Recognition of garlic clove orientation based on machine vision

Abstract

To address the problem of low garlic clove-head-turning rates in mechanized sowing of garlic, this study proposes an algorithm for identifying garlic clove orientation using machine vision. First, the algorithm performs preprocessing operations such as image enhancement and color space conversion; then, it identifies and labels the garlic clove regions by determining the connected regions. Each garlic clove region is derived for independent analysis, and three garlic clove features of each clove are calculated. Then, based on the distinctiveness of each feature parameter with clove recognition, comprehensive feature parameters are obtained by fusing the information of the three feature parameters. Finally, garlic clove head is identified on the basis of these comprehensive feature parameters. Furthermore, two garlic varieties, ‘Cangshan’ and ‘Jinxiang’ garlic, are tested by capturing images of both single and multiple garlic cloves. The experimental results indicate that the average recognition rate of the algorithm is 97.44%, and the average running time is 0.61 s. The correct recognition rate of ‘Jinxiang’ garlic is 94.51%, while that of ‘Cangshan’ garlic is 100%. In summary, the algorithm has strong adaptability and high accuracy, and it can provide a useful reference for an intelligent garlic clove-head-turning mechanism for garlic planters.