"We are developing software that will monitor dangerous algae and various biochemical anomalies for public health," said Cai, who noted other applications for this software might range from studying cancer tumor growth to tracking serial killers. "The other very important part of this research is to adapt our findings to homeland security where we may use this method to track the bioterrorism activities pertaining to the nation's waterways."
Cai said his NASA/ESTO-funded research team is developing a spatiotemporal data mining system to track harmful ocean objects from NASA's SeaWiFS satellite images and NOAA's oceanographic data. These harmful ocean objects, known as red tide, include a naturally occurring microscopic algae Karenia brevis, which form only in the Gulf of Mexico and release toxins deadly to fish and marine mammals. The red tide algae are also potentially harmful to humans if ingested through unmanaged shellfish. On the worst days, the irritating vapors from red tide algae or the noxious smell from the rotting carcasses of poisoned fish linger along the coast, impacting tourism.
Nationwide, all forms of red tide algae cause commercial fisheries to lose $18 million a year. More than $20 million was spent on public healthcare to handle thousands of cases in which humans ingested shellfish poisoned by red tide between 1987 and 1992, according to NOAA's 2005 Economic Statistics Report.
At present, NOAA researchers must manually scan through thousands of images in order to test and evaluate models that determine where red tide is impacting both humans and sea life.
"We are pleased to work with Carnegie Mellon r esearchers to improve our monitoring of red tide," said Richard P. Stumpf, an oceanographer with the NOAA National Ocean Service and a co-investigator for this project. "In fact, a huge, huge problem for us is trying to validate any algorithm for the harmful algae. Evaluating the imagery is very time-consuming."
Stumpf said he is collaborating with Carnegie Mellon researchers to devise a way to simply convert harmful algae data into an image field that can be easily accessed to map out the deadly algae's location and where it may travel. The new software is designed to incorporate multiple computer vision algorithms and the domain knowledge from NOAA scientists.
"Spatiotemporal data mining extracts changing spatial patterns from continuous data flow. It is a rapidly growing field, which includes inverse physics, machine learning and human-computer interaction," Cai said.