Connectivity, defined as the movement of individuals among populations or habitats, is a crucial ecological process that underpins the function of ecosystems. Animal movements promote a wide array of ecological outcomes, from genetic diversity to ecosystem recovery after disturbance. Therefore, it is necessary to have a clear understanding of the capabilities, scale, frequency, and locations of animal movements. In aquatic ecosystems, obtaining and analysing movement data is challenging because constantly changing environmental conditions hamper the use of traditional frameworks and methods. The study of animal movement in dynamic aquatic ecosystems also requires large volumes of data because animal movements cover different magnitudes, directions, and spatial levels of ecological organisation. As a result, new data collection and processing technologies are being developed to increase our understanding of this complex ecological process. Among new technologies, computer vision, machine learning and deep learning have received increased attention for their robust capabilities for rapidly processing large volumes of underwater imagery. Computer vision (CV) techniques are particularly suited to animal movement research because they can capture and process large amounts of raw data from underwater imagery. Despite their potential, CV techniques are only now beginning to be assessed in studies of aquatic animal movement, and their integration with appropriate statistical frameworks for behavioural analyses is required. In this thesis, I aim to identify, develop and apply CV techniques to measure animal movement in aquatic ecosystems. The focus is on measuring fish movements in connectivity corridors in estuarine systems. Fish movement research provides fundamental information about fisheries stocks, the status of protected areas, and the impact of habitat loss. Connectivity corridors are hotspots of fish migration, colonisation, feeding and reproduction. Yet much of fish behaviour in aquatic ecosystems remains hard to observe and timeconsuming to document manually. Connectivity corridors are a challenging but useful case study to test novel computer vision techniques for tracking fish. I first explored the current uses of CV techniques in fish movement studies and identified the benefits of CV for fish movement research. While the uptake of CV in fish movement studies has been slow, CV techniques provide two key benefits: 1) rapid, accurate and reliable datasets and 2) complementary information with traditional data collection techniques. Then, I developed a CV pipeline that automatically detects and tracks fish from underwater imagery. The pipeline has an 84% accuracy at detecting and subsequently tracking fish and provides large, raw movement datasets useful for ecological insight. To translate the raw movement data into behavioural events, I developed a new methodology for applying structural equation models to infer latent behavioural states of fish from observations of behavioural indicators. The statistical models accurately predicted behavioural events such as foraging (a slow, sinuous movement near the substrate) and fine-scale migrations (a fast, directional movement near the surface). Finally, I applied the CV pipeline to study the fine-scale movement and predation dynamics of fish at piped weirs in multiple estuaries. I used multi-species occupancy models to characterise fine-scale temporal changes in predator-prey co ccurrence and determined if behavioural differences could be detected at different categories of predator-prey co-occurrences. The fine-scale temporal changes of predator-prey co-occurrence varied among sampling days and locations, but I nevertheless identified that prey exhibited significantly different behaviours that depended on the probability of co-occurring predators. Overall, I bridged the gap between the development and application of new technologies for ecological research. CV can help us improve our understanding of critical interconnections among habitats and help researchers and managers increase data availability into conservation ecology and decision making. CV has the capacity to inform data-driven decisions that directly influence the health and productivity of marine ecosystems.