[Nondestructive detection of grey mold of eggplant based on ground multi-spectral imaging sensor]

Botrytis cinerea Pers. is a worldwide fungus. It is a severe threat to eggplant. Chemistry methods can do an accurate identification, however they are time-consuming, require execution by professionals and are high cost. The present paper presents the development of a ground based multi-spectral imaging sensor for the grey mold detection. Three channels (green, red, near-infrared) of crop images were acquired. Two algorithm systems were developed. The objective of the image processing is to obtain a binary image, which could point out the location of symptoms as accurately as possible. Two image processing methods were developed. It could be seen that method 1 can diagnose the symptoms accurately even if the symptoms are small while method 2 can only diagnose the symptoms with a certain extent area and the detection of symptoms is not very accurate. However, the images processed by method 1 showed some error diagnoses while the method 2 did not. It was concluded that both methods have some advantages and disadvantages. In the agriculture practice, the diagnosis environment will be more complex than in the greenhouse. Some things such as dry soil and perished leaf fragment will disturb the symptom detection by naked eyes when the grower stands away from the leaf. Two image process methods can diagnose the symptoms clearly although the position based on method 2 was a little deviated. It was concluded that the symptoms were well detected using multi-spectra imaging technique even there were some disturbances. Thus, multi-spectral imaging technique is available for the symptoms detection of grey mold on eggplant leaves.