Underwater video camera surveys are a well-used technique to collect information on the distribution and abundance of fish species. Using baits to attract fish can cause bias and may limit the number of species observed. When cameras are used without baits to sample fish assemblages the distance from camera can affect identification for various species, particularly as turbidity levels increase. Image resolution can be used to overcome distance from camera problems by recording at high resolution and digitally zooming during playback. However, recording at high resolutions can reduce battery life and produce large amounts of video data that subsequently requires manipulation, storage and maintenance. Therefore determining the lowest resolution required to correctly identify fish can reduce file size, lower storage costs, improve handling time and improve battery life.
To determine the most efficient resolution to identify fish species, we propose a survey asking participants to correctly identify a range of fish species from photos at differing resolutions (low, medium, high). Participants will self-assess their identifications skills placing themselves in one of three categories (no experience, limited experience identifying fishes, and a high level of identifying fishes). Images of fish species will be selected and categorised based on features that alter the difficulty of identification including shape, pattern, colour, background, and difficulty in differentiating from similar species. The minimum number of pixels required to identify fish species can then be used to determine the appropriate video resolution. The study will also inform the maximum distance a fish can be identified and how changes in turbidity will affect detection rates of different sized species.
The poster will promote this aspect of my PhD and provide an opportunity to recruit participants.