A custom high-resolution snapshot computed tomography imaging spectrometer (CTIS) capable of capturing hyperspectral images - with dimensions 456 × 471 × 417 covering the wavelength range from 438 to 740 nm - through an iterative reconstruction algorithm is tested in a real-world application. The hyperspectral cubes are reconstructed using a sparse implementation of the expectation-maximization algorithm. In the literature, CTIS has previously been limited to laboratory environments with control of external parameters, ideal imaging conditions and imaging of relatively simple objects. Therefore, knowledge about the performance, capabilities and limitations of CTIS in real-world applications (outdoor illumination) is sparse. The real-world application comprises imaging of apples in apple trees and fake plastic apples, which are visually indistinguishable to the naked eye. The results show that real apples are distinguishable from fake plastic apples based on the recon- structed spectral signature. Furthermore, an upgraded CTIS with dimensions of 277 × 278 × 430 is used to investigate the feasibility of the system through the acquisition of CTIS images in the controlled environment of the laboratory. Additionally, several shortcomings of the current system are highlighted and discussed, and improvements to circumvent the shortcomings are proposed. This work shows the capabilities and potential of CTIS in real-world applications and paves the way for future real-time snapshot hyperspectral imaging.
|