作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2012, Vol. 38 ›› Issue (06): 204-206. doi: 10.3969/j.issn.1000-3428.2012.06.067

• 人工智能及识别技术 • 上一篇    下一篇

基于小波变换和稀疏光流法的火灾烟雾检测

姚太伟 a,王慧琴 a,b,胡 燕 a,b   

  1. (西安建筑科技大学 a. 信息与控制工程学院;b. 管理学院,西安 710055)
  • 收稿日期:2011-06-22 出版日期:2012-03-20 发布日期:2012-03-20
  • 作者简介:姚太伟(1986-),男,硕士研究生,主研方向:图像处理,机器视觉;王慧琴,教授、博士生导师;胡 燕,博士研究生
  • 基金资助:
    陕西省科学技术研究发展计划基金资助项目(2011K17- 04-01);陕西省教育厅专项基金资助项目(08JK319);陕西省西安市碑林区科学技术基金资助项目(GX1104)

Fire Smoke Detection Based on Wavelet Transform and Sparse Optical Flow Method

YAO Tai-wei a, WANG Hui-qin a,b, HU Yan a,b   

  1. (a. School of Information and Control Engineering; b. School of Management, Xi’an Univ. of Architecture and Technology, Xi’an 710055, China)
  • Received:2011-06-22 Online:2012-03-20 Published:2012-03-20

摘要: 传统图像型火灾烟雾检测算法不适用于存在灯光、水蒸气等噪声的图像。为此,通过分析早期火灾烟雾的运动规律,采用分块和背景自适应相结合的方法,提取运动前景,然后分别在RGB空间和HSV空间建立烟雾的颜色模型和亮度变化模型,分割出烟雾疑似区域。在灰度空间使用二维离散小波变换对烟雾疑似区域及对应的背景区域进行能量分析,提取高频和低频能量的比值。用Lucas-Kanade稀疏光流算法跟踪运动区域,获取烟雾运动区域的平均偏移量和相位分布作为烟雾识别特征,根据烟雾识别规则,判断监控区域是否有火灾发生。实验结果表明,该方法具有较强的鲁棒性,能够提高烟雾检测的准确率。

关键词: 块分割, 背景自适应, 小波变换, 稀疏光流, 平均偏移量

Abstract: Through analyzing the movement rule of early fire smoke, traditional single background difference method and adjacent frame difference method can not effectively solve the light and noise influence and background can not update itself. This paper presents a new segmentation method based on block and background self-adaptive, extracting the movement part. In order to segment the smoke suspected area, it establishes the smoke color model in RGB space and the brightness change model in HSV space respectively. The high/low frequency energy eigenvalues is obtained through discrete wavelet transform in gray space. Meanwhile, the average displacement distance value and phase value are obtained based on L-K optical flow algorithm. The eigenvalues are combined and used to determine whether the suspicious region is smoke or not. Many experiments are carried out and result proves the robustness of the method is strong. The method improves the accuracy of smoke detection.

Key words: block segmentation, background self-adaptive, wavelet transform, sparse optical flow, average offset

中图分类号: