Ficusdeltoidea (Jack) Moraceae Varietal Identification Using Statistical Recognition Approach
Plant species identification is one of important application in pattern recognition. Selection of relevant features for classification is a classical problem in statistical pattern recognition and data mining. There are two main branches in solving this problem which is by using feature selection or...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English English |
| Published: |
IDOSI Publications
2017
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/18476/ http://umpir.ump.edu.my/id/eprint/18476/ http://umpir.ump.edu.my/id/eprint/18476/1/Ficusdeltoidea%20%28Jack%29%20Moraceae%20Varietal%20Identification%20Using%20Statistical%20Recognition%20Approach.pdf http://umpir.ump.edu.my/id/eprint/18476/7/Ficusdeltoidea%20%28Jack%29%20Moraceae%20Varietal%20Identification%20Using%20Statistical%20Recognition%20Approach%201.pdf |
| Summary: | Plant species identification is one of important application in pattern recognition. Selection of relevant features for classification is a classical problem in statistical pattern recognition and data mining. There are two main branches in solving this problem which is by using feature selection or feature extraction. Currently, in statistical plant species recognition domain, researchers focus mainly on providing an automatic system using feature extraction method such as Principal Component Analysis. This paper presents a hybrid of filter and wrapper in feature selection as well as an empirical comparison of feature selection wrapper method using Sequential Forward Selection and feature extraction method using Principal Component Analysis on a benchmark of 420 images of Ficus deltoidea leaf with 6 varieties. At first, the leaf images are processed using image pre processing techniques. Then, 23 leaf features are extracted such as shape, texture and vein. Finally, in classification process, different feature selection and feature extraction techniques are computed using Support Vector Machine and Nearest Neighbor classifiers. The recognition results reveal that a combination of filter and wrapper approach in feature selection outperformed the other approaches for F. deltoidea varietal identification. |
|---|