Automated Iracking Of Multiple C-Elegans

This paper presents a method for model based automated tracking of multiple worm-like creatures. These methods are essential for accurate quantitative analysis into the genetic basis of behavior that involve more than one organism. An accurate worm model is designed using the geometry of planar curves and nonlinear estimation of the model's parameters are performed using a central difference Kalman filter (CDKF). The filter can naturally be expanded to estimate the locations of multiple worms and determine when they are occluding each other. The predicted location of the models at each iteration allows for an efficient method to determine the regions that are undergoing occlusions. Experiments on actual C. Elegans mating sequence data demonstrate the quality of the proposed method.