A New SIFT-Based Camera Calibration Method for Hybrid Dual-Camera

Camera networks, consisting of various types of camera systems. play an important role in security surveillance system. This paper presents a new calibration method for hybrid multi-camera system; particularly, a video surveillance system with a static camera and a dynamic camera which is used for environment-monitoring and security purpose. The first static wide angle camera covers the complete scene, whereas the second dynamic camera, Pan-Tilt-Zoom (PTZ) camera provides multi-view-angle and multi-resolution images of the complete scene. The new proposed calibration method is based on Lowe's Scale invariant Feature Transform (SIFT) algorithm and keypoints are selected based on the measurement of their stability. To improve the accuracy and robustness, a simple noise (unwanted keypoints) filtering technique using trigonometry theorem has also been adopted in the proposed system. From the obtained experimental results, it is shown that great improvement, in term of the determination and detection rate (from 55.71% to 94.87%) in camera networks calibration, has been achieved.