针对现有的真火训练评估系统难以直接准确评价出实际灭火效果，以及部分真火训练环境比较空旷的特点，提出了一种基于视频图像处理技术识别火焰的算法，该算法能够直接检测出火焰大小，从而实现对训练效果更加直观准确的评估。通过对比分析的方法，确定采用YCbCr彩色模型算法，该算法拥有其自身独特的优点，即减轻改变照明度造成的负面影响并且提高火焰检测的准确率。为提高评估的准确性，针对火灾检测技术的难点和关键点提出了基于视频图像的解决办法，即先对火焰的圆形度、火焰边界的粗糙度和尖角个数分别进行特征提取，再进行动态特征提取，完成数据归一化处理。采取训练集对分类器展开训练，再根据训练集得到的模型对测试集进行预测，最终获取分类识别率。以某实际真火训练的1000张图片为样本，对模型中的火焰图像处理效果进行验证。实验结果表明：基于饱和度的规则和颜色模型能够更准确地体现火焰行为，在不利的照明背景条件下其鲁棒性更强。对比激光束检测、超声波检测和水量检测等方法，基于图像的火焰识别评估方式具有不易变形、定量直观、准确性高及操作方便等优点。

Since it is difficult to evaluate the extinguishing effect of existing real fire training system,this article introduces a flame recognition algorithm based on image processing technology.It can detect the fire size directly and evaluate the extinguishing effect more visually and exactly.Through comparison and analysis,the YCbCr color model algorithm is adopted.This algorithm has its own unique advantages in reducing the negative impact caused by changing of illumination and improving the accuracy of flame detection.To improve the evaluation accuracy,aiming at the difficulties and key points of fire detection technology,a new method is proposed to solve the image processing.It extracts the feature of the roundness of the flame,the roughness of the flame boundary and the number of sharp corners respectively,and then extracts the dynamic feature to complete the data normalization.It adopts the training set to train the classifier and predicts the test set by modeling obtained by training set.Finally it obtains the classification recognition rate.To verify the flame image processing effect of the model,1000 pictures from a real fire training are adopted as samples.The experimental result shows that the rules and color models based on saturation can reflect the flame behavior more accurately,and it is more robust under unfavorable lighting background conditions.Compared with methods such as laser beam detection,ultrasonic detection and water flow detection,imagebased flame recognition and evaluation method has more advantages such as high resistance to deforming,being quantitative and intuitive,higher accuracy and convenience of operation.

TP391.41