计算机科学与应用

Vol.3 No.5 (August 2013)

一种基于多种特征组合的烟雾检测方法
A Method of Smoke Detection Based on Various Features Combination

 

作者:

郑雪梅 , 杨 胜 :湖南大学信息科学与工程学院

 

关键词:

小波特征纹理特征BP神经网络分类器烟雾检测Wavelet Feature Texture Feature BP Neural Network Classifier Smoke Detection

 

摘要:

视频中类似烟雾的区域很大程度上增加了烟雾检测的误差,为了提高烟雾检测的准确性,利用BP神经网络分类器,一种基于小波特征、烟雾纹理特征以及Y分量均值特征相结合的烟雾检测方法被提出。首先对视频序列进行运动区域提取;然后对疑似区域提取小波特征、纹理特征以及Y分量均值特征,形成一种新的多特征组合向量;最后将特征向量输入到BP神经网络分类器进行检测。实验表明,通过这种组合特征的方法,检测结果更有效。

Smoke-like regions greatly increase smoke detection errors in the video. In order to improve the accuracy of smoke detection, a smoke detection method based on BP neural network is proposed, combining the wavelet feature, smoke texture feature and mean of Y component pixel value. Firstly, moving regions in the video sequences are ex- tracted; secondly, the wavelet feature and texture feature of suspected regions are extracted, then a new kind of multi-feature vector is formed. Finally, feature vector is input into the BP neural network classifier for smoke detection. The experiments show that smoke detection results are more effective by combing various features.

文章引用:

郑雪梅 , 杨 胜 (2013) 一种基于多种特征组合的烟雾检测方法。 计算机科学与应用, 3, 239-243. doi: 10.12677/CSA.2013.35041

 

参考文献

分享
Top