An improved algorithm for smoke layer identification in building fire condition

The smoke layer identification is an important issue in building fire hazard analysis. The existing ways of smoke layer recognition are reviewed, and a new smoke layer recognition means based on Otsu method and its 2D refinement are presented in this paper. All the methods are tested in an implementation case, and the obtained smoke layer regions are compared. The result indicates that the Otsu method is more adaptable and credible than the existing methods, and its 2D refinement inherits the advantages of original Otsu method, with enhanced capability of reflecting the smoke motion trends. The 2D Otsu method is suitable for smoke layer recognition.