基于Multilinear PCA的烟雾图像识别算法  

Smoke Image Detection Algorithm Based on Multi-linear Principal Component Analysis

作　　者：戚永刚[1] 胡伟[2] 

机构地区：[1]德州学院外语系,山东德州253023 [2]湖南第一师范学院科研处,湖南长沙410002

出　　处：《电视技术》2014年第11期193-197,共5页Video Engineering

基　　金：国家自然科学基金项目(F020704);湖南省教育厅科学研究项目(10C0526;11C0280)

摘　　要：为了提高火灾烟雾图像识别正确率,更好地预防火灾,提出一种基于multilinear PCA的火灾烟雾图像识别算法。在提取烟雾疑似区域的基础上,首先分别提取烟雾的静态和动态特征,然后采用multilinear PCA对特征进行融合,消除其中冗余特征,并根据选择特征进行训练样本构造,最后将训练样本输入支持向量机建立烟雾图像分类器,并对测试样本进行识别。仿真结果表明,相对于其他烟雾图像识别算法,提出的算法不仅提高了烟雾图像识别正确率和识别效率,而且具有良好的抗干扰能力,可以满足不同环境下烟雾图像识别要求。In order to improve the fire smoke image recognition accuracy and prevent fire, a fire smoke image recognition algorithm based on multiple linear principal component analysis is proposed. Firstly, the static and dynamic features of the smoke suspected area is extracted respectively, and then multiple linear principal component analysis is used to eliminate the redundant features, and training sample is built according to the features, and finally the training samples are input support vector machine （SVM） to build the smoke image classifier, and the test samples are used to identified the perform- ance. Compared to other smoke image recognition algorithm, the simulation results show that this algorithm can improve the recognition efficiency and ac- curacy recognition of the smoke image, and has good anti-interference ability, and it can meet the requirements of the smoke image recognition under different environment.

关 键 词：烟雾图像 多线性主成分分析 支持向量机 特征提取 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]