Computer Stereovision System for 3D Tracking of Free-Swimming Zebrafish

Zebrafish (Danio rerio) has been gaining an enormous attention, as a promising model organism, in studies relevant to human diseases, drug tests, and other studies. However, behavior studies of this fish, using computer vision, is still challenging due to its small size, and fast and unpredictable movement. This paper proposed a new stereovision system for 3D tracking of free-swimming zebrafish. The proposed system that comprises two fixed cameras takes two images separated by a baseline. Centroids of the target to be tracked at both images are identified and used to extract the third coordinate of the 3D model using simple triangulation to derive the target's depth. Validity of the derived stereo model is assessed experimentally and compared to an existing orthogonal model. Performance of the developed stereovision system that is assessed experimentally demonstrated high correlation (82%) with that of an orthogonal vision system but with a significant reduction in the complexity of the system setup as well as less space requirement for experimentation.