Optimization of Lighting Parameters for Imaging with the Optical Identification Module

Hyperspectral analysis is a non-invasive method that reduces losses and improves fruit quality by accurately identifying defects during sorting. A key requirement for obtaining reliable data is stable, uniform illumination provided by specialized light sources with a controlled spectrum. The integration of such systems into automated production lines minimizes human error, increases productivity, and supports the sustainable development of the agricultural sector. ( Research purpose ) The study aims to substantiate the selection of parameters for the hyperspectrometer and the light source within the illumination system. ( Materials and methods ) The optical identification module used in this study consists of stepper motors, rack-and-pinion and screw drives with bearings, and a platform with rubber rollers, the speed of which is regulated by a three-phase motor powered through a frequency converter. The stand suspension can move both horizontally and vertically at a preset speed. SpecGrabber and CubeCreator software was used to collect and process data during scanning, enabling subsequent image analysis using Gelion software. ( Results and discussion ) A hyperspectrometer was selected for the identification module, and the main light sources for the illumination system were determined. ( Conclusions ) The calculated luminous flux is 934 watts per square meter, which falls within the sensitivity range of 100–1500 watts per square meter for the Complementary Metal-Oxide-Semiconductor (CMOS) detector. This confirms that the camera can capture hyperspectral data under the specified exposure and illumination conditions. It was determined that four halogen lamps should be installed in the illumination module, providing an illuminance level of 3010 lux. At this lighting level, reliable spectral graphs were obtained for both healthy and diseased fruit areas. Additionally, the short exposure time of 2.1 milliseconds per spectrometer frame resulted in a total scanning time of less than 2 seconds.