基于红外图像和逆向投影算法的室内人体跌倒检测方法

陈涵; 余磊; 彭泗田; 聂宏; 欧巧凤; 熊邦书

基于红外图像和逆向投影算法的室内人体跌倒检测方法

陈涵^1,,
余磊^1, ,,
彭泗田¹,
聂宏²,
欧巧凤¹,
熊邦书¹

1.
南昌航空大学图像处理与模式识别江西省重点实验室，江西南昌 330063
2.
北爱荷华大学技术系，美国爱荷华州 50614-0178

基金项目:

国家自然科学基金 62162044

江西省自然科学基金 20202BAB202016

南昌航空大学研究生创新专项基金 YC2019030

详细信息

作者简介:
陈涵（1996-），男，硕士研究生，研究方向：基于红外图像的人体姿态识别。E-mail：chenhan_1996@163.com

通讯作者:
余磊（1984-），男，副教授，主要从事图像处理及应用。E-mail：yulei@nchu.edu.cn

中图分类号: TP391.4
计量
- 文章访问数: 499
- HTML全文浏览量: 86
- PDF下载量: 63
- 被引次数: 0
出版历程
- 收稿日期: 2020-11-10
- 修回日期: 2021-01-12
- 刊出日期: 2021-10-20

Indoor Human Fall Detection Method Based on Infrared Images and Back-Projection Algorithm

1.
Key Laboratory of Image Processing and Pattern Recognition of Jiangxi Province Nanchang Hangkong University, Nanchang 330063, China
2.
Department of Technology, University of Northern Iowa, Iowa, 50614-0178, USA

摘要

摘要: 研究表明跌倒是我国老年人伤害的主要原因，缩短跌倒到救治的时间能降低跌倒造成的伤害。为此，室内老年人跌倒检测需求逐年增加。红外传感器具有受光照影响小，保护隐私等优点，越来越广泛地应用于室内人体跌倒检测中。然而，由于红外图像存在分辨率低、信噪比差等缺陷，导致传统方法的检测精度较低。针对这个问题，本文提出一种基于逆向投影算法的室内人体跌倒检测方法。首先，通过人体温度计算出人体与传感器之间的距离；其次，结合图像信息，逆推出人体在真实世界的高度；最后，对获取的人体真实高度数据进行平滑处理，并根据其变化情况进行跌倒检测。实验结果表明，本文所提方法的检测准确率达到98.57%，优于传统非逆向投影方法，其性能完全可以应用于实际检测中。
- 人体跌倒检测 /
- 逆向投影算法 /
- 红外图像 /
- 人体定位 /
- 特征平滑
Abstract: Falls are reported to be a major cause of injury in China's elderly population. Shortening the time between the fall and subsequent treatment can reduce injuries caused by falls; therefore, the demand for indoor fall detection is increasing annually. Infrared image-based human fall detection methods are becoming increasingly popular owing to advantages such as being unaffected by light and non-intrusive. However, the traditional methods perform low accuracy because it is difficult to extract the features from the infrared videos at low resolution and high noise. Hence, this paper proposes an indoor human fall detection method based on a back-projection algorithm. First, the distance between the human body and the sensor is calculated using the human body temperature. Second, the height of the human body in the real world is reversely deduced using image information. Finally, the human body height data are smoothed and used for fall detection based on the height variation. The experimental results show that the detection accuracy of the proposed method is 98.57%, which is better than that of traditional projection methods. Therefore, it can be used for detection in real-life situations.
- human fall detection /
- back projection algorithm /
- infrared image /
- human localization /
- feature smoothing

