Investigations on the detection of sulphur fumigation from copra images using features fusion method with artificial intelligence techniques

Abstract

Sulphur fumigation is one of the common methods of food newlinepreservation. This process leads to enormous health effects and also creates newlinemore impacts on the environment. Copra is a form of dried coconut that is newlineexposed to sulphur fumigation, maintains its colour, and restricts the newlinemicrobial agents for a long period. There is a need for technological newlineintervention to detect Sulphur-fumigated copra in order to avoid several newlinehealth issues when it is consumed and used as an external resource. Avoiding newlinethe traditional expensive and time-consuming process, the sulphur-fumigated newlinecopra is differentiated from the normally dried copra by combining image newlineprocessing and artificial intelligence techniques. newlineInitially, to examine the drying process from coconut to copra inhouse, newlineit is dried using a tray drier at regular intervals. Major and minor axes newlineare calculated from the copra images at different drying stages to differentiate newlinethe sulphur and normally dried copra. This work also encompasses the newlinecollection of images from the copra drying industry, pre-processed using newlinenoise filters and also segmented to analyse the fumigated process in copra. newlineThe collected images from the drying industry undergo extraction of newlinetexture features using Gray Level Co-Occurrence Matrix (GLCM) and newlineclassification using K-Nearest Neighbour (KNN), Adaptive Neuro-Fuzzy newlineInference System (ANFIS) and Neural Network Pattern Recognition system newline(NNPR). NNPR classifies the sulphur-fumigated copra with an accuracy of newline92.3%. Levenberg-Marquardt (LM) is the best learning algorithm for neural newlinenetworks, with an R2 value of 0.9407 and MSE of about 0.0287. newline newline

Description

Keywords

Citation

item.page.endorsement

item.page.review

item.page.supplemented

item.page.referenced