A classification model for lightning images without branches based on the tortuosity metric

• Published in: Modeling earth systems and environment 2024-04, Vol.10 (2), p.2447-2461
• Main Authors: Orozco-Gomez, Diego, Bolanos, F., Herrera-Murcia, Javier, Espinosa-Bedoya, Albeiro
• Format: Article
• Language: English
• Subjects: Chemistry and Earth Sciences; Classification; Computer Science; Detection; Earth and Environmental Science; Earth Sciences; Earth System Sciences; Ecosystems; Electromagnetic fields; Environment; Lightning; Lightning detection; Math. Appl. in Environmental Science; Mathematical Applications in the Physical Sciences; Natural phenomena; Original Article; Performance indices; Physics; Shape; Statistics for Engineering; Storms; Tortuosity; Wind waves
• ISSN: 2363-6203; 2363-6211
• DOI: 10.1007/s40808-023-01900-5

Lightning detection is important in order to develop protection systems aimed to living beings and to assure electronic and electrical devices function correctly in electrical storms. Currently, technologies aimed to detect lightning that are based on magnetic and electromagnetic field measurements and do not always provide information about the event occurrence. Using cameras to acquire photographs and video recordings of lightning could serve as a redundant system in lightning detection. In order to determine whether or not the object detected visually corresponds to lightning, a shape descriptor could be used. Because images of lightning have an elongated, thin, and polygonal shape, this paper proposes using a classification model for lightning images without branches that can be achieved by applying a tortuosity metric. The method to find the appropriate classification model begins with measuring tortuosity indexes using a set of segmented and thinning lightning images without branches to select the appropriate range for each tortuosity index. These measurements are then used in two image sets to evaluate the classification model performance: one corresponds to lightning photographs, and the second set has images associated with other natural phenomena such as rainbows, light reflections in clouds and on water surfaces, and wind waves. The results are promising; the model shows a Precision of 85.3448%, Recall of 91.6666%, an F1-Score of 88.3928% with the combination of the tortuosity indexes τ 2 and τ 4 .