The semi-automated fish census system

Underwater Visual Census

Fish species count, population density and biomass are crucial variables for assessing the health of a tropical marine environment. Underwater Visual Census (UVC) conducted with reef areas measure these variables to provide quantifiable metrics for determining the success of reef protection and rehabilitation initiatives. The monitoring results serve as inputs for the formulation of strategies for improving reef resilience and productivity of reef fish communities.

The Problem

UVC methods suffer from the following biases that tend to underestimate fish density measurements: fish behavioral response to diver’s presence (agitation or attraction), diver’s skill in accurately counting and identifying fish, and diver fatigue. Quite often the diver is burdened with the decision to include or exclude an individual fish as it moves into or away from the sample area. In difficult cases, the diver has to decide whether the fish was inside or outside the area when it was first sighted.

The Solution

The semi-automated fish population density, species identification and biomass estimation system from raw input underwater videos. Through this method, there is minimal disturbance in the water environment, thus greatly reducing observer bias. With Fish-I, even divers with minimal knowledge of fish can obtain high quality population and species distribution measurements using a stereo camera rig and fish video analyzer software that we have developed. A permanent visual record of the census is available for later review and archiving.

How it works

How it works
The fish-i system consists of two major components: a camera rig consisting of 2 to 8 cameras oriented parallel to the seabed. Mounted on a simple stand with 4 legs; and a video analyzer software that performs stereo analysis and image recognition on the collected videos. The main output of the software are video clips containing estimates of fish size, density, and species diversity that may be exported in CSV format for end users.

Sample Output

Sample Output

Where we've worked

Visited Sites