vioft2nntf2t|tblJournal|Abstract_paper|0xf4ff091429000000b331010001000100
Agriculture is the largest economic sector in India. Coconut is one of the most demanded fruit amongst all. The dried coconut, copra is the main source of coconut oil. Naturally it contains 70% of moisture content and it is dried to about 7% for production of coconut oil. The sulphur is added as preservative which acts as anti-microbial agent for preventing bacteria, fungus etc. Sulphur is a toxic food preservative which restricts lung performance and leads to direct allergenic reactions. The survey of World Health Organisation says that 65% of asthmatic children are sensitive to sulphur and 75% of children exposed to sulphur exhibits changes in their behaviour. The sulphur fumigation over coconut affects human both externally and internally. Fumigation leads to cancer and environmental pollution. In order to prevent this devastating effect, copra is examined using image processing. The proposed idea is to identify the presence and percentage of sulphur region present in copra. The region of interest is segmented by method of superimposition thereby segmenting white layers in copra. The RGB colour features are extracted to differentiate the sulphur added copra from normal copra. The coconut is dried under 60? in a tray drier and shapes of copra decreases at regular interval of time are extracted using image processing. The decreasing percentage of shape features are measured to identify the sulphur added in the copra. The k-means clustering technique is used to discriminate the copra at different levels. The segmented patch area is measured to determine the percentage of sulphur present in copra. The percentage of sulphur over copra is divided into three levels (low sulphur added region, medium sulphur added region and high sulphur added region). The K-Nearest Neighbour classification is also used to classify the sulphur added copra at different levels. The proposed algorithm classifies the sulphur added copra at three different levels with 86% accuracy.