Design and experiment of adjustment device based on machine vision for garlic clove direction

Abstract Garlic is an important economic crop whose planting areas has been annually increasing. Studies have shown that the direction of garlic cloves at the time of sowing has important effects on the germination time, yield and visual appearance of garlic. In order to ensure that the garlic cloves are upright when garlic is planted, an adjustment device based on computer vision has been designed to re-direct garlic cloves. As the garlic clove enters the adjustment device, images are collected by an industrial camera, from which the direction of the clove is identified through image analysis. In order to effectively identify the clove direction in images, a multi-feature algorithm is proposed. This algorithm is found to have higher accuracy than the single-feature recognition method, especially for garlic varieties with large individual differences. Once the clove’s direction is known, the adjustment device manipulates the direction of garlic clove to move it into the ideal planting position. Experimental results showed that the adjustment rate of ‘Jinxiang’ and ‘Cangshan’ garlic was 94.6% and 97.5%, respectively, and the average adjustment time was 1.13 s and 1.24 s, respectively. The proposed method of clove adjustment will not only contribute to improved precision planting outcomes for garlic but could also be extended to other crops whose yield levels are crucially affected by the seed direction.