摘要
基于视频图像分析,提出了一种融合火焰识别与背景干扰滤除的算法。系统分析了火焰图像的视觉特征,重点利用其高亮特性,通过图像灰度化、最优阈值分割及二值化等预处理技术,实现了可疑区域的有效分割。针对复杂环境下的干扰问题,创新性地构建了GMD三特征融合识别模型:提取火焰的几何特征(Geometric)、形态特征(Morphological)和动态特征(Dynamic)作为判别依据。基于实际火情视频数据的实验验证表明,该方法能有效识别多场景下的火焰目标,特别是通过GMD特征变化规律的量化分析,可准确区分正常用火与火灾隐患,具有较高的环境适应性和识别准确率。
Based on video image analysis,a fusion algorithm of flame recognition and background interference filtering is proposed.Firstly,the visual features of flame images were analyzed systematically,with a focus on utilizing their high brightness characteristics.Through pre-processing techniques such as image grayscale,optimal threshold segmentation,and binarization,effective segmentation of suspicious areas was achieved.Aiming at the interference problem in complex environments,an innovative GMD three feature fusion recognition model was constructed:extracting the geometric,morphological,and dynamic features of flames as the discrimination basis.Experimental verification based on actual fire video data shows that this method can effectively identify flame targets in multiple scenarios,especially through quantitative analysis of GMD feature changes,it can accurately distinguish between normal fire use and fire hazards,with high environmental adaptability and recognition accuracy.
作者
何勇溢
祁倩
HE Yongyi;QI Qian(Jiangmen Polytechnic,Jiangmen,Guangdong 529000,China)
关键词
GMD特征
火焰识别
视频图像
变化规律
智能识别系统
GMD feature
flame recognition
video image
change pattern
intelligent recognition system
