6533b85bfe1ef96bd12bb0e5

RESEARCH PRODUCT

Design and modelling a of bimodal camera system

subject

description

In computer vision, researchers face a challenging concern due to the limited field of view (FoV) when using standard cameras in capturing scenes. This problem of narrow FoV often encountered within applications that involve tasks which are related to the robot vision. Many authors have attempted, in different approaches, to solve this issue by fusing two cameras. However, they failed to improve the registration and fusing errors, where the obtained results still limited and not accurate enough. To deal with this limitation and decrease these errors, we propose a new method of camera calibration based on Interior-Point Optimisation (IPO) algorithm using transformation matrix estimation. This proposal conducts us to the design and modelling of a bi-modal camera system. Our calibration method leads to develop a hybrid camera system with full FoV, which can analyse scene details with high resolution within a window. The method combines multi-cameras setups composed of different characteristics and geometries. It increases the accuracy of the fusion between cameras while handling different levels of optical distortion. We evaluate our method in terms of quality of the sphere shape by comparing its accuracy to those obtained with two state-of-the-art approaches. The obtained results show that the registration errors recorded by our method are much lesser compared to other methods. As an example, we estimated an error of 0.1 with our method while we have 0.3 and 0.7 with method [199] and method [200] respectively. These results demonstrate clearly that our approach improves the accuracy of the state-of-the-art methods.