一种基于关键点的红外图像人体摔倒检测方法

徐世文; 王姮; 张华; 庞杰

一种基于关键点的红外图像人体摔倒检测方法

西南科技大学信息工程学院，四川绵阳 621000

详细信息

作者简介:
徐世文（1994-），男，四川省成都市人，硕士研究生，主要研究方向为计算机视觉与图像处理、深度学习。E-mail：1411761943@qq.com

通讯作者:
王姮（1971-），女，硕士，教授，主要研究方向为机器人技术及应用、自动化技术研究。E-mail：wh839@qq.com

中图分类号: TP391.4
计量
- 文章访问数: 375
- HTML全文浏览量: 130
- PDF下载量: 89
- 被引次数: 0
出版历程
- 收稿日期: 2020-02-18
- 修回日期: 2020-02-21
- 刊出日期: 2021-10-20

Human Fall Detection Method Based on Key Points in Infrared Images

School of Information Engineering, Southwest University of Science and Technology, Mianyang 621000, China

摘要

摘要: 针对已有人体摔倒检测方法在复杂环境场景下易受光照影响、适应性差、误检率高等问题，提出了一种基于关键点估计的红外图像人体摔倒检测方法。该方法采用红外图像，有效避免了光照等因素的影响，经过神经网络找到人体目标中心点，然后回归人体目标属性，如目标尺寸、标签等，从而得到检测结果。使用红外相机采集不同情况下的人体摔倒图像，建立红外图像人体摔倒数据集并使用提出的方法进行检测，识别率达到97%以上。实验结果表明提出的方法在红外图像人体摔倒检测中具有较高的精度与速度。
- 红外图像 /
- 关键点估计 /
- 摔倒检测 /
- 神经网络
Abstract: To address the problems with existing human fall detection methods for complex environments, which are susceptible to light, poor adaptability, and high false detection rates, an infrared image human fall detection method based on key point estimation is proposed. This method uses infrared images, which effectively eliminates the influence of factors such as lighting; first, the center point of the human target is found through a neural network, and second, the human target attributes, such as the target size and label, are regressed to obtain detection results. An infrared camera was used to collect human body fall images in different situations and establish datasets containing infrared images of human falls. The proposed method was used for experiments; the recognition rate exceeded 97%. The experimental results show that the proposed method has a higher accuracy and speed than other two methods in infrared image human fall detection.
- infrared image /
- key point estimation /
- fall detection /
- neural network

HTML全文

图 1 Centernet网络整体架构

Figure 1. Overall network architecture of Centernet

下载: 全尺寸图片幻灯片

图 2 红外相机以及电源

Figure 2. Infrared camera and power supply

下载: 全尺寸图片幻灯片

图 3 红外数据录制场景

Figure 3. Infrared data recording scene

下载: 全尺寸图片幻灯片

图 4 部分人体摔倒数据集

Figure 4. Partial human fall dataset

下载: 全尺寸图片幻灯片

图 5 人体摔倒检测效果

Figure 5. Human fall detection effect

下载: 全尺寸图片幻灯片

图 6 本文算法的P-R曲线

Figure 6. The P-R curve of the algorithm in this paper

下载: 全尺寸图片幻灯片

表 1 对比实验结果

Table 1. Comparison of experimental results

Algorithm Accuracy/(%) Time/(ms/frame)

Yolo v3 96.9 0.0421

Faster RCNN 95.7 0.441

Ours 98.4 0.0462

下载: 导出CSV

参考文献(16)

