A Study on Image-Processing-Based Flame Height Estimation and Its Applicability to Full-Scale Facade Fire Tests

This study examined the applicability of an image-processing-based method for estimating the flame height in full-scale facade fire tests. Fire experiments were conducted using the same wood fuel source under free-burning conditions and spatial conditions simulating KS F 8414, and the recorded videos were analyzed using Python-based OpenCV. The flame regions were detected using the HSV color space, morphological image processing, and a seed-based region growing algorithm. The maximum and mean flame heights were estimated on the basis of the relationship between the reference heights and pixel coordinates in the images. The original images, Final Flame Masks, and overlay results were also compared to assess the adequacy of the detected flame regions and flame-tip positions. The temporal changes in the flame height could be quantitatively estimated under different spatial conditions, thus demonstrating the applicability of the proposed image-processing-based method for flame height estimation.