|
Detection of Chlorotic Cassava Leaves using Image Processing and Discriminant Analysis |
|---|---|
| รหัสดีโอไอ | |
| Creator | Wanrat Abdullakasim |
| Title | Detection of Chlorotic Cassava Leaves using Image Processing and Discriminant Analysis |
| Publisher | Thai Society of Agricultural Engineering |
| Publication Year | 2558 |
| Journal Title | Thai Society of Agricultural Engineering Journal |
| Journal Vol. | 21 |
| Journal No. | 2 |
| Page no. | 50-59 |
| Keyword | Cassava, Crop monitoring, Leaf chlorosis, Image processing, Discriminant analysis |
| ISSN | 1685-408X |
| Abstract | Cassava (Manihot esculenta Crantz) has been an important industrial crop for Thailand with a tendency of increasing production scale. A profitable cassava production in the future requires not only effective cultivation practices but also an efficient crop protection system, suggesting the necessity of an automated pests and diseases monitoring technology. Modern surveillance operation is usually performed by field imagery and analysis to detect atypical symptoms on the plants. The objective of this study was therefore to assess the feasibility to detect diseased cassava plants in situ by means of conventional image analysis. An image processing technique has been developed for distinguishing healthy and chlorotic leaves which is common symptoms of many cassava diseases. Color images of healthy and diseased cassava leaves were captured in fields with a resolution of 640?480 pixels and overlaid with squared grids of 80?80 pixels. Various color indices including red (r), green (g), and blue (b) chromatic coordinates, contrast indices rg, gb, (gb)/?rg? and 2grb, and hue (H), saturation (S), and intensity (I) were calculated for each grid. The discriminant analysis of principal components method was used to classify the healthy and chlorotic leaves. Total accuracy of the image classification was then evaluated based on Brier score. The results showed that the developed algorithm correctly identified 84.70% of healthy leaves and 79.90% of chlorotic leaves, giving a Brier score of 0.1654. A critical comparison with the neural network classification in an earlier study done by the authors is herein discussed. |
