Churchill Beluga Boat Drone Imagery

Aerial imagery surveys are commonly used in marine mammal research to determine population size, habitat distribution and habitat use. Analysis of aerial photos involves hours of manually identifying individuals present in each image and converting raw counts into useable biological statistics. Our research proposes the use of deep learning algorithms as an assistive technology to increase the efficiency of the marine mammal research workflow. To test the feasibility of this proposal, the existing YOLOv4 convolutional neural network model was trained to detect belugas, kayaks and motorized boats in unmanned aerial vehicle (UAV) imagery. Computer-based object detection achieved an average precision of 61.17% for belugas, 98.58% for boats, and 95.97% for kayaks. We then tested the performance of computer vision tracking of belugas and manned watercraft in UAV videos using the DeepSORT tracking algorithm, achieving a multiple object tracking accuracy (MOTA) ranging from 37% – 88% and multiple object tracking precision (MOTP) between 63% – 86%. Results from this research indicate that deep learning technology can perform at a similar caliber as human annotators in beluga and watercraft detection and tracking, allowing for larger datasets to be processed within a fraction of the time.

航空影像调查在海洋哺乳动物研究中被广泛采用，以确定种群规模、栖息地分布及栖息地利用情况。航空照片的分析涉及对每张图像中存在的个体进行数小时的手动识别，并将原始计数转换为可用的生物统计学数据。本研究提出利用深度学习算法作为辅助技术，以提高海洋哺乳动物研究工作流程的效率。为了验证此提议的可行性，我们训练了现有的YOLOv4卷积神经网络模型，使其能够在无人机（UAV）影像中检测鲸鱼、皮划艇和机动船只。基于计算机的对象检测技术在鲸鱼检测方面达到了61.17%的平均精度，在船只检测方面达到了98.58%，在皮划艇检测方面达到了95.97%。随后，我们使用DeepSORT跟踪算法测试了在无人机视频中利用计算机视觉跟踪鲸鱼和载人水上交通工具的性能，实现了37%至88%的多目标跟踪准确率（MOTA）和63%至86%的多目标跟踪精确率（MOTP）。本研究结果表明，深度学习技术能够在鲸鱼和水面交通工具的检测与跟踪方面达到与人类标注者相当的水平，从而使得在大数据集的处理上能够显著缩短所需时间。