Decision Support System for In Situ Melon’s Fruit Harvesting Time Based on Fuzzy Logic and Single Shot Detector (DSS).

Abstract

Decision Support Systems (DSS) are wildly used in agriculture, industry, and healthcare, due to their efficient information management and decision-making activities. Melon has a considerably large genetic variability and is widely used in scientific research such as biology and genetics. In this study, we aimed to develop a system to detect the right harvesting time for melon fruits because the quality cannot be maintained after harvesting. Hence, it would be best if the harvesting is carried out at the right time, which will not affect the fruit harvested earlier. Secondly, melon fruit harvesting is done every day and not on the same day for all fruits due to environmental factors. The harvest DSS will classify melon’s categories on the plant regarding melon’s skin color by the guidance of the expert. The essential sensory factor for freshness and maturity is color, as mentioned in many studies. We use the skin color to classify melon fruit while it is on the plant into three categories: Ripe, About to Ripe, and Under Ripe, using a melon image. Firstly, localizing the melon on the plant using SSD and extracting its skin color. The input of fuzzy logic is the extracted color channels means values. The extraction of color is a challenging job because some fruits blocked with plant’s stem, root, or leaves, so, we divided the process of reaching the purest melon skin into several steps: detecting the melon fruit on the plant, segmenting the detected melon and removing the unwanted part from the skin, and finally, applying image normalization and calculate mean values for melon’s skin color. In this study, we achieved 91-100% accuracy in classifying melon ripeness levels. All images source were captured using iPhone rear camera.