基于机器学习的跌倒检测算法

基于机器学习的跌倒检测算法(任务书,开题报告,论文13000字)
摘  要
随着我国医疗技术的不断提高，人口老龄化问题也越发明显。现今老年人意外跌倒已成为威胁到老年人身心健康的一个重大隐患，近年来跌倒检测识别也逐渐成为当下研究领域中的热门研究方向之一。
本文对人体跌倒检测识别的整个过程所用到的主要相关检测方法进行了分析和总结，并且参考了近年来国内外对跌倒检测算法方面的许多文献资料，提出了一种基于机器学习的在视频中对人体跌倒进行检测的算法。在本文中使用了一种基于卷积神经网络的开源深度学习框架——Darknet，通过Darknet来帮助实现对人体目标的识别和相关人体行为的检测。文中参考了传统的目标提取方法，先用背景差分法和相关形态学算法来进行目标骨架的提取，然后再采用基于人体比例的判断依据来初步对人体跌倒的情况进行判断，再使用以被检测对象躯体的运动趋势的检测依据来进行跌倒情况的精准判断。该跌倒检测算法效果已达到了预期研究目标。
关键词：机器学习；跌倒检测；目标检测；Darknet；YOLO

Abstract
With the continuous improvement of medical technology in China, the problem of population aging has become more apparent. The accidental fall of the elderly today has become a major hidden danger to the physical and mental health of the elderly. In recent years, the detection of fall detection has gradually become one of the hot research directions in the current research field.
This paper analyzes and summarizes the main related detection methods used in the whole process of human fall detection and identification, and refers to many literatures on fall detection algorithms in recent years, and proposes a machine learning based on video. An algorithm for detecting human fall. In this paper, a deep-depth deep learning framework based on convolutional neural network, Darknet, is used to help identify human targets and detect human behavior through Darknet. In this paper, the traditional target extraction method is referenced. The background difference method and the related morphological algorithm are used to extract the target skeleton, and then the judgment based on the proportion of the human body is used to judge the fall of the human body and then used to be detected. The detection of the movement trend of the subject's body is based on the accurate judgment of the fall situation. The effect of the fall detection algorithm has reached the expected research goal.
Key Words：Machine learning；Fall detection；Target Detection；Darknet；YOLO
 
目录
第1章 绪论    6
1.1选题背景与意义    6
1.2国内外研究状况    6
1.2.1国内研究状况    6
     1.2.2国外研究状况    7
1.3论文主要工作    7
第2章 基于神经网络的深度学习框架Darknet    8
第3章 实时目标检测    10
3.1 Yolo概述    10
3.2 Yolo核心思想    10
3.3 Yolo与OpenCV实现目标检测    14
第4章 跌倒检测算法在Darknet框架上的实现    16
4.1人体跌倒过程分析    16
4.2 跌倒检测算法剖析    17
4.3 跌倒检测算法的设计    17
4.4 Darknet框架搭建与配置    20
第5章 算法测试结果    22
5.1 实验结果展示    22
5.2 算法需要改进的地方    22
第6章 总结与展望    23
6.1 论文总结    23
6.2 跌倒检测发展方向展望    23
参考文献    24
致谢    25