【作者】 张虎；

【导师】 谭克俊；

【作者基本信息】 大连海事大学 ， 电子科学与技术， 2013， 硕士

【摘要】 火灾是世界各国人民所面临的一个共同的灾难性问题,火焰的实时检测与准确识别,对早期火灾的发现从而降低火灾所造成的损失有很大帮助。基于视频图像的火焰检测方法融合了图像处理、计算机视觉、模式识别和数字信号处理等方面的技术,具有检测准确率高、抗干扰能力强、通用性好等优点,近年来逐步取代传统的感烟、感温、感光、以及红外等火灾探测方法,成为消防行业的热点研究领域。本文在认真研究国内外视频图像火焰检测技术的设计思想和实现方法的基础上,提出了一种基于FPGA的视频火焰检测系统实现方案。该方案充分利用了FPGA并行处理能力强,内部资源丰富,逻辑可重新配置的优点,全部设计均采用Verilog HDL编程实现,与DSP、ARM等嵌入式处理器相比,具有运算速度快,实时性好,系统升级方便灵活等特点。本系统以Altera公司的Cyclone Ⅱ系列EP2C35为核心,主要由D5M数字摄像头初始化模块、图像采集模块、图像转换模块、多端口SDRAM控制器模块、VGA接口控制模块、火焰检测及定位等六个模块组成。其中火焰检测及定位模块又包含色彩空间转换、火焰识别模块和火焰定位模块。在火焰检测算法上,选择颜色特征判别和火焰目标运动特征作为火焰区域识别的判断条件。最终将火焰识别及检测结果在VGA显示器上实时显示观察。本系统的开发是基于Altera公司DE2硬件开发平台以及QuartusⅡ软件开发环境。无论是视频图像的采集与控制,还是火焰检测算法的实现均使用纯硬件的方法实现,没有嵌入NiosⅡ软核,从而使系统具备了很高的实时性和可靠性。论文最后给出了火焰视频图像检测的实验结果,并对实验结果及系统硬件资源消耗情况进行了分析。更多还原

【Abstract】 There is a common catastrophic problem for the people all over the world that fire can damage people’s lives, so it is some great help if there is an early discovery via an accurate identification. The video image fusion based flame detection method merges image processing, computer vision, pattern recognition and digital signal processing technology. This kind of detection has a good advantage of high accuracy detection, anti-interference ability, versatility, etc. It has taken place of the traditional sense smoke, temperature, light-sensitive and infrared fire detection gradually and becomes hot in research field of fire industry in recent year.This paper proposes a new system implementation method of video flame detection based on FPGA on a condition of finishing some researches. These researches include the creative idea and implementation methods in this industry at home and abroad. This new method make full use of excellent parallel processing ability, rich internal resources and reconfigurable logic by Verilog HDL programming only. Compared with the DSP and ARM embedded processor, it has a good advantage of computing speed, real-time plan and flexibility in upgrade.This system takes EP2C35, Altera’s Cyclone Ⅱ series as the core. It is made up of six modules. They are digital camera initialization module D5M, image acquisition module, image conversion module, multi-port controller module SDRAM, interface control module VGA, flame detection and localization. In addition, flame detection and localization contains a color space conversion, flame recognition module and flame localization module. In flame detection algorithm, the judgment conditions of flame area identification are the color characteristics of discrimination and the flame target motion characteristics. The purpose of this system is to realize flame identification and detection result shown on a VGA monitor instantly.The development of this system is based on Altera hardware DE2platform and QuartusII software development environment. It has a great real-time ability and reliability, because there is no Nios II soft-core in video image acquisition and control, of the implementation of the lame detection algorithm. Finally, it is shown the performance of he flame video image detection and analysis of the experimental results and system hardware resource consumption.更多还原