Hardware Implementation and Feature Localization for Near-Infrared Cameras Calibration

The Near-Infrared Camera(NIRC) is insensitive to the visible light spectrum. The lack of corresponding calibration device and extraction method of the feature points limit its application. Based on this, We design a near-infrared calibration device by attaching LEDs to the printed circuit board through Surface Mounted Technology(SMT). However, near-infrared LEDs generate scattering halos when powered on, so we improved this adverse effect by applying a Gaussian Convolution Kernel. Then, we detected subpixel edges of the feature points based on Zernike Moment. Finally, we perform ellipse fitting on subpixel edges to obtain more accurate center coordinates of the feature points.