火焰及干扰图像形态特征分析研究

Research on shape feature of flame image and noise image

对不同材质的火焰与干扰图像的形态特征进行分析对比,选择面积、周长、圆形度、边界粗糙度、矩形度、尖角数、相似度、质心位移对火灾探测作为候选特征,经过分析排除面积、周长作为候选特征,建立其他6个形态特征的弱分类器,运用AdaBoost算法计算出各个弱分类器的权值,根据弱分类器的权值进行排序,提出形态特征选择的优先性,排列顺序为相似度、圆形度、质心位移、尖角数、边界粗糙度、矩形度。

Compared the shape feature of difference flame image and difference noise image. Area, perimeter, rotundity, solidi- ty, extent, number of sharp corners, similarity and barycenter displacement were chosen as shape feature for video fire detec- tion. Area and perimeter were removed by analyzing. Weaker classify based on difference shape feature were built. Weight Of weaker classify was calculated by AdaBoost. Rank of weight is built to propose the priority of difference shape feature, namely similarity, rotundity, baryeenter displacement, number of sharp corners, solidity, extent.