HTML全文

图 1 屋顶传感器采集实验结果图

Figure 1. Experimental results collected by the sensor on the roof

下载: 全尺寸图片幻灯片

图 2 墙侧传感器安装示意图

Figure 2. Schematic diagram of sensor installation on the wall

下载: 全尺寸图片幻灯片

图 3 墙角传感器安装示意图

Figure 3. Schematic diagram of sensor installation in the corner

下载: 全尺寸图片幻灯片

图 4 跌倒检测算法流程图

Figure 4. Flow chart of human fall detection algorithm

下载: 全尺寸图片幻灯片

图 5 红外图像及其相关处理

Figure 5. Thermal images and processing

下载: 全尺寸图片幻灯片

图 6 人体定位示意图

Figure 6. Human body positioning diagram

下载: 全尺寸图片幻灯片

图 7 两种行为过程示意图

Figure 7. Diagrams of two behavioral processes

下载: 全尺寸图片幻灯片

图 8 两种行为人体像素高度变化

Figure 8. Changes in human pixel height under two behaviors

下载: 全尺寸图片幻灯片

图 9 逆向投影整体示意图

Figure 9. Overall schematic diagram of back projection

下载: 全尺寸图片幻灯片

图 10 测量点位置示意图

Figure 10. Schematic diagram of measuring point location

下载: 全尺寸图片幻灯片

图 11 图像单位转换示意图

Figure 11. Schematic diagram of image unit conversion

下载: 全尺寸图片幻灯片

图 12 逆向投影局部示意图

Figure 12. Partial schematic diagram of back projection

下载: 全尺寸图片幻灯片

图 13 两种行为人体真实高度变化

Figure 13. Changes in human real height under two behaviors

下载: 全尺寸图片幻灯片

图 14 平滑后两种行为人体真实高度变化

Figure 14. Changes in human pixel height after smoothing under two behaviors

下载: 全尺寸图片幻灯片

图 15 行走路线示意图

Figure 15. Schematic diagram of walking routes

下载: 全尺寸图片幻灯片

图 16 跌倒方向示意图

Figure 16. Schematic diagram of falling directions

下载: 全尺寸图片幻灯片

表 1 行走实验数据

Table 1. Walking experiment data

Route	Human body height change by non-back projection algorithm Δ_w1/pixel			Human body height change by back projection algorithm Δ_w2/mm
Route	Minimum	Maximum	Average value	Minimum	Maximum	Average value
1	6	8	7	43	165	106
2	9	11	10	48	187	118
3	12	14	13	52	223	138
4	8	10	9	45	182	115
5	5	7	6	41	158	102

下载: 导出CSV

表 2 跌倒实验数据

Table 2. Fall experiment data

Route	Human body height change by non-back projection algorithm Δ_f1/pixel			Human body height change by back projection algorithm Δ_f1/mm
Route	Minimum	Maximum	Average value	Minimum	Maximum	Average value
1	17	20	18	991	1475	1233
2	16	18	17	986	1465	1128
3	12	15	13	886	1174	1032
4	15	17	16	975	1253	1114
5	17	19	18	988	1484	1238

下载: 导出CSV

表 3 动作要求

Table 3. Action requirements

Fall action	Fall down
Non-fall action	Walk back and forth
	Sit on the chair
	Standing
	Lying in bed
	Squat fast

下载: 导出CSV

表 4 检测情况混淆矩阵

Table 4. Confusion matrix of detection results

Real state	Predictor state
Real state	Fall	Non-fallng
Fall	79	1
Walk	0	40
Sit	0	40
Stand	0	40
Lay	0	40
Squat	3	37

下载: 导出CSV

表 5 本文方法与其他方法对比结果

Table 5. Results comparison

	Method of this paper	Reference [16]
Sensitivity/(%)	98.75	91.25
Specificity/(%)	98.50	86.50
Accuracy/(%)	98.57	87.86

下载: 导出CSV

参考文献(16)

