建筑物高空抛物视频监控技术研究

论文总字数：25973字

摘 要

建筑物高空抛物时有发生，危害极大。本文研究利用运动目标视频检测技术来实时监测这个过程。

运动目标检测是计算机视觉技术的一个分支，就是对所录视频内的运动物体，如人和交通工具或者坠物等，进行实时与动态的观测，并将其分类，然后分析与总结运动特征。

本次研究的主要技术路线是视频获取，视频处理，运动检测与危险判断和视频的后处理工作。在此次研究中用到的主要算法是基于帧间差分法的运动检测。具体的实现方式是采集模拟高空抛物视频，使用视频编辑软件对原始视频进行预处理，使用Open CV软件对采集的视频进行图像的处理，主要包括图像去噪与图像二值化，图像去噪主要使用的算法是中值滤波，接着对视频序列通过帧间差分法来进行运动检测，最后通过一定的物理知识对高空抛物进行危险判断与报警。

通过本次研究结果分析，本文所研究的成果能满足实验开始之前所预设的目标：基本能完成建筑物高空抛物检测的目标，对大部分坠物有明显的判断，同时计算出的关于速度、楼层等的问题与实际情况几乎吻合，在危险判定正确时可以触发报警系统进行报警。

本次实验通过对不同算法进行比较，得出结论：针对建筑物高空抛物视频监控，基于帧间差分的方法明显优于基于背景减法的运动检测；此算法在处理颜色与背景颜色接近的物体是会出现明显错误；此算法对危害程度的判断不够严谨。此方法适合单一运动目标检测。

关键词:高空抛物 运动目标检测 帧差法 中值滤波 危险判断

RESEARCH ON HIGH ALTITUDE PARABOLIC VIDEO MONITORING

TECHNOLOGY FOR BUILDINGS

Abstract

Buildings at high altitude parabolic has occurred, great harm. In this paper, we use moving object video detection technology to monitor the process in real time.

Moving object detection is a branch of computer visual technology is on the recorded video moving objects, such as people and vehicles or falling objects, such as, real-time and dynamic observation and the classification, then analyzed and summarized the motion features.

The main technical route of this research is video capture, video processing, motion detection and the post processing of video. The main algorithms used in this study are motion detection based on inter frame difference method. The specific implementation is acquisition simulation parabolic aerial video, the use of video editing software to preprocess the original video, using open CV software to capture the video image processing, mainly including image de noise and image binarization, image denoising using the algorithm is median filter, then on the video sequence of frame difference method to detect motion. Finally through a certain physical knowledge of the height of parabolic risk judgment and alarm.

According to the study results, the results can meet the experiment prior to the start of the preset goal: can basically complete the goal of throwing objects from high buildings detection. For most sinkers have clear judgment, at the same time, we calculate the speed floors on the problems and the actual situation almost coincide with the and in dangerous judge correctly can trigger the alarm system alarm.

Different algorithms are compared through the experiment, draws the conclusion: for throwing objects from high buildings in video surveillance, based on inter frame difference method is obviously better than that of motion detection based on background subtraction; this algorithm in processing color and background color close to the object will appear obvious errors; this algorithm judgment on the degree of harm not strict enough. This method is suitable for single moving target detection.

Key words：Altitude parabolic Moving object detection frame difference method,

median filtering ,hazard

目 录

摘要 I

Abstract II

第一章 绪论 1

1.1课题背景与意义 1

1.2研究现状 1

1.3技术平台介绍 2

1.3.1 Microsoft Visual Studio 2010 2

1.3.2 Open CV2.4.9 2

1.4 本文的研究内容 2

1.5 本研究的技术路线及技术路线图 3

1.6本章小结 3

第二章 视频获取与处理 4

2.1视频的获取 4

2.2视频的预处理 4

2.3视频处理 4

2.3.1图像增强 5

2.3.2图像滤波 5

2.3.3图像去噪小结 7

2.3.4图像二值化 9

2.3.5图像分割 10

2.4本章小结 10

第三章 运动物体检测与报警 12

3.1运动物体检测方法 12

3.2检测结果保存 14

3.3危险程度判断 15

3.3.1质心的获取 15

3.3.2危险判断 16

3.3.3报警 17

3.4本章小结 17

第四章 Open CV编程实现与结果分析 18

4.1代码实现 18

4.2结果分析 22

4.3本章小结 22

第五章 总结 23

5.1实验总结 23

5.2存在的问题与解决方法 23

5.2.1体积问题 23

5.2.2颜色问题 25

5.2.3视角问题 25

5.2.4视频保存 25

5.3展望 26

致谢 27

参考文献 28

第一章 绪 论

1.1课题背景与意义

剩余内容已隐藏，请支付后下载全文，论文总字数：25973字