Abstract: In order to overcome the disadvantage that it needs long time for Support Vector Machine(SVM) to solve problems when the data quantity is large, this paper puts forward a video fire smoke recognition algorithms based on Least Squares Support Vector Machine(LS-SVM). Through a second segmentation of suspicious smoke areas, the color feature, correlation coefficient and area change rate are selected as the input feature vector. The dimension of input vector and the training time has been reduced. Experimental results show that the algorithm enhances the classification speed and identify accuracy.

Key words: Least Squares Support Vector Machine(LS-SVM), feature extraction, fire recognition, image-based fire smoke, smoke detection

摘要： 支持向量机在数据量较大时求解时间很长。针对该问题，提出一种基于最小二乘支持向量机的视频火灾烟雾识别算法。对烟雾的可疑区域进行二次分割，选取颜色特征、相关系数和面积变化率作为特征输入向量，由此降低输入向量维数，缩短训练时间。实验结果表明，该算法具有较快的分类速度和较高的识别准确率。

关键词: 最小二乘支持向量机, 特征提取, 火灾识别, 图像型火灾烟雾, 烟雾探测

CLC Number: 

• TP301.6