[1]	Santos G, Endo P, Monteiro K, et al. Accelerometer-based human fall detection using convolutional neural networks[J]. Sensors, 2019, 19(7): 1644. doi: 10.3390/s19071644
[2]	Gia T N, Sarker V K, Tcarenko I, et al. Energy efficient wearable sensor node for IoT-based fall detection systems[J]. Microprocessors and Microsystems, 2018, 56: 34-46. doi: 10.1016/j.micpro.2017.10.014
[3]	Nadee C, Chamnongthai K. Multi sensor system for automatic fall detection[C]//2015 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA) of IEEE, 2015: DOI: 10.1109/APSIPA.2015.7415408.
[4]	Tzeng H W, CHEN M Y, CHE J Y. Design of fall detection system with floor pressure and infrared im age[C]//2010 International Conference on System Science and Engineering of IEEE, 2010: 131-135.
[5]	Kerdegari H, Samsudin K, Rahman Ramli A, et al. Development of wearable human fall detection system using multilayer perceptron neural network[J]. International Journal of Computational Intelligence Systems, 2013, 6(1): 127-136. doi: 10.1080/18756891.2013.761769
[6]	LIU Chengyin, JIANG Zhaoshuo, SU Xiangxiang, et al. Detection of human fall using floor vibration and multi-features semi-supervised SVM[J]. Sensors, 2019, 19(17): 3720(doi: 10.3390/s19173720).
[7]	Mazurek P, Wagner J, Morawski R Z. Use of kinematic and mel- cepstrum-related features for fall detection based on data from infrared depth sensors[J]. Biomedical Signal Processing and Control, 2018, 40: 102-110. doi: 10.1016/j.bspc.2017.09.006
[8]	MIN W, CUI H, RAO H, et al. Detection of human falls on furniture using scene analysis based on deep learning and activity characteristics[J/OL]. IEEE Access, 2018, 6: 9324-9335.
[9]	FENG W, LIU R, ZHU M. Fall detection for elderly person care in a vision-based home surveillance environment using a monocular camera[J]. Signal, Image and Video Processing, 2014, 8(6): 1129-1138. doi: 10.1007/s11760-014-0645-4
[10]	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): 30-35. https://www.cnki.com.cn/Article/CJFDTOTAL-GCTX201806005.htm DENG Zhifeng, MIN Weidong, ZOU Song. A fall detection method based on CNN and human elliptical contour motion features[J]. Journal of Graphics, 2018, 39(6): 30-35. https://www.cnki.com.cn/Article/CJFDTOTAL-GCTX201806005.htm
[11]	卫少洁, 周永霞. 一种结合Alphapose和LSTM的人体摔倒检测模型[J]. 小型微型计算机系统, 2019, 40(9): 1886-1890. doi: 10.3969/j.issn.1000-1220.2019.09.014 WEI Shaojie, ZHOU Yongxia. A human fall detection model combining alphapose and LSTM[J]. Minicomputer System, 2019, 40(9): 1886-1890. doi: 10.3969/j.issn.1000-1220.2019.09.014
[12]	赵芹, 周涛, 舒勤. 飞机红外图像的目标识别及姿态判断[J]. 红外技术, 2007, 29(3): 167-169. doi: 10.3969/j.issn.1001-8891.2007.03.012 ZHAO Qin, ZHOU Tao, SHU Qin. Target recognition and attitude judgment of aircraft infrared image[J]. Infrared Technology, 2007, 29(3): 167-169. doi: 10.3969/j.issn.1001-8891.2007.03.012
[13]	ZHOU X, WANG D, Krhenbühl P. Objects as points [J/OL]. [2019-04-25]. arXiv: 1904.07850(https://arxiv.org/abs/1904.07850).
[14]	YU F, WANG D, Shelhamer E, et al. Deep layer aggregation[J/OL]. [2019-01-04]. arXiv: 1707.06484(https://arxiv.org/abs/1707.06484)
[15]	Law H, DENG J. CornerNet: detecting objects as paired keypoints[J]. International Journal of Computer Vision, 2020, 128(3): 642-656. doi: 10.1007/s11263-019-01204-1
[16]	LIN T Y, Goyal P, Girshick R, et al. Focal loss for dense object detection[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017: DOI: 10.1109/ICCV. 2017.324. http://fcv2011.ulsan.ac.kr/files/announcement/657/1708.02002.pdf