Identification and Grading of Manage Using Image Processing

Fruit grading for commercialization is currently conducted through manual operations prone to inconsistent grading and human error, due to fatigue and the tedious nature of the task. Automation in agriculture especially for post-harvest yield inspection has played a vital role in reducing such error and at the same time, ensuring produce such as fruits and vegetables are graded based on commercial standards. This project has developed an image processing algorithm for a systematic maturity identification of "Mangga Susu Thai Gold" mangos. The criteria of mangos to be assessed by the grading algorithm are color and weight. Classification of these mangos are conducted based on standards set by Federal Agriculture Marketing Authority (FAMA) mango ripeness index, Project activities have started using a proposed activity flow for algorithm development using Python and the experimental chamber setup for actual mange data collection. Actual mango data collection is focused on gamnering data such as mange weight, and skin color. Experimental chamber for image acquisition is developed in building the image dataset. 40 random samples of "Mangga Susu Gold Thai mangos are sampled. Features such as maximum colour component values, pixel area and perimeter are extracted using a feature extraction algorithm for compilation into separate "sv" files for classifier and prediction models training and testing. 3 classes are selected using silhouette analysis in labelling the mango features as training references for classifiers. Classification is conducted where a combination of LAB and SVM yielded best results (100% accuracy).