, 2020, 47(10):28-33.
[3]	张达峰, 刘宇红, 张荣芬. 基于深度学习的智能辅助驾驶系统[J]. 电子科技, 2018, 31(10):60-63.
	Zhang Dafeng, Liu Yuhong, Zhang Rongfen. Intelligent assistant driving system based on deep learning[J]. Electronic Science and Technology, 2018, 31(10):60-63.
[4]	陶映帆, 胡鹏飞, 杨文明. 智能家居中的计算机视觉技术[J]. 人工智能, 2020(5):30-38.
	Tao Yingfan, Hu Pengfei, Yang Wenming. Computer vision technology in smart home[J]. AI-View, 2020(5):30-38.
[5]	胡振寰. 基于深度学习算法的遮挡行人检测[D]. 柳州: 广西科技大学, 2019.
	Hu Zhenhuan. Occlusion pedestrian detection based on deep learning algorithm[D]. Liuzhou: Guangxi University of Science and Technology, 2019.
[6]	周文凯, 韩芳, 孔维健. 基于Faster-RCNN的极验点选式验证码识别[J]. 电子科技, 2019, 32(9):42-45.
	Zhou Wenkai, Han Fang, Kong Weijian. Point-selective geetest CAPTCHA recognition based on Faster-RCNN[J]. Electronic Science and Technology, 2019, 32(9):42-45.
[7]	李光曜. 基于深度学习的遥感图像目标检测方法研究[D]. 石家庄: 石家庄铁道大学, 2019.
	Li Guangyao. Target detection method of remote sensing image based on deep learning[D]. Shijiazhuang: Shijiazhuang Tiedao University, 2019.
[8]	杨璐. 基于SSD的人脸检测算法研究[D]. 镇江: 江苏科技大学, 2019.
	Yang Lu. Research on face detection algorithm based on SSD algorithm[D]. Zhenjiang: Jiangsu University of Science and Technology, 2019.
[9]	吴华运, 任德均, 付磊, 等. 基于改进型SSD算法的空瓶表面缺陷检测[J]. 计算机与现代化, 2020(4):121-126.
	Wu Huayun, Ren Dejun, Fu Lei, et al. Surface defect detection of empty bottles based on improved SSD algorithm[J]. Computer and Modernization, 2020(4):121-126.
[10]	刘学, 李范鸣, 刘士建. 改进的SSD红外图像行人检测算法[C]. 上海: 上海市红外与遥感学会, 2019.
	Liu Xue, LiFanming, LiuShijian. Improved SSD infrared image pedestrian detection algorithm[C]. Shanghai: Proceedings of the Shanghai Institute of Infrared and Remote Sensing, 2019.
[11]	苟先太, 黄巍, 刘琪芬. 基于改进SSD模型的牛脸检测[J]. 计算机工程与设计, 2020, 41(3):833-837.
	Gou Xiantai, Huang Wei, Liu Qifen. Cattle face detection based on improved SSD model[J]. Computer Engineering and Design, 2020, 41(3):833-837.
[12]	Zhai S, Shang D, Wang S, et al. DF-SSD: An improved SSD object detection algorithm based on densenet and feature fusion[J]. IEEE Access, 2020(8):24344-24357.
[13]	Li Y L, Wang S. HAR-Net: Joint learning of hybrid attention for single-stage object detection[J]. IEEE Transactions on Image Processing, 2020, 29(11):3092-3103. doi: 10.1109/TIP.2019.2957850
[14]	沈新烽, 姜平, 周根荣. 改进SSD算法在零部件检测中的应用研究[J]. 计算机工程与应用, 2020, 56(12):1-10.
	Shen Xinfeng, Jiang Ping, Zhou Genrong. Application of improved SSD algorithm in parts detection[J]. Computer Engineering and Applications, 2020, 56(12):1-10.
[15]	Kanimozhi S, Gayathri G, Mala T. Multiple real-time object identification using Single shot multi-box detection[C]. Chennai: Proceedings of the 2019 International Conference on Computational Intelligence in Data Science, 2019.
[16]	郭川磊. 基于转置卷积操作改进的SSD目标检测算法研究[D]. 成都: 成都信息工程大学, 2019.
	Guo Chuanlei. Research on improved SSD target detection algorithm based on transposed convolution operation[D]. Chengdu: Chengdu University of Information Technology, 2019.
[17]	焦鑫, 杨伟东, 刘全周, 等. 用于ADAS实时目标车辆检测的改进SSD算法[J]. 汽车安全与节能学报, 2020, 11(3):337-344.
	Jiao Xin, Yang Weidong, Liu Quanzhou, et al. Improved SSD algorithm for real-time target vehicle detection in ADAS[J]. Journal of Automotive Safety and Energy, 2020, 11(3):337-344.
[18]	黎阳, 沈烨, 刘敏, 等. 融合运动信息与表观信息的多目标跟踪算法[J]. 电子科技, 2020, 33(9):21-24.
	Li Yang, Shen Ye, Liu Min, et al. Multi-target tracking algorithm by combining motion information and apparent information[J]. Electronic Science and Technology, 2020, 33(9):21-24.
[19]	He K, Zhang X, Ren S, et al. Identity mappings in deep residual networks[C]. Amsterdam: European Conference on Computer Vision, 2016.
[20]	Huang G, Liu Z, Laurens V, et al. Densely connected convolutional networks[C]. Honolulu: IEEE Conference on Computer Vision and Pattern Recognition, 2017.
[21]	Chen Y, Li J, Xiao H, et al. Dual path networks[C]. New York: Proceedings of the Thirty-first International Conference on Neural Information Processing Systems, 2017.

	Input_ shape	Output_ shape	Out_chan nels	Sub_blo ck_num	Kernel
Conv2d	512	256	16	1	3×3
DPN_Block1	256	128	16	1	3×3
DPN_Block2	128	64	32	2	3×3
DPN_Block3	64	64	48	3	3×3
DPN_Block4	64	64	64	3	5×5
DPN_Block5	64	32	96	4	5×5
DPN_Block6	32	16	128	4	5×5
Dept_Block7	16	8	192	1	5×5
Dept_Block8	8	4	320	1	3×3
Dept_Block9	4	2	512	1	3×3
Max_Pooling	2	1	1 024	1	-

名称	型号
OS	Centos7.8
CPU	i9-7900X
GPU	GTX 2080Ti
Python	3.7.6
Keras	2.3.1
Tensorflow	1.15

类别	模型名称	F₁	Recall	Precision
地漏	经典SSD	0.55	40.00%	85.71%
地漏	改进的SSD	0.64	46.67%	100.00%
莲蓬头	经典SSD	0.74	60.53%	95.83%
莲蓬头	改进的SSD	0.87	76.32%	100.00%