Neural Network Scans 160,000 Images in Search for Missing Helicopter in Primorye
LizaAlert, a volunteer search-and-rescue organization, has used a neural network to analyze drone imagery in the search for a missing Robinson R44 helicopter in Russia’s Primorye region.
Drone operators conducted continuous aerial imaging with overlapping coverage, capturing each area two or three times. A neural network developed by SberLife Insurance processed the images and flagged more than 160,000 points of interest, including hundreds of frames with a high probability of showing the helicopter or its wreckage.
Mikhailov said rescuers in Russia’s Far East spend about an hour each shift reviewing the selected images. There are also about 60 hours of video recordings, which a LizaAlert review team will examine manually by the end of the year.
The crew of a Robinson R44 operated by Granat airline stopped communicating on June 21 in Primorye’s Terneysky District.
