Recognition of Mangoes and Oranges Colour and Texture Features and Locality Preserving Projection

Abstract

In this paper, a recognition system for classifying and predicting mangoes and oranges has been developed. With the use of support vector machine (SVM) and decision tree algorithm (DTA), classification was done on the images of the fruits gathered locally and publicly into defective, ripe, unripe for local and ripe and unripe for public datasets. The proposed system involves some stages such as, pre-processing, feature extraction and classification, implementation was done by resizing images, removal of background distortion, and extracting colour and texture components for each image. Histogram and Haralick texture features were extracted from each pre-processed image as a feature vector and used as inputs for transformation. Also, the locality preserving projection (LoPP) was computed on the extracted local features and used as feature for classification. A One-against-One multi-class SVM and fine tree DTA classifier with 30% held out was used for classification. The performance of the proposed method was tested on 328 mangoes and oranges sample images obtained locally and 149 images of public data. Based on the experiment carried out various success rates were recorded on different levels but an excellent classification accuracy of 100% and 92.9% was obtained on the public dataset, 91.3% and 90.2% and 91.1% on the local dataset, 91.3% and 92.2% on the local dataset using LoPP for mango and orange predictions. Mangoes and oranges were categorised, results obtained was 88.6%, 80.4% and 85.6% for public, local and LoPP on local datasets..