| Autor: |
MANGLA, Monika, PUNJ, Deepika, SETHI, Shilpa |
| Předmět: |
|
| Zdroj: |
Annals of the Faculty of Engineering Hunedoara - International Journal of Engineering; Aug2021, Issue 3, p121-128, 8p |
| Abstrakt: |
Plants are very susceptible to the onslaught of diseases; diseases of plants can be distinguished from the buds on the leaves. The disease is visually recognizable as it has a unique color and texture features. But visual recognition has the disadvantage is that it is difficult to recognize the similarity between a disease and a disease other so as to affect the inaccuracy of the identified disease. On research this is where a system that can identify diseases and provide information is a treatment solution in the prevention or treatment of leaf diseases through digital image identification using supervised classification. The imagery will be previously identified through the RGB colour transformation (Red Green Blue) to HSV (Hue Saturation Value), HSV (Hue Saturation Value) to Greyscale, and feature extraction processes texture of GLCM (Gray Level Co-occurrence Matrix). The extraction results of the texture feature are classified over Gaussian Processes and Fuzzy C-Mean using CNN (Convolutional Neural Network) for determining the disease caused by various factors on leaves. The test was conducted with 200 samples of various category of leaves imagery, 160 imagery as training data and 40 imagery as test data. Test results show that the CNN using Gaussian function method has an average percentage of 97.5% accuracy, 95.45% precision, recall 95% and error 5%. While FCM filter produces 95% accuracy, precision 90.83%, recall 90% and error 10%. From the test results, it can be stated that at this research Gaussian Function is a better classifier than Fuzzy C Means. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
