Five methods to calibrate overlapping surveillance cameras1

Calibration of surveillance cameras requires using appropriate methods to handle different limitations, such as large distortion of the camera lenses, because of a wide field of view. Another limitation is blur in the image which can decrease the sharpness because of fixed focus lenses. To deal with these limitations a fully automatic calibration method from Svoboda is compared with a specialized technique for internal calibration and with a method for external calibration. The former is the Matlab Calibration Toolbox which uses a planar checkerboard pattern as a calibration object. The latter calculates the external camera parameters with a-priori knowledge of the internal parameters. Finally Svoboda’s method is compared to another method which estimates the rotation of cameras with the use of an inertial sensor. Through experimental evaluation it will be shown that the specialized methods for internal and external calibration will outperform the fully automatic calibration method from Svoboda, both in terms of accuracy and robustness under the stated limitations. has been funded by the Vienna Science and Technology Fund (WWTF) through project ICT08-030.