General coordination of the formation is a critical asset in swarm robotics. When specialized agents are considered, with robots dedicated to specific tasks, formation control and efficient transition in the leadership of the swarm must be achieved. A task switching approach is presented in this paper and is formulated by evolving the definition of specialization to match with targets recognition, such as detecting special landmarks via embedded sensors, or the presence of an object in the environment. Specialization zones are generally defined as circular areas around each detected target corresponding to a specialized task to be dealt with by a specific robot. Entering within the influence area of a specialization zone triggers the switching of the leader of the formation. Under the assumption of specialized agents, each robot in the swarm is the specialist associated with one given specialized target. Appropriateness, smoothness and safety of the transition remain the major factors of performance considered. The framework is further refined by making the targets, and therefore the corresponding zone of influence, dynamic, which leads to the consideration of combined specialization areas. The robots control and leader switching processes are studied in relation with targets recognition, and adapted to accommodate real-time dynamic leadership and formation control as targets move and interact one with each other. The proposed system is validated in simulation to demonstrate that the group of robots effectively coordinate themselves around targets and dynamically allocate the appropriate specialized agent.