Geolocation, the process of determining where a photo or video was taken, is a major problem with widespread applications. Counter terrorism, environmental monitoring, disaster relief and robot navigation all have a need for reliable photograph geolocation. But current methods of geolocation are severely limited.
I developed a computer program that geolocates photographs by horizon matching. The algorithm compares the horizon line in a photograph against millions of virtual horizon lines simulated from terrain elevation data, looking for the best match to determine the location of the camera within several hundred meters. The algorithm is fast, accurate, and reliable. In a test of 100 different photos, 83% geolocated successfully. Unlike most other algorithms, it can geolocate any ordinary photograph—the only requirement is a visible horizon line.
After working on the algorithm for over a year, I competed at the Intel International Science and Engineering Fair (ISEF) and won second place in the computer science category. Around the same time, I submitted a paper to the 2012 International Geoscience and Remote Sensing Symposium (IGARSS). The paper was accepted for publication and I presented my work at the conference in Munich, Germany in July. In the fall, I was selected as a regional finalist in the Siemens Competition in Math, Science, and Technology. I won the regional competition at MIT and went on to compete at the National Finals in Washington, D.C., where I won third place and a $40,000 scholarship.
I filed a provisional patent application and my work is currently patent pending. I plan to continue developing it in the future, along with other research projects. For example, I’m currently a student intern at the National Institutes of Health, where I work in a chemistry lab developing drugs to treat colorectal cancer. I’ll be attending Harvard University in the fall with a major in engineering.