摘要： 为实现电力机房重点区域指示灯安全事件的监控,提出一种音视频融合检测方法。采用监控区域彩色图像非线性变换和最大类间方差法自动阈值分割技术进行指示灯定位,将局部区域统计直方图作为视频特征向量。采用连续多帧梅尔频率倒谱系数建立监控区域的音频特征向量。利用主成分分析对连续多帧的音视频融合特征向量进行降维处理,并借助支持向量机对不同安全事件进行分类检测。实验结果表明,与单独采用音频或者视频进行安全事件检测的方法相比,该方法具有较高的检测率和较低的误检率。
关键词:
电力机房,
彩色图像非线性变换,
最大类间方差法,
梅尔频率倒谱系数,
音视频融合,
安全监控
Abstract: An audio-video fusion detection method is proposed in this paper for indicator light security incident surveillance in key areas of the electrical room.The optimal indicator range is determined by color image nonlinear transform for monitoring area and OTSU automatic threshold segmentation technique.The local gray scale histograms are taken as the video feature vector,and the Mel-frequency Cepstral Coefficien(MFCC) for multi frame is selected as the audio feature vector.The audio feature vector and the video feature vector are combined into a relatively large-scale vector.Principal Component Analysis (PCA) is employed for dimensionality reduction of the fused feature vector,and Support Vector Machine(SVM) is used to classify and detect the security incidents.Experimental result show that the proposed method obtains higher detection rate and lower false positive rate over the traditional single security incident detection(only audio-based or video-based) methods.
Key words:
electrical room,
color image nonlinear transformation,
OTSU,
Mel-frequency Cepstral Coefficient(MFCC),
audio-video fusion,
security monitoring
中图分类号: