At a time where industrial processes are becoming more and more automated, the demand for non-contact measuring systems is increasing rapidly. In addition to capacitive and inductive sensors, optical sensors in particular are used in a large number of applications. Cameras are of high importance in the field of optical sensors, as they can be used in many cases for process monitoring and quality assurance in a wide variety of products. By using multispectral cameras, the information content obtained can be increased significantly compared to monochrome or RGB-cameras, since these are taking images in several spectral regions even outside the visible range.
In this work, a multispectral camera which analyzes twelve spectral regions, especially in the visible range, is designed and built. The essential functionality is based on illuminating the measured object with light-emitting-diodes of certain wavelengths and taking the images with a monochrome camera
The illumination is characterized at different ambient temperatures and the achieved reproducibility of the illumination spectra is assessed. Subsequently, images are taken from a reference object with a known reflection spectrum and a dark calibration of the camera is carried out.
To demonstrate the functionallity of the multispectral camera, images of different measurment objects are taken and their reflection spectra are assessed. Using a method for largely information-preserving data reduction, these images and spectra are further processed and the results obtained are discussed.
Keywords: multispectral camera, multispectral imaging, LED characterization, reflection spectrum, principal component analysis
August 16, 2022