Cabbage Discernment Using CNN for Vegebot Application

Abstract

Agriculture could provide a unique scope for the headway of the robotic system. Robots are a perfect substitute for human resources to a great extent as they deploy unmanned sensing and machinery system. During the period of plantation and harvesting of crops, the involvement and participation of humans get decreased nowadays. This in turn reduces the cost of labour with improved yield. Agricultural robotics has unique challenges when compared to the robotics in the industrial applications. Cabbage is a cruciferous and most fashionable winter vegetable grown in India. Though India is one of the leading cabbage-producing and cabbage-consuming countries, there is no evidence of automation in cabbage harvesting. The conventional method of using knife for cabbage harvesting is practised till date. Automation in cabbage fields is risky, and its complication level is too high. Complexity starts with identifying the cabbage head size for harvesting. Using a sharp knife for cutting the cabbage head and leaving the stem uneven at some places, even it may damage the cabbage itself. Sometimes, it leads to irregular cabbages, which is futile to meet the market standards. Due to the increased difficulty in harvesting cabbage through labours, automatic cabbage identification and harvesting robot is the only solution. As a part of automation, convolutional neural network is implemented for cabbage localization and classification. In this paper, cabbage localization and classification is performed using CNN, by collecting the different datasets from the cabbage field. The field test results show that the cabbage classification is achieved successfully.