[1]	邓志锋, 闵卫东, 邹松. 一种基于CNN和人体椭圆轮廓运动特征的摔倒检测方法[J]. 图学学报, 2018, 39(6): 1042-1047. https://www.cnki.com.cn/Article/CJFDTOTAL-GCTX201806005.htm DENG Zhifeng, MIN Weidong, ZOU Song. A fall detection method based on CNN and motion features of human elliptical contour[J]. Journal of Graphics, 2018, 39(6): 1042-1047. https://www.cnki.com.cn/Article/CJFDTOTAL-GCTX201806005.htm
[2]	Giovanna S, Ivanoe D, Giuseppe D P. A supervised approach to automatically extract a set of rules to support fall detection in an mHealth system[J]. Applied Soft Computing, 2015, 34: 205-216. doi: 10.1016/j.asoc.2015.04.060
[3]	胡双杰, 秦建邦, 郭薇. 基于特征自动提取的跌倒检测算法[J]. 传感技术学报, 2018, 31(12): 1842-1847. https://www.cnki.com.cn/Article/CJFDTOTAL-CGJS201812011.htm HU Shuangjie, QIN Jianbang, GUO Wei. A fall detection algorithm with automatic feature extraction[J]. Chinese Journal of Sensors and Actuators, 2018, 31(12): 1842-1847. https://www.cnki.com.cn/Article/CJFDTOTAL-CGJS201812011.htm
[4]	郑毅, 李凤, 张丽, 等. 基于长短时记忆网络的人体姿态检测方法[J]. 计算机应用, 2018, 38(6): 1568-1574. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJY201806008.htm ZHENG Yi, LI Feng, ZHANG Li, et al. Human posture detection method based on long short term memory network[J]. Journal of Computer Applications, 2018, 38(6): 1568-1574. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJY201806008.htm
[5]	Giuffrida D, Benetti G, Martini D D, et al. Fall detection with supervised machine learning using wearable sensor[C]//Proceedings of 17th International Conference on Industrial Informatics (INDIN) of IEEE, 2019: 253-259.
[6]	Kumar V S, Acharya K G, Sandeep B, et al. Wearable sensor-based human fall detection wireless system[J]. Wireless Communication Networks and Internet of Things, 2019, 493: 217-234 doi: 10.1007/978-981-10-8663-2_23
[7]	Mehmood A, Nadeem A, Ashraf M, et al. A novel fall detection algorithm for elderly using Shimmer wearable sensors[J]. Health and Technology, 2019, 9(4): 631-646. doi: 10.1007/s12553-019-00298-4
[8]	Kerdjidj O, Ramzan N, Ghanem K, et al. Fall detection and human activity classification using wearable sensors and compressed sensing[J]. Journal of Ambient Intelligence and Humanized Computing, 2020, 11(1): 349-361. doi: 10.1007/s12652-019-01214-4
[9]	MIN W D, CUI H, RAO H, et al. Detection of human falls on furniture using scene analysis based on deep learning and activity characteristics[J]. IEEE Access, 2018, 6(99): 9324-9335. http://www.onacademic.com/detail/journal_1000040195826110_71bc.html
[10]	KONG Y Q, HUANG J H, HUANG S S, et al. Learning spatiotemporal representations for human fall detection in surveillance video[J]. Journal of Visual Communication and Image Representation, 2019, 59: 215-230. doi: 10.1016/j.jvcir.2019.01.024
[11]	QIU Z, LIANG X Q, CHEN Q Q, et al. Old man fall detection based on surveillance video object tracking[C]//Proceedings of 10th International Symposium on Parallel Architectures, Algorithms and Programming, 2019: 159-167.
[12]	FAN K B, WANG P, ZHUANG S. Human fall detection using slow feature analysis[J]. Multimedia Tools and Applications, 2019, 78: 9101-9128. doi: 10.1007/s11042-018-5638-9
[13]	Shota M, Jihoon H and Tomoaki O. A fall detection system using low resolution infrared array sensor[C]//Proceedings of 25th International Symposium on Personal, Indoor and Mobile Radio Communications of IEEE, 2014: 2109-2113.
[14]	杨任兵, 程文播, 钱庆, 等. 红外图像中基于多特征提取的跌倒检测算法研究[J]. 红外技术, 2017, 39(12): 1131-1138. http://hwjs.nvir.cn/article/id/hwjs201712011 YANG Renbing, CHENG Wenbo, QIAN Qing, et al. Research on fall detection algorithm based on multi-feature extraction in infrared image[J]. Infrared Technology, 2017, 39(12): 1131-1138. http://hwjs.nvir.cn/article/id/hwjs201712011
[15]	王召军, 许志猛, 陈良琴. 基于红外阵列传感器的人体行为识别系统研究[J]. 红外技术, 2020, 42(3): 231-237. http://hwjs.nvir.cn/article/id/hwjs202003005 WANG Zhaojun, XU Zhimeng, CHEN Liangqin. Human behavior recognition system based on infrared array sensors[J]. Infrared Technology, 2020, 42(3): 231-237. http://hwjs.nvir.cn/article/id/hwjs202003005
[16]	LIANG Q S, YU L, ZHAI X P, et al. Activity recognition based on thermopile imaging array sensor[C]//Proceedings of International Conference on Electro/Information Technology (EIT) of IEEE, 2018: 770-773.