基于机器视觉的早期火焰探测算法研究

A study on the algorithm to detect fires in early stage based on MV

目的针对基于机器视觉(MVFDS)的火焰探测系统存在着虚警误报的问题提出了一种增强型算法,用以改善系统性能。方法分析了新算法的5个附加功能,即用于减少虚警误报的3种功能:突变探测功能、前景平均色素计算功能、前景色素比计算功能,以及用于改善系统功效和灵敏度的两个功能:设置前景边界框功能和三级触发报警功能。结果给出了基于新算法的决策过程和实验效果。结论文中采用的参数计算和决策方法一方面减少了虚警率,另一方面利用前景框提高了处理速度,同时三级触发报警使用户可以灵活地采用最佳方案应对各种火焰情况,使系统的探测性能得到提高。

Aim To propose an enhanced algorithm to eliminate the possibilities of known false alarms to improve the features of MVFDS and analyze the new algorithm＇s five additional features added to the previous system theodicy. Methods The additional features include sudden change detector, foreground mean color element computation and foreground color element ratio computation in order to eliminate the false alarms in several cases and foreground bounding box and three-level alarm trigger aim to improve the efficiency and sensitivity of the system. Results The parameters calculation and decision-making methods on the one hand reduce the false alarm rate, on the other hand used the prospect frame raise the processing speed, while the three triggered alarm allows users to use the best way to dealing with all kinds of flame flexibility, and improved the system Detection performance. Conclusion The effect of simulated experiment was briefly introduced and the process of decision was made based on the new algorithm.