摘要
针对目前森林火灾烟雾检测算法存在的不足,提出基于图像增强与多特征的森林火灾烟雾检测算法。该算法利用小波变换后不同尺度高频分量之间的相关度对小波系数进行修正,并通过小波重构实现图像增强;计算图像的分形特征和基于灰度共生矩阵的图像纹理特征作为烟雾识别的特征;把特征输入支持向量机,通过机器学习和大样本量实现火灾烟雾的识别。复杂背景下的森林火灾烟雾检测结果表明该算法能够有效实现森林火灾烟雾检测。
Against the disadvantages of forest smoke detection algorithm,an algorithm based on image enhancement and multi-feature of texture was put forward.The method used the correlation between high frequency components of different scales after wavelet transform to modify wavelet coefficients,and accomplished image enhancement with decomposing the image by wavelet;fractal features of computed images and texture features based on grayscale symbiotic matrix were computed as the features for smoke recognition;the features were input to SVM,and based on machine learning and large sample,the smoke recognition was realized.Forest smoke detection with complex background showed that the algorithm can accomplish the forest smoke detection effectively.
作者
朱磊
ZHU lei(Zhengzhou Fire Detachment, Henan Zhengzhou 450000, China)
出处
《消防科学与技术》
CAS
北大核心
2018年第2期225-228,共4页
Fire Science and Technology
关键词
森林火灾
图像增强
分形特征
烟雾检测
forest fire
image enhancement
fractal feature
smoke detection