Modular calibration board for automated calibration of hyperspectral cameras

Abstract

This work presents an adaptable and fully automated approach for calibrating hyperspectral images (HSI) acquired using a pushbroom (line-scan) system. The proposed method enables simultaneous spectral, intensity, and spatial calibration while determining the spectral and spatial resolution of the system, all within a single scan. Calibration is performed using a flexible multi-purpose calibration board designed to provide necessary reference data. Spectral calibration is achieved using an array of LEDs spanning the spectral range from visible to shortwave infrared (Vis-SWIR), enabling precise wavelength registration. Additionally, a narrowband laser diode determines the system’s spectral resolution at one wavelength. Intensity calibration corrects for illumination inhomogeneities and pixel sensitivity variations using a highly reflective Restan plate. Spatial calibration ensures an accurate aspect ratio using a checkerboard pattern, while modulation transfer functions (MTFs) derived from line grids and edge transitions quantify the spatial resolution. A grayscale calibration module evaluates sensor intensity linearity. The calibration process is fully automated through a Python-based software package, which utilizes an ANN-driven object detection model to accurately extract pixel coordinates of each module. This modular approach allows for precise calibration of line-scan systems and can be adapted to other systems, e.g., snapshot. The method is evaluated using Newtec Vis-SWIR Oculus hyperspectral line-scan cameras, demonstrating consistent and accurate calibration.