Guava defect detection using hyperspectral imaging with fluorescent light source

Hyperspectral imaging with spectral region of 400nm to 1000nm has been built for guava defect detection. Fluorescent light was chosen as the light source for image acquisition. Due to the low reflectance spectrum of fluorescent light source in infrared region, analysis was carried out in visible region with spectral from 431nm to 642nm. Image captured in this experiment shown uneven illumination where half of the fruit has brighter area. Hyperspectral image was evaluated using principal component analysis (PCA) in order to select several wavelengths that could potentially be used in multispectral imaging system. Individual image was built with selected wavelengths and each image was went through thresholding. Dilation and erosion was applied to individual image and image combination was done to segment out the defect areas. Experiment shown that multispectral imaging is possible been built up by selecting the most contribution wavelength from hyperspectral imaging. Algorithm in this experiment could be further improved with the study of morphology operation in order to get the actual defects size.