没有合适的资源?快使用搜索试试~ 我知道了~
云计算-面向复杂公共区域的群体聚集性计算方法研究.pdf
1.该资源内容由用户上传,如若侵权请联系客服进行举报
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
版权申诉
0 下载量 171 浏览量
2022-07-03
21:21:00
上传
评论
收藏 6.63MB PDF 举报
温馨提示
试读
68页
云计算-面向复杂公共区域的群体聚集性计算方法研究.pdf
资源推荐
资源详情
资源评论
摘要
I
摘要
随着城镇人口的增多和城市规模的不断扩大,发生在公共区域的群体聚集
情形越来越常见,伴随而来发生在人群中的恐怖暴力和踩踏等群体性事件严重
影响了社会的安全与稳定,针对此类场景中的群体聚集行为分析逐渐成为智能
视频监控领域新的研究热点。
由于公共区域内的目标对象类型较多、群体运动模式复杂多变,使得针对
这些区域内的群体聚集状态分析变得非常困难。针对该问题,本文提出了一种
基于运动目标轨迹信息的群体聚集状态计算方法。首先,采用了一种新的全局
特征来表示公共区域内的群体运动,该特征可以充分描述目标兴趣点的时空和
运动信息;其次,提出了一种对目标兴趣点先聚类分组再计算聚集性的策略,
使得单个群体的聚集性计算更加一致有效;最后,提出了一种更为全面的群体
聚集性描述子,用于对人群聚集状态进行详细描述,并以该描述子为基础,实
现了群体运动状态演化分析与异常行为检测。
在本文提出的群体聚集状态计算方法中,所采用的目标运动信息是利用 KLT
算法提取特征点的方法得到的,并基于这些特征点对人群密度做近似估计。这
种方法能够处理摄像头距行人较远的场景,但当摄像头距行人较近时,这种方
法就会存在较大误差。为对后续人群密度以及人群状态分析提供准确的数据,
本文采用对行人进行跟踪的方法来提取这些信息。然而,由于监控视频中的目
标对象姿态多变、稠密场景下个体间遮挡严重,传统的检测跟踪方法并不能很
好的解决这些问题。因此,针对以上问题,本文提供了一种基于深度学习和卡
尔曼滤波的行人目标跟踪算法。首先使用卷积神经网络对行人进行检测;然后
针对检测过程中出现的漏检现象,采用卡尔曼滤波算法来预测行人当前的位置,
最后使用匈牙利分配算法关联视频相邻帧中的行人,从而实时的抽取行人运动
信息。
实验结果表明,本文所提出的群体运动特征提取方法能够很好地应用于中
高密度场景下的人群运动特征计算;结合这些精确的人群运动特征,本文提出
的群体聚集性描述子能够应用于各种复杂公共区域内的群体行为分析,可为公
共安全管理提供真实可靠的参考。
关键词:群体运动 全局特征提取 聚集状态计算 行人跟踪
万方数据
Abstract
II
Abstract
With the explosive growth of the population and the scale of cities, the crowd
stampede and terrorist attacks in public areas have become more dangerous and
seriously affect the safety and stability of the society. Therefore, the analysis of crowd
aggregation behavior has been a new research focus in the field of intelligent video
surveillance.
However, such public area scenes not only contain moving crowd but also
contain other types of objects. The sizes of these objects are usually small, which
make their appearances quite similar. Moreover, the individuals in a crowd move
randomly and often occlude each other. All the above factors make the analysis of
crowd aggregation very difficult. In this thesis, we propose a novel crowd aggregation
algorithm based on the moving object trajectory to tackle this problem in three
aspects. Firstly, a new global feature is used to represent the moving crowd. This
feature can well describe the spatial and the temporal motion information of
points-of-interest. Then, a strategy is adopted to cluster the interesting points firstly
and then calculate the collectiveness. This makes the collectiveness computation of
individual groups more consistent and effective. Finally, a more comprehensive
collective descriptor is proposed to provide a detailed description of the crowd status.
Based on the proposed descriptor, the authors realize the evolution analysis of the
group movement and the crowd abnormal detection.
The proposed crowd aggregation status computing method, in which the object
motion can be obtained via KLT to extract the feature points estimating the
population density, can tackle the situation that the camera is far away from the
pedestrians. However, this method may lead to big error when the camera is close to
the pedestrians. In order to provide accurate data for the analysis of the crowd status
and the population density, we track the pedestrians to extract the motion information.
In the video surveillance system, the traditional detecting and tracking methods
cannot deal with these problems effectively because of the varied posture and the
serious occlusion of the individual in the dense scene. In this thesis, we present a
万方数据
Abstract
III
novel pedestrian tracking algorithm which detects the pedestrian by means of the
convolution neural network, predicts the current positions of pedestrians via the
Kalman filter, and associates the pedestrians in the adjacent frames of the video to
extract the motion information in real time.
The experiment results show that the proposed methods are able to deal with the
feature extraction of crowd movement in the dense scene. The proposed algorithms
based on the extracted features can analyze the crowd aggregation status in various
public areas, which provides a reliable reference for the public safety management.
Key words: Crowd movement, Global feature extraction, Aggregation computation,
Pedestrian tracking
万方数据
目录
IV
目录
摘要 ..................................................... I
Abstract ................................................ II
目录 .................................................... IV
图目录 .................................................. VI
1 绪论 ................................................... 1
1.1 研究背景及意义 ............................................. 1
1.2 国内外研究现状 ............................................. 3
1.3 本文主要研究内容 ........................................... 6
1.4 本文的组织结构 ............................................. 8
2 相关技术 .............................................. 10
2.1 KLT 跟踪算法 ............................................... 10
2.2 卷积神经网络 .............................................. 13
2.3 本章小结 .................................................. 20
3 基于运动目标轨迹信息的群体聚集状态计算 ................ 21
3.1 公共场景特征点提取 ........................................ 21
3.2 基于图方法的场景特征点聚类 ................................ 23
3.3 群体聚集性描述子计算 ...................................... 24
3.4 实验与结果分析 ............................................ 26
3.4.1 群体聚类结果 ................................................ 27
3.4.2 聚类参数分析 ................................................ 27
3.4.3 群体描述子分析 .............................................. 30
3.5 群体聚集性描述子在不同场景下的应用 ........................ 32
3.5.1 人群运动演化分析 ............................................ 32
3.5.2 人群异常状态检测 ............................................ 34
3.6 本章小结 .................................................. 36
万方数据
目录
V
4 面向群体聚集状态计算的行人目标跟踪 .................... 38
4.1 基于卷积神经网络的行人目标检测 ............................ 38
4.2 基于卡尔曼滤波的遮挡行人预测 .............................. 43
4.3 基于相似度计算的行人目标关联 .............................. 47
4.4 实验与结果分析 ............................................ 48
4.4.1 实验数据集 .................................................. 48
4.4.2 实验设置 .................................................... 49
4.4.3 结果分析 .................................................... 50
4.5 本章小结 .................................................. 55
5 总结与展望 ............................................ 56
5.1 论文总结 .................................................. 56
5.2 未来展望 .................................................. 56
参考文献 ................................................ 58
个人简历、在学期间发表的学术论文与研究成果 .............. 63
致谢 .................................................... 64
万方数据
剩余67页未读,继续阅读
资源评论
programxh
- 粉丝: 17
- 资源: 1万+
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功