Directional sensors in visual sensor networks (VSNs) spend most of their energy for two major tasks: sensing and communication. As sensing regions of directional sensors often overlap, the sensing task can be rotationally scheduled among different sensors. Thus, efficient scheduling of directional sensors became one of the fundamental interesting problems attracting many researchers over the past decade. One possible scheduling scheme can be achieved by partitioning the directional sensors into a maximal number of disjoint set covers each of which is capable of monitoring all the targets using minimal number of sensors. The derived set covers can be activated successively in a round robin fashion. As a result, the larger the number of disjoint sets, the longer the lifetime of the network. Moreover, fault tolerance can be easily achieved by activating more than one set covers at the same time. In this paper, we solve the problem of finding maximum possible disjoint set covers using minimal number of directional sensors. We formulate the problem as an Integer Linear Programming (ILP) problem. The ILP partitions the sensors into optimum number of disjoint set covers. As ILP does not scale well over large problem instances, we design a new target oriented heuristic and modify couple of existing heuristics for generating near-optimal disjoint set covers. We compare the performance of all three heuristics using extensive simulations.