摘要
将CMYK彩色模型下的火焰颜色特征与粒子群优化(Particle swarm optimization,PSO)算法求最优适应度函数相结合,通过合理配置检测窗口邻域中颜色的M,Y各低阶矩的特征在欧氏距离分类适应度函数中的因子及PSO中各参数,使森林监测图像的可疑火灾着火点的初定位控制在0.05 s以内,最后通过对火灾样本关键图与初定位可疑着火点的颜色互信息的计算而得到进一步的校验和确定。实验表明改进的PSO算法具有良好的火焰区域定位效果及运算效率,易于移植到前端探测的嵌入式火灾探测系统。
A combination algorithm of color property is proposed based on CMYK color space and particle swarm optimization(PSO). The parameters of PSO and M, Y elements are imple- mented and configured as the adaptation function factors in Euclidean distance classification. Neighborhood regionrs color lower order moments are used as the features. Hence the time needed for primary orientation of suspected forest flame in forest monitoring image can be con- trolled in 0.05 s. The primary orientation of forest flame is verifcated by color mutual informa- tion calculation between key template image and searched window region image. Experiments show that the algorithm with better orientation result and higher calculation efficiency can be easily transfered into the embedded fire detecting system.
出处
《数据采集与处理》
CSCD
北大核心
2013年第3期324-329,共6页
Journal of Data Acquisition and Processing
基金
国家自然科学基金(61262031)资助项目
江西省研究生创新专项资金(YC2012-S081)资助项目
关键词
图像匹配
森林火灾
颜色空间
粒子群优化
互信息
image matching
forest flame
color space
particle swarm optimization
mutual infnrmation