Tracking moving objects is one of the important applications in sensor networks. A key issue in designing such sensor applications is to reduce energy consumption as the available power of a sensor is supplied by a battery and is limited. Many methods have been proposed to track moving objects, however, the improvement in energy consumption is still insignificant, particularly for those moving objects with irregularly moving behavior. In this paper, we tackle the object tracking problem under the theme of minimizing energy consumption for sensor networks. We first design an alert-enabling sensor model so that distributed sensors are able to participate in the decision making for tracking objects. We then propose a novel alert-based object tracking (AbOT) scheme, to minimize the consumed energy for the object tracking applications. Our performance results reveal that AbOT can significantly save energy and still have high accuracy under various conditions.