Advancing Acoustic Fish Tracking with Deep Learning

摘要

Tracking the movements of marine animals in space and time plays an increasingly important role in the sustainable management of ocean resources. In the context of fishery operations research, there is a desire to utilize fish tracking technology to gain a better understanding offish movement, allowing fisheries to optimize both operations and sustainability by improving the targeting of desired fish stocks while reducing bycatch.However, existing fish tracking solutions do not work well at the spatial scales of interest to the fisheries (～ 100 km~2). Innovasea, a leading ocean tech company headquartered in Halifax, N.S., is seeking to solve this problem through the Ocean Aware project supported by Canada's Ocean Supercluster. Innovasea has invented an improved fish tracking solution that employs high-frequency signals and a novel encoding scheme. The new solution has proven highly resilient to noise and hence enabled the detection of tagged fish at farther ranges and/ or in noisier acoustic environments than previously possible. Therefore, it has the potential to enable the tracking of fish at the spatial scales of open ocean fisheries. However, the new solution has proven difficult to scale because the novel coding scheme requires manual analysis to reconstruct accurate trajectories.