【作者】 孙强；

【导师】 张静；

【作者基本信息】 哈尔滨理工大学 ， 控制工程， 2016， 硕士

【摘要】 选煤厂是火灾易发生场所,具有空间广阔、空气流动性强等特征,传统火灾探测器在这种工况环境下很难发挥作用,而且基于常规传感器的火险检测实时性差,对火灾初期的烟雾无法进行侦测。近些年,随着图形学科与数字视觉的不断发展,数字图形图像技术被越来越多的工程实践与软件开发采用,本文应用计算机图像处理方法对选煤厂进行火险检测研究,以解决空旷环境下的烟雾侦测问题。论文在分析火险像素特征基础上,以Intel公司推出的一款视觉库Open CV作为开发工具,进行带有角度检测的火险检测算法研究。具体研究内容如下:首先,论文简要介绍了课题研究的背景与相关技术的发展现状,通过传统检测手段与图像处理技术的对比分析,阐述了应用图像处理技术进行火险检测具有重大的现实意义。其次,论文对涉及的概念、理论知识及基本算法进行了阐述,对采用的图像处理工具Open CV及多线程技术进行了简单介绍,为论文研究提供了可靠的理论依据。再次,论文对火险前景区域的动态特性与像素特征进行分析;通过分析前景图像与背景图像,判断前景物体是否满足火焰的动态特性;将火焰的像素特征进行YCb Cr颜色空间转换,得到转换后火焰图像满足的条件,并指出火焰的三甬道渐近线具有递减规律;文中在Windows平台对几种发光物体进行检测,得出像素特征算法能够有效检测出火险的发生,但容易造成对其他物体的错误检测。最后,在像素特征算法的基础上,论文提出了带有角度测量的检测算法,并对经典的Harris检测算法做进一步优化,利用Harris强度值估算火焰尖角的角度,并通过查表法计算出具体角度值,得出满足条件的火焰尖角个数;文中综合火焰的尖角特性和像素特征,实现了带有角度测量的检测算法,进而判断火险是否发生,通过在实验平台上与像素特征算法对比得出,带有角度测量的火险检测算法能够有效提高自动报警的准确性,及时完成有效的火灾探测。更多还原

【Abstract】 Coal preparation plant is fire prone places, has broad space and air flow characteristics such as strong, traditional fire detectors in the working environment is difficult to play a role and conventional sensor real-time fire detection based on difference, on the initial fire smoke cannot detect. In recent years, with the continuous development of digital graphic discipline and visual,digital graphics technology is using more and more engineering practice and software development, the application of computer image processing method of coal preparation plant fire detection research, in order to solve the open environment of smoke detection problem. Based on the analysis of the feature of fire pixels based on, Intel launched a vision library opencv as a development tool,with angle detection of fire detection algorithm research. The specific contents are as follows:Firstly, this paper briefly introduces the research background and related technologies of the development status, through the comparison of traditional detection methods and image processing technology analysis, elaborates on the application of image processing technology for fire detection is of great practical significance.Secondly, the paper relates to the concept, theory and basic algorithm are described, on the image processing tools and multi thread technology are briefly introduced in this paper, provides a reliable theoretical basis for the research.Thirdly, analysis of the fire foreground region dynamic characteristics and pixel feature; through the analysis of the foreground image and background image, judge foreground objects is to satisfy the flame dynamic characteristics;flame pixel color space conversion, conversion after the flame image meet the conditions, and pointed out that the flame three corridor asymptote with law ofdiminishing; on the windows platform of several luminous objects were detected that pixel feature algorithm can effectively detect the fire occurrence, but easy to cause the error checking on other objects.Finally, in the pixel feature algorithm based on, this paper puts forward the detection algorithm with measurement of the angle, and the classical Harris detection algorithm do further optimization, the Harris strength value flame angle angle estimation, and through the look-up table method to calculate the specific angle value, that satisfy the conditions of flame angle number; the integrated flame angle characteristic and pixel features, the realization of the detection algorithm with measurement of the angle, and then judge whether there is a fire,through in experimental platform and feature pixels algorithm compared to that with angle measurement of the fire detection algorithm can effectively improve the accuracy of automatic alarm, timely completion of effective fire detection.更多还原