Automatic Measurement of Fish Size Using Stereo Vision

In aquaculture, in order to optimize the feeding of fish, it is essential to know their size with precision. Most of the current systems either do not automatically determine the size of the fish in real time or the error in measurement is excessive. The proposed system enables automatic and real-time measurement using artificial vision techniques. To attain this goal, a two stages methodology was developed with the aim of optimizing time and development costs. In a first stage the results obtained in test environments only were collected and compared in order to obtain a good basis for the development and improvement of the measurement software. In a second stage, work was started in a real-world environment and a recursive process was started in which, based on the information obtained, both the test environments and the measurement software were improved. In this work the results obtained in test environments are shown, leaving for a later work the problem of measurement in the field.