Plant disease identification using autoencoder
Plant diseases limit the crop production and have received more attention from experts and farmers. Plant disease identification is carried out by experienced people or needs microscopic identification. However, trained people or professionals are not always available, and the manual approach may le...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2021
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/99441/1/JaniceOngAunNeeMKE2021.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Plant diseases limit the crop production and have received more attention from experts and farmers. Plant disease identification is carried out by experienced people or needs microscopic identification. However, trained people or professionals are not always available, and the manual approach may lead to bias or errors, costly and time-consuming, especially when some of plant disease symptoms are similar. It has also not easily been understood and identified that attacking crop could be due to parasitic organisms like fungus or bacteria besides the insect. To reduce the damage on the crops, plant disease early detection should be carried out in an automated way for early detection, prevention and control. Many methods have been proposed to do automated detection, but it is not easy to target which feature is the best for the classification. Thus, the objective of this project is to develop an automatic feature extraction method in identifying the severity of two types of plant diseases, namely early blight and late blight, which are caused by microorganism attacks. The main classifier module will be governed by autoencoders as an automatic feature extraction to identify the plant diseases. The MATLAB software was used to develop the autoencoder module. With the data set ready from Plant Village leaf images, this project identified two plant diseases into three severity levels, low, mild and severe at 72.7% accuracy. |
|---|