基于YCbCr与角点检测的火灾识别算法研究

Research on Fire Identification Algorithm Based on YCbCr and Corner Detection

目前火灾识别技术已成为火灾预警与有效控制的重要手段之一,但传统的火灾识别方法存在易受噪声干扰及受数据规模限制的问题。本文提出了一种基于YCbCr与角点检测的火灾识别算法。通过YCbCr提取疑似火焰区域,实现疑似火焰区域内的颜色、面积变化率、尖角个数和圆形度的计算与分析。在尖角个数特征提取方法基础上,提出了一种基于角点检测的方法,使用近似值代替精确值的算法得到尖角个数。利用提出的基于交叉验证的网格搜索方法获得最优参数对,以实现支持向量机(SVM)对数据逐帧地检测。实验结果表明,通过本文提出的方法能够获得噪点更少、更清晰、识别度更高的火焰图像,预测正确率达到95%以上。因此,该方法可用于常见火灾检测。

At present,fire recognition technology has become one of the most important tools for early warning and effective control of fires.However,traditional methods for fire recognition are vulnerable to noise disturbance and limited by data volume.A fire recognition algorithm is proposed herein based on YCbCr and corner detection.Through YCbCr extraction of suspected flame area,the color,area change rate,number of sharp corners and roundness in the suspected flame area are calculated and analyzed.On top of the feature extraction method of the number of sharp corners,a corner detection-based method is proposed that obtains the number of sharp corners by using approximation instead of the exact value.The optimal parameter pairs are obtained using the proposed cross-validation-based grid search,so as to enable frame-by-frame detection of data by support vector machine(SVM).According to the experimental results,the proposed method can obtain clearer flame images of less noise and higher degree of recognition,with a prediction accuracy of above 95%.Therefore,the method can serve as a common tool for fire detection.