A System for the Analysis of Images containing Atmospheric Discharges (SAIDA)

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 3 February 2014
Hall C3 (The Georgia World Congress Center )
Diego Matos Guedes Sr., Centro Universitário da FEI, Sao Bernardo do Campo, Sao Paulo, Brazil; and R. B. B. Gin and R. A. C. Bianchi
Manuscript (459.3 kB)

The SAIDA system started being developed in 2011, with the purpose of automatically analyze video files containing atmospheric discharges images, using C language, OpenCV image/video libraries, and the Artificial Neural Network Model. The purpose of this automation is to reduce the human cost, because of the large number of videos to be analyzed manually in the laboratory in which the project started. The FEI's Atmospheric Electricity laboratory has a tower on its top, with 7 video cameras installed that automatically records São Bernardo do Campo's sky, Brazil, 24 hours per day, 7 days per week, where 1 hour record contains about 100.000 frames. Based on result of previous projects, which were developed in the same environment, it was found that the Artificial Neural Network is the best technique to classify atmospheric discharges. The SAIDA system is able to detect lightning flashes with a margin of 95% and classify the lightning flashes as horizontal and vertical with a margin of 90% using the developed Artificial Neural Network. The SAIDA also identify the number of strokes and the luminosity level of each flash. The system also rejects irrelevant events, such as birds passing though the camera and being recorded, noises, and so on. In the Aug/2011-Dez/2012 period, results showed that the SAIDA system achieved a 50% speed improvement based on the manual analysis. Nowadays, the SAIDA system is being optimized to be faster and more accurate, with the development of new algorithms which reduces the misinterpretation of lightning flashes (false positive events), and the enhancement of the algorithms already developed. The system improvements will be tested and the results are promising.