Very frequently, research in biology and ethology requires visual tracking of live animals, such as insects, rodents and fish. The challenge is particularly difficult in the aquatic environment because of the light scattering effect of water: reflections and refractions, low contrast and resolution images, noise, and also due to the unpredictable animal behavior. This paper describes a new tracking method that is based in start and stop detection. When the animals stop or move very slowly, a precise tracker based on AdaBoost classifiers is applied. Otherwise, when they move faster, a more efficient process based on background detection is used. A control system is responsible for deciding which method should be used in each instant. The experimental results using sea cucumbers (Echinodermata, Holothuroidea), zebrafish and rodents, show the efficiency and accuracy of the proposed method, which is able to deal with complex situations.