College of Computer Science and Technology, Jilin University, Changchun, China

---

Abstract—In this paper, a novel smoke detection method based on wavelet energy and optical flow is proposed. Firstly, smoke motion is extracted by dual background modeling. Then, candidate smoke regions are determined. Thirdly, two basically smoke eigen-values are calculated by using wavelet transformation tools and Lucas-Kanade optical flow method. The motion directions are estimated in every small region. These two eigen-values can effectively express image texture and move orientation feature, respectively. Lastly, these two eigen-values are given different weights. The experimental results have proved that the proposed algorithm have a robust and better appearance.

Index Terms—Smoke detection; Optical Flow; Wavelet Transformation; Smoke Eigen-values; Dual Background Modeling.

[PDF]

Cite: Wen-Hui Li, Bo Fu, Lin-Chang Xiao and Ying Wang, Pei-Xun Liu, "A Video Smoke Detection Algorithm Based on Wavelet Energy and Optical Flow Eigen-values," Journal of Software vol. 8, no. 1, pp. 63-70, 2013.