您的位置:山东大学 -> 科技期刊社 -> 《山东大学学报(工学版)》

山东大学学报 (工学版) ›› 2020, Vol. 50 ›› Issue (1): 28-34.doi: 10.6040/j.issn.1672-3961.0.2019.179

• 控制科学与工程 • 上一篇    下一篇

基于空间隐患分布与运动意图解析的危险评估方法

赵越男1(),陈桂友1,*(),孙琛1,卢宁1,廖立伟2   

  1. 1. 山东大学控制科学与工程学院, 山东 济南 250061
    2. 北京大学软件与微电子学院, 北京 100871
  • 收稿日期:2019-04-19 出版日期:2020-02-20 发布日期:2020-02-14
  • 通讯作者: 陈桂友 E-mail:201734497@mail.sdu.edu.cn;chenguiyou@sdu.edu.cn
  • 作者简介:赵越男(1993-),男,山东枣庄人,硕士研究生,主要研究方向为计算机视觉,机器学习与模式识别. E-mail:201734497@mail.sdu.edu.cn
  • 基金资助:
    山东省重点研发计划项目(2017GGX90105);山东省自然科学基金(ZR2017MF014)

Risk assessment method based on spatial hidden danger distribution and motion intention analysis

Yuenan ZHAO1(),Guiyou CHEN1,*(),Chen SUN1,Ning LU1,Liwei LIAO2   

  1. 1. School of Control Science and Engineering, Shandong University, Jinan 250061, Shandong, China
    2. School of Software and Microelectronics, Peking University, Beijing 100871, China
  • Received:2019-04-19 Online:2020-02-20 Published:2020-02-14
  • Contact: Guiyou CHEN E-mail:201734497@mail.sdu.edu.cn;chenguiyou@sdu.edu.cn
  • Supported by:
    山东省重点研发计划项目(2017GGX90105);山东省自然科学基金(ZR2017MF014)

摘要:

针对智能监控领域中存在的目标检测速度慢、安全性分析缺失等问题,提出一种基于空间隐患分布与运动意图解析的危险评估方法。使用k-means++及背景消除法增强YOLO(you only look once)v3算法的目标检测能力,完成人体目标和危险物的识别及定位;利用卡尔曼滤波预测人体目标的移动轨迹,以移动偏向角及人体目标与危险物的距离构建人的运动模式;建立危险度评估模型,依据不同的运动行为模式评估其危险程度。试验结果显示,增强YOLOv3算法在测试集上对各类目标的检测精准度与召回率均超过95%,交并比(intersection over union, IOU)提升7%,帧率达31.3帧/s,满足系统的实时性要求;本研究提出的危险度评估方法能够较好拟合多种运动模式的风险递变规律,危险度评估结果融合了看护目标的运动意图,使得对不同运动模式的安全性评估更加合理。

关键词: 智能监控, YOLOv3, 卡尔曼滤波, 空间隐患, 运动意图, 危险度评估

Abstract:

To solve problems of slow detection and lack of behavioral safety analysis in intelligent nursing, a risk assessment method based on spatial hidden danger distribution and motion intention analysis was proposed. The k-means++ algorithm and background elimination method were used to enhance the object detection capability of YOLO(you only look once) v3, which led to the classification and localization of human targets as well as dangerous objects. The Kalman filter was used to predict the moving trajectory, where two parameters, namely the deviation angle of motion and the distance between the human target and danger, were used to construct the human movement patterns. The risk assessment model was established, and the degree of danger was evaluated according to different movement behavior patterns. Experimental results showed that for identifying different objects in the test set, both the detection precision and the recall rate of the enhanced YOLOv3 algorithm were over 95%. An increasement of IOU(intersection over union) at 7% was witnessed, and frames rate reached 31.3 frames/s. These results proved the real-time performance of the system. Since the proposed risk assessment model incorporated motion intentions of the human target, this method was expected to boost the performance in fitting the risk progression of different movement patterns, making the risk assessment more reasonable.

Key words: intelligent surveillance, YOLOv3, Kalman filter, spatial hidden danger, motion intention, risk assessment

中图分类号: 

  • TP391.4

图1

YOLOv3结构图"

图2

算法工作流程"

图3

聚类效果分析"

图4

改进的YOLOv3目标检测效果"

图5

IOU箱线分布图"

表1

不同算法框架对比"

不同框架 平均IOU 精度 召回率 帧率/(帧·s-1)
YOLOv3-416[15] 0.71 0.95 0.95 33.4
SSD-321[13] 0.65 0.89 0.91 16.4
Faster-RCNN[12] 0.80 0.94 0.95 6.6
Casc-RCNN[20] 0.82 0.96 0.96 8.0
增强YOLOv3-416 0.78 0.96 0.97 31.3

图6

真实轨迹、预测轨迹对比"

图7

参数kd影响分析"

图8

参数σ影响分析"

表2

三种交互方式危险度变化情况"

危险情形 d/pix θ R/%
Cl 410. 17 32. 35 34.04
C2 375. 20 23. 83 40. 90
C3 339. 27 22. 63 48. 09
C4 293.44 14.37 58.08
C5 243.51 0. 16 69. 03
C6 186. 98 23. 48 79. 27
Cl 143. 70 7.91 87. 75
C8 104. 97 9.60 93. 13
A1 149. 29 114. 18 62.80
A2 130. 50 126. 43 60.29
A3 158. 10 148. 30 49. 36
A4 187.02 126. 88 53.74
A5 216. 72 144. 73 44. 17
A6 251.81 125. 33 45.43
A7 238. 58 176. 94 27. 66
A8 303. 30 128. 75 37. 18
B1 292.49 44. 24 55.79
B2 262.21 42.09 62. 25
B3 231.67 49. 05 67.33
B4 204. 43 59. 19 70. 55
B5 183.45 66. 42 72. 57
B6 168.08 79.21 71.64
B7 159.23 98. 25 67.05
B8 163. 83 111.08 62. 11
B9 182. 95 125. 36 54.77
BIO 219. 62 141.77 44. 76
Bll 256. 68 146. 49 38. 74
B12 284. 98 164. 52 30. 60

图9

危险运动模式报警"

