Automated Detection of Whale Blows in Infrared Video

The gray whale migration generates many hours of video that must be examined carefully to identify passing whales and get an accurate count of their numbers. Unfortunately, humans lack the necessary attention span. To solve this problem, a detector will be developed that can process hours of video, detecting whale blows while minimizing false positives and negatives. With an accurate record of the time and position of whale blows, it is possible to calculate the numbers of animals in each group, their average swimming speed, and, in the longer view, to measure the species abundance. Sonalysts, along with the University of Dayton’s School of Engineering and Whale Acoustics, will design a software architecture that encourages the loose coupling of pluggable components for: input processing of sensor streams, analysis algorithms, visualization systems, and storage components to support the workflow of whale population monitoring. This solution will draw upon our research-grade Heuristic Intelligent Storage System, a distributed, intelligent, large stream data ingest, data mining and storage system. This framework, coupled with image processing algorithms and whale population density estimation expertise, will define a video processing system that will produce more accurate results with less labor.