Automated block detection in lahars through 3-bands spectral analysis of video images

Lahars present a particular problem for automated block detection using machine vision, because their patterns and apparent sizes and shapes vary with the flow. To address this difficulty, we developed a simple algorithm to detect blocks in lahars at Semeru Volcano based on traditional RGB (Red Green Blue) 25 frame-per-second videos. The detection algorithm is based on the combination of two methods: (1) a determination of blocks by color segregation, and (2) objects movements' detection. After image processing, we converted the data into a signal that emphasizes the presence of large blocks. Results compare very well with video-images, since 100% of peaks of amplitude > 300 corresponded to large blocks traveling down the channel.This method ultimately aims at contributing to lahars automatic detection systems, where little funding is available, since it only requires a traditional camera and a computer.