| Scope and Method of Study. The thesis proposes algorithms to perform distributed calibration of a camera sensor network. Unlike other calibration techniques which involve non-linear computations, we propose simple techniques which can be implemented on low power processors. The concept of pin-hole camera model and generation of a reference coordinate system that are used in this thesis, were inspired from some existing work. A cooperative friendly target equipped with a dead reckoning position sensor, moves around in the environment and communicates its coordinate information at three distinct locations in the field of view of a camera sensor node. The sensor node uses the concept of slope angle of a straight line to solve for its location and orientation. This method does not require any initial guesses to calculate the external parameters of a camera. The correctness of the method was verified experimentally and its scalability for a large scale sensor network was validated through simulation.;Findings and Conclusions. During the literature review, we came up with the following findings. Most of the state of the art calibration techniques have the disadvantage that they either require the presence of landmarks to perform calibration or those which do not use landmarks either give only the relative locations and orientations or require complex non-linear computations. The algorithms proposed in this thesis are relatively fast and involve simple mathematical calculations. Also, the proposed calibration algorithm will not require any initial guesses and hence avoids the possibility of getting stuck at a local minimum. As the expressions for the final mathematical solutions are fairly simple, they can be programmed on low power processors which are at least able to support floating point arithmetic operations. |
