This paper is to study the tracking of mobile objects and associated coverage problems in visual sensor networks. When sensors equipped with cameras are randomly deployed in a monitoring environment, many cameras are involved in covering the same mobile object. Images of the object may be captured by different camera nodes in different orientations simultaneously, and the captured images are then sent back to a base station or server. However, achieving characteristic point coverage of objects invariably involves a great deal of redundant image data consuming the transmission energy of the visual sensor network. One approach is proposed to overcome this problem. The minimum number of camera nodes required for set coverage can be determined by predicting the direction and speed of the mobile object. The simulation results show that this protocol reduces transmission cost while preserving the maximum coverage range of mobile objects.