1 黄凯奇, 陈晓棠, 康运锋, 等. 智能视频监控技术综述[J]. 计算机学报, 2015, 38 (6): 1093- 1118.
HUANG Kaiqi , CHEN Xiaotang , KANG Yunfeng , et al. Intelligent visual surveillance: a review[J]. Chinese Journal of Computers, 2015, 38 (6): 1093- 1118.
2 尹宏鹏, 陈波, 柴毅, 等. 基于视觉的目标检测与跟踪综述[J]. 自动化学报, 2016, 42 (10): 1466- 1489.
YIN Hongpeng , CHEN Bo , CHAI Yi , et al. Vision-based object detection and tracking: a review[J]. Acta Automatica Scinica, 2016, 42 (10): 1466- 1489.
3 LI Ce , HAN Zhenjun , YE Qixiang , et al. Visual abnormal behavior detection based on trajectory sparse reconstruction analysis[J]. Neurocomputing, 2013, 119 (16): 94- 100.
4 LEE D G , SUK H I , PARK S K , et al. Motion influence map for unusual human activity detection and localization in crowded scenes[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2015, 25 (10): 1- 1.
doi: 10.1109/TCSVT.2015.2481139
5 COSAR S , DONATIELLO G , BOGORNY V , et al. Towards abnormal trajectory and event detection in video surveillance[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2017, 27 (3): 683- 695.
doi: 10.1109/TCSVT.2016.2589859
6 XIE Dan , SHU Tianmin , TODOROVIC S , et al. Learning and inferring "Dark Matter" and predicting human intents and trajectories in videos[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40 (7): 1639- 1652.
doi: 10.1109/TPAMI.2017.2728788
7 AHMED S A , DOGRA D P , KAR S , et al. Surveillance scene representation and trajectory abnormality detection using aggregation of multiple concepts[J]. Expert Systems with Applications, 2018, 101, 43- 55.
doi: 10.1016/j.eswa.2018.02.013
8 REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, USA: IEEE Press, 2016: 779-788.
9 FELZENSZWALB P, MCALLESTER D, RAMANAN D. A discriminatively trained, multiscale, deformable part model[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Anchorage, USA: IEEE Press, 2008: 1-8.
10 GIRSHICK R, DONAHUE J, DARREL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Columbus, USA: IEEE Press, 2014: 580-587.
11 GIRSHICK R. Fast R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision. Santiago, Chile: IEEE Press, 2015: 1440-1448.
12 REN Shaoqing , HE Kaiming , GIRSHICK R , et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39 (6): 1137- 1149.
doi: 10.1109/TPAMI.2016.2577031
13 LIU Wei, ANGUELOV D, ERHAN D, et al. SSD: single shot multiBox detector[C]//Proceedings of European Conference on Computer Vision. Amsterdam, Holland: Springer Cham, 2016: 21-37.
14 LIN T Y , GOYAL P , GIRSHICK R , et al. Focal loss for dense object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, (99): 2999- 3007.
15 REDMON J, FARHADI A. YOLOv3: an incremental improvement[EB/OL]. (2018-04-08) [2018-09-27]. https://arxiv.org/pdf/1804.02767.pdf.
16 HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, USA: IEEE Press, 2016: 770-778.
17 ZHAO Xia, NI Yingting, JIA Haihang. Modified object detection method based on YOLO[C]//Proceedings of Chinese Conference on Computer Vision. Singapore: Springer, 2017: 233-244.
18 ARTHUR D, VASSILVITSKII S. K-means++: the advantages of careful seeding[C]//Proceedings of Eighteenth Acm-siam Symposium on Discrete Algorithms. New Orleans, USA: Society for Industrial and Applied Mathematics, 2007: 1027-1035.
19 REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]//Proceedings of Conference on Computer Vision and Pattern Recognition. Honolulu, USA: IEEE Press, 2017: 6517-6525.
20 CAI Zhaowei, VASCONCELOS N. Cascade R-CNN: delving into high quality object detection[C]//Proceedings of Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE Press, 2018: 6154-6162.
[1] 董满,刘淑琴. 基于UKF和AH法的磁悬浮人工心脏泵用锂电池SOC估计复合算法[J]. 山东大学学报(工学版), 2018, 48(2): 121-127.
[2] 邓俊武,张玉民,张红娣,杜晓坤. X尾翼无人机的故障诊断和容错控制方法[J]. 山东大学学报(工学版), 2017, 47(5): 166-172.
[3] 黄洁,何潇. 基于两阶段TKF的测量死区下离散时间系统的故障估计[J]. 山东大学学报(工学版), 2017, 47(5): 136-142.
[4] 李洪阳,何潇. 基于SCKF方法的非线性随机动态系统故障诊断方法[J]. 山东大学学报(工学版), 2017, 47(5): 130-135.
[5] 兰义华,任浩征*,张勇,赵雪峰. 一种基于“当前”模型的改进卡尔曼滤波算法[J]. 山东大学学报(工学版), 2012, 42(5): 12-17.
[6] 丁信忠1,张承瑞1*,李虎修1,于乐华2. 永磁同步电机的转动惯量辨识及状态估计[J]. 山东大学学报(工学版), 2012, 42(2): 70-76.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
[1] 王素玉,艾兴,赵军,李作丽,刘增文 . 高速立铣3Cr2Mo模具钢切削力建模及预测[J]. 山东大学学报(工学版), 2006, 36(1): 1 -5 .
[2] 张永花,王安玲,刘福平 . 低频非均匀电磁波在导电界面的反射相角[J]. 山东大学学报(工学版), 2006, 36(2): 22 -25 .
[3] 施来顺,万忠义 . 新型甜菜碱型沥青乳化剂的合成与性能测试[J]. 山东大学学报(工学版), 2008, 38(4): 112 -115 .
[4] 李梁,罗奇鸣,陈恩红. 对象级搜索中基于图的对象排序模型(英文)[J]. 山东大学学报(工学版), 2009, 39(1): 15 -21 .
[5] 陈瑞,李红伟,田靖. 磁极数对径向磁轴承承载力的影响[J]. 山东大学学报(工学版), 2018, 48(2): 81 -85 .
[6] 浦剑1 ,张军平1 ,黄华2 . 超分辨率算法研究综述[J]. 山东大学学报(工学版), 2009, 39(1): 27 -32 .
[7] 秦通,孙丰荣*,王丽梅,王庆浩,李新彩. 基于极大圆盘引导的形状插值实现三维表面重建[J]. 山东大学学报(工学版), 2010, 40(3): 1 -5 .
[8] 刘文亮,朱维红,陈涤,张泓泉. 基于雷达图像的运动目标形态检测及跟踪技术[J]. 山东大学学报(工学版), 2010, 40(3): 31 -36 .
[9] 王丽君,黄奇成,王兆旭 . 敏感性问题中的均方误差与模型比较[J]. 山东大学学报(工学版), 2006, 36(6): 51 -56 .
[10] 孙殿柱,朱昌志,李延瑞 . 散乱点云边界特征快速提取算法[J]. 山东大学学报(工学版), 2009, 39(1): 84 -86 .