Neural Networks-Based Framework for Detecting Chemically Ripened Banana Fruits

Abstract

Fruits are one of the most nutritious as well as naturally available foods, which are generally consumed in raw form. However, in the present competitive world, almost 80% of fruits are ripened using hazardous chemicals such as Calcium carbide (CaC$_2$) by greedy traders which cause serious health issues. Further, the regular consumption of fruits ripened using Calcium carbide can cause cancer due to the presence of traces of poisonous gases such as Arsenic and Phosphorous. On the other hand, in the existing literature, only few efforts are carried out toward identification of chemically ripened fruits using computer vision-based techniques. To solve this problem, this article proposes a new framework, which can identify the artificially ripened banana fruits by means of employing different visual features including color, shape, and edge histograms in an integrated manner. The proposed framework is implemented on a real dataset consisting of banana images using neural network-based algorithm. The experimental results in terms of accuracy, cross-entropy, and confidence level measures demonstrate the efficiency of the proposed system.