Multi-spectral imaging for artificial ripened banana detection

Abstract

Ripening is a natural process that makes fruits edible and nutritious. With increasing demand, the practice to employ artificial ripening of fruits have been increased recently in the market chain to fulfill the needs of the consumer. Artificial ripening not only reduces the quality of fruits but also increases the health-related risk especially Calcium carbide(CaC sub(2)), an artificial ripening agent, inherits the carcinogenic properties. Although the problem of detecting artificial ripening of fruits is important, the conventional methods based on chemical analysis are not feasible. In this paper, we present the use of multi-spectral imaging in eight narrow spectrum bands across VIS and NIR range to detect the artificial ripened banana. To present this approach, we introduce a newly constructed multi-spectral images collected from naturally and artificially ripened banana samples. The extensive experiments are performed on the large scale data set consists of 5760 banana samples by performing 10 fold cross-validation. The obtained average classification accuracy of 97.5 percent demonstrates the significance of multi-spectral imaging for differentiating natural and artificial ripened banana.