Computer vision-based detection of marine vessels in the Canadian Arctic using visual and acoustic data

点击翻译按钮获取中文摘要

We aim to develop a robust and accurate system for the autonomous detection of marine vessels in the western Canadian Arctic using visual and passive acoustics data. Among the potential outcomes are the creation of novel technologies integrating computer vision and acoustic signal processing, such as computational systems that can recognize both regular-sized as well as small (50 pixels or less within images of 921x1080 pixels) vessels under a variety of environmental conditions. From an ecological standpoint, the outcome of this project will provide insights that will enable our partners to assess the influence of increased vessel traffic on local marine mammals such as beluga whales. This project will have a significant impact, since it will provide a cost-efficient and more accurate alternative to the traditional data analysis of acoustic and visual data, which is typically performed manually by environmental scientists. This is a topic of high importance to both Canada and our partner organizations, the Wildlife Conservation Society Canada (WCS) and Environment and Climate Change Canada (ECCC). The geographic region of interest has recently suffered an increase in anthropogenic activities, due to the sea ice reduction. Therefore, in order for our partners to carry out their core missions of informing the public and advocating for wildlife conservation policies and planning, the ecological impacts of marine traffic have to be better quantified, monitored and studied. Our proposal makes a direct contribution towards the quantitative analysis of marine traffic.