Abstract
In the early stage of fruit rot in satsuma mandarins, there is little change in the appearance of the fruit compared with that of normal fruit, so that it is difficult for an inspector to detect the fruit rot even if using an RGB color camera. Therefore, in order to detect fruit rot, we developed a device that is easily able to measure a spectral image from visible to near infrared. Using the device, the average absorbance of a normal part or a rotted part was calculated from the spectral image, and multiple linear regression (MLR) was performed. From the result of MLR, three wavelengths were selected as variables for the calibration model and used to create an image of each sample by applying the reflectance of the three wavelengths. It was found that the rotted parts of 27 out of 36 samples could be detected using these images.
