Home
Mining Environment
Research Projects
Education
Case Studies
Tools
publications
Staff
ITSC
Contact Us

 

NASA logo

ITSC logo
Information Technology and Systems Center The University of Alabama in Huntsville
Huntsville, AL 35899 (256) 824-6868
info@itsc.uah.edu

UAH logo

 

Lightning Detection Using OLS

The objective of the lightning detection procedure is to identify lightning streaks in night time portions of Operational Line Scan (OLS), which is carried by Defense Meteorological Satellite Program (DMSP) satellites. The OLS sensor produces a visible and thermal image. Due to the scan frequency of the sensor, lighting shows up as bright horizontal streaks, typically of one or two pixel width, in the visible images. Unfortunately, other things also show up as bright spots. City lights and moonlight reflected off of clouds both produce bright areas in images. Lightning is only visible in nighttime images of OLS since daylight saturates the sensor.

After evaluating several alternatives, an approach based on morphological filtering and gradient detection was selected. Both the visible and thermal bands of the OLS sensor were used in the detection process. Gradient detection and dilation were used on the thermal image to find the areas that were in or near areas of cloud cover. All other areas were removed from consideration, and thus ground features such as roadways and city lights were screened out. Gradient detection in the direction of satellite propagation was then applied to the visible image, and bright horizontal streaks were extracted. Finally, texture measures were used to identify areas of patchy cloud cover, which exhibited many small bright streaks, some of which were horizontal and thus produced false positives. A small set of training and test images was available for which human experts had classified lightning events. A genetic algorithm was used to tune the parameters of the classification procedure using the training images. An objective function provided a measure of classification performance. False positives were assigned a penalty five times greater than that assigned to false negatives. Results of training and testing are shown in the figure below.

Collaboration with Domain Expert:
Dr. Steve Goodman (NASA/MSFC GHCC)