基于人体检测的异常行为识别

摘要异常行为识别技术已经得到了广泛的应用，并且将在诸如智能监控、自动导航、生产自动化等领域有着更广阔的应用前景。本文在现有的的科研成果上，对运动目标识别、特征提取、建模与匹配识别等方面进行了研究，并对单一对象的运动视频进行了异常识别的编程实现。68472

异常行为识别基本可分为三个阶段：运动目标检测、特征提取和分类识别。本文提出了在第一阶段采用混合高斯模型进行背景减除的方式，分离前景并获得运动目标。在第二阶段提取关键帧的 Hu 不变矩的前三项作为特征。最終对所有样本数据建立隐马尔科夫模型，自定义一些模型类别为异常行为，用概率的方法识别出测试数据属于某一模型，进而判断是否是异常行为。

毕业论文关键词行为识别背景差分高斯混合模型特征提取隐马尔科夫模型

Title Abnormal behavior recognition based on human detection

Abstract

Abnormal behavior recognition technology has been widely used, and will have a broad application prospect in the fields such as intelligent monitoring, automatic navigation and production automation. In this paper, on the existing research results, I study at the moving target recognition, feature extraction, modeling and matching, etc. Then I program the abnormal behavior recognition of motion video with single object.

Abnormal behavior recognition can basicly be pided into three stages: moving object detection, feature extraction and classification recognition. I proposed to use the gaussian mixture model in the first stage for background deduction, then I and separate foreground and get the moving target. In the second stage, I extract the former three of the Hu invariant moment from key frames as characteristics. Eventually, I respectively model all sample data as hidden markov model and classify some modes as abnormal. I use probability method to identify which model the test data belongs to and then judge whether it is the abnormal behavior.

Keywords behavior recognition background subtraction Gaussian mixture model(GMM) feature extraction hidden makov model(HMM)

1 绪论 1

1.1 研究背景和意义 1

1.2 关键技术研究现状 · 3

1.3 本文主要研究内容 · 4

1.4 本文实验环境 · 5

1.5 本文内容安排 · 7

2 运动目标检测 8

2.1 引言 · 8

2.2 背景差分 · 基于人体检测的异常行为识别:http://www.youerw.com/jisuanji/lunwen_77006.html