Non-planar target for multi-camera network calibration

Abstract

Rapid calibration of multi-camera systems using planar targets is typically impractical. Here we report a non-planar target for rapid calibration of inward-looking visual sensor networks (VSNs), designed using a novel VSN simulator and verified experimentally. The target consists of a large central sphere, surrounded by several smaller, rigidly mounted spheres. Rotation about an axis through the center of this sphere changes the location of the outer spheres, while keeping a fixed reference point in the center to permit extrinsic parameter extraction. Using the simulation, different target configurations are tested for field-of-view (FOV) coverage for each camera. Experimental results using a constructed operational prototype are consistent with traditional planar target calibrations, but require only a single view of the target to extract the intrinsic parameters.