Marine imagery is important for assessing the distributions and relative abundance of fish in low accessibility habitats. Taxon specific detection biases are pervasive sources of data uncertainty arising when the selected camera type is not suitable for capturing the diagnostic characteristics of a target fish species. However, there are no standardised methods of objectively assessing and comparing these biases between types of marine imagery.
We developed a method of comparing detection biases between stereo-still and video imagery based on morphometric characteristic visibility. Using a case study of deep towed camera imagery from CSIRO’s RV Investigator SEA-MES voyage (IN2023_V05), we evaluated detection biases of 13 taxonomic groups of demersal fishes.
The most informative camera parameters for detection were resolution for stereo-stills and capture frequency for video. The diagnostic characteristics that influenced characteristic visibility the most were high-contrast patterns, size and movement. Only the rattail detection counts were significant towards stereo-still imagery due to the taxa's smaller sizes. Both stereo-still and video imagery are therefore suitable for detecting all other fish taxa.
This method of assessing detection biases has broadscale applicability across various marine habitats and species and will aid scientists in selecting suitable camera types for optimal detection of target fish.