Local Vegetable Detection: A Computer Vision Approach

Abstract

In this thesis work it describes about a unique system that can detect local vegetable using computer vision approach. To enhance the identification process and promote the usability of the graphical user interface compared to existing manual system is the main motive of the system. Different convolutional neural networks have been tested and retrained to classify an object. It’s our goal to implement a system which can suggest optimal recipes based on the identified vegetables. There are verities of methods that can detect the local vegetables, but for particular and user friendly process we decide to use Random Forest Classifier method to identify the local vegetable. We use three feature extraction methods which are Hu Moments, Color Histogram and Haralick Textures to extract the features. We calculate the accuracy by the help of confusion matrix. As the domain of this research model, local vegetable detection are classified, and 96% accuracy are achieved which can obviously help in our modern life along with proper introduction.