Image Tracking Study on Courtship Behavior of Drosophila

Background: In recent years, there have been extensive studies aimed at decoding the DNA. Identifying the genetic cause of specific changes in a simple organism like Drosophila may help scientists recognize how multiple gene interactions may make some people more susceptible to heart disease or cancer. Investigators have devised experiments to observe changes in the gene networks in mutant Drosophila that responds differently to light, or have lower or higher locomotor activity. However, these studies focused on the behavior of the individual fly or on pair-wise interactions in the study of aggression or courtship. The behavior of these activities has been captured on film and inspected by a well-trained researcher after repeatedly watching the recorded film. Some studies also focused on ways to reduce the inspection time and increase the accuracy of the behavior experiment. Methodology: In this study, the behavior of drosophila during courtship was analyzed automatically by machine vision. We investigated the position and behavior discrimination during courtship using the captured images. Identification of the characteristics of drosophila, including sex, size, heading direction, and wing angles, can be computed using image analysis techniques that employ the Gaussian mixture model. The behavior of multiple drosophilae can also be analyzed simultaneously using the motion-prediction model and the variation constraint of heading direction. Conclusions: The overlapped fruit flies can be identified based on the relationship between body centers. Moreover, the behaviors and profiles can be correctly recognized by image processing based on the constraints of the wing angle and the size of the body. Therefore, the behavior of the male fruit flies can be discriminated when two or three fruit flies form a close cluster. In this study, the courtship behavior, including wing songs and attempts, can currently be distinguished with accuracies of 95.8% and 90%, respectively.