Intra-Class Recognition of Fruits Using DCNN for Commercial Trace Back-System

Intra-class recognition of fruit using image processing and computer vision techniques, is considered challenging task due to similarities between various type of fruits and external environmental changes such as lighting. Mainly sub-type of the same fruit shows a much similarities between each other so, it's more difficult task to distinguish than when different types with same color of fruits are involved. It creates mismatches between training and test set. The problem become more difficult when lighting changes which tend to change the actual characteristic of the fruits like contour shape. To solve the problem of intra-class recognition, we proposed a deep neural network model with only few layers which learn optimal features from an input image adaptively. Our proposed method has been tested on 2 different fruit types and 2 sub classes of each fruit. To show the robustness of the classifier on sub classes of fruits. Extensive experiments have been done on dataset of 5602 apple fruit images & 4292 kiwi fruit images. The proposed model shows satisfactory performance across different fruit types and sub-types.