Shallow and Deep Learning Architecture for Pests Identification on Pomelo Leaf

In this research, we proposed using two methods for the problem of pest identification from leaf patterns. Firstly, we use a traditional recognition shallow architecture with extracted three features: Color moments, Color correlograms, Zernike moments, then these features used to classifying by SVM algorithm. Secondly, we apply a deep convolutional neural network (CNN) for recognition purpose. We consider four different kind of pests in pomelo leaf: black bugs, snails, mealybugs, scales insects, each with 400 images and 700 images leaves are not pestilent. The introduction of a CNN avoids the use of handcrafted feature extractors as it is standard in state of the art pipeline and this approach improves the accuracy of the referred pipeline. These results show that both proposed methods achieve promising results and can be applied to identify the pests in reality.