E-catalog        

Библиогр.: 35 назв.

Burning and glowing firebrands generated by wildland and urban fires may lead to the initiation of spot fnes and the ignition of structures. One of the ways to obtain this infonnation is to process tliennal video files. Earlier, a number of algorithms were developed for the analysis of the characteristics of fu'ebrands under field conditions. However, they had certain disadvantages. In this regard, this work is devoted to the development of new algorithms and their testing. For this purpose, semi-field experiments were conducted using an apparatus for generating firebrands to obtain the necessary theimal video files. The thermograms were processed to create an annotated IR video base that was further used to test the detector and the tracker. To detect firebrands in the thermograms, the Laplacian of Gaussian and Difference of Gaussians (DoG) algorithms were tested. To estimate the accuracy of detectors, an original approach involving the application of the FI score was used. The analysis showed that both algorithms can provide the necessary accuracy for the detection of firebrands and are comparable in time and accuracy, but the DoG algorithm is easier controlled and implemented. Different firebrand tracking algorithms have been developed and tested. In particular, a Hungarian algorithm-based tracker that can track firebrands between frames with high accuracy is implemented. The comparison of the algorithms showed that Hungarian algorithm-based trackers more accurately tracked the movement of particles.