Automated Analysis of Leaf Shape, Texture, and Color Features for Plant Classification.

The main purpose of this research is to apply image processing for plant identification in agriculture. This application field has so far received less attention rather than the other image processing applications domains. This is called the plant identification system. In the plant identification system, the conventional technique is dealt with looking at the leaves and fruits of the plants. However, it does not take into account as a cost effective approach because of its time consumption. The image processing technique can lead to identify the specimens more quickly and classify them through a visual machine method. This paper proposes a methodology for identifying the plant leaf images through several items including GIST and Local Binary Pattern (LBP) features, three kinds of geometric features, as well as color moments, vein features, and texture features based on lacunarity. After completion of the processing phase, the features are normalized, and then Pbest-guide binary particle swarm optimization (PBPSO) is developed as a novel method for reduction of the features. In the next phase, these features are employed for classification of the plant species. Different machine learning classifiers are evaluated including k-nearest neighbor, decision tree, naive Bayes, and multi-SVM. We tested our proposed technique on Flavia and Folio leaf datasets. The final results demonstrated that the decision tree has the best performance. The results of the experiments reveal that the proposed algorithm shows the accuracy of 98.58% and 90.02% for the "Flavia" and "Folio" datasets, respectively.