This thesis investigates the coordination approaches of multiple mobile and autonomous robots, especially resource-limited small-scale UAVs, for the surveillance of pre-defined ground targets in a given environment. A key research issue in surveillance task is the coordination among the robots to determine the target's time varying locations. The research focuses on two applications of surveillance: (i) cooperative search of stationary targets, and (ii) cooperative observation of moving targets. The objective in cooperative search is to minimize the time and errors in finding the locations of stationary targets. The objective of cooperative observation is to maximize the collective time and quality of observation of moving targets. The thesis identifies various factors and application scenarios that affect the performance of multi-UAV surveillance systems and proposes a distributed strategy for merging delayed and incomplete information collected by a team of UAVs. The thesis uses an analytic derivation of the number of required observations to declare the absence or existence of a target in a given region and to plan the paths of UAVs. It performs an exploration of the algorithmic design space and analyzes the effects of centralized and distributed coordination on the cooperative search of stationary targets in the presence of sensing and communication limitations. Additionally, the thesis proposes an approach based on the quad-tree data-structure to model the paths of the UAVs for observing moving targets with different resolutions.