Visual Sensor Placement and Orientation Optimization for Surveillance Systems

More and more surveillance systems need visual sensors like cameras to monitor the activities in a large area. It is critical and non-trivial to explore the optimal placement of visual sensors due to the directional and larger sensing range (called FoV: Field of View) comparing to those traditional omnidirectional sensors like temperature sensors (called scalar sensors). Most current research on sensor placement optimization is based on maximizing coverage area to save cost. However, different locations may have different priorities at different time. To optimize the visual sensor placement according to the variant detection requirements, this paper proposes a new model of FoV within which the detection capability (DC) is non-uniformly distributed similar to human´s eyes. Based on the new FoV model, the visual sensor placement optimization problem can be solved by a nonlinear programming algorithm. Moreover, when the number and positions of visual sensors are known, the orientation of each visual sensor can be optimized through a least-square problem, which is efficient for real-time monitoring. The effectiveness of the proposed method was proved by experiments and the averaged relative error of DC is about 3.4%.