Abstract:
Major economic and production losses of yield occur because of diseases on the tomato
production. Nowadays there are several diseases, such as septoria leaf spot (S. lycopersici
speg), anthracnose, early blight, and late blight that significantly affect this crop and
constrain its production. To increase production and quality of tomato, it is important to
manage such a harmful disease. For the management of such diseases, it is necessary to
detect a specific disease. In our country, farmers are not professionally trained to detect
symptoms of diseases, and thus they need to get expert advice. Ironically experts use naked
eye observations or visual assessment, which is prone to error to estimate disease severity. In
this thesis work, a new approach is introduced for detection and measurement of the severity
of S. lycopersici speg on tomato leaf. Otsu and adaptive thresholding techniques were
employed to segment the leaf area and the spots of lesion regions, respectively, in tomato
leaves. The proposed method (image based) includes image acquisition, diseased leaf area
calculation, total leaf area calculation and calculation of disease severity. In this study, 144
infected leaves were collected from the glasshouse at various stages after early detection and
the disease severity levels of each leaf were done by two mechanisms (image-based and visual
eye estimation). When compared, both methods showed that 87.50% of the samples were
under the disease severity levels of one, was conducted by digital imaging technique, and
there were no samples under the disease levels of four or above for both techniques. The
relative error of disease severities (%) measured by image based and the corresponding visual assessment technique was 34% i.e. the average accuracy of agreement of the two
methods was 66% which was confirmed by this experiment.