Research On Garlic Automatic Cutting Technique Based On Machine Vision

China is largest garlic producer and exporter in the world,However,China's garlic exports are dominated by primary commodities,resulting in low prices for export products and greatly reduced economic benefits.In order to increase the added value of garlic for export,it is necessary to completely remove the garlic roots.At present,domestic and foreign research and production units' mechanization research on garlic mainly focuses on the sowing and harvesting of garlic.As a result,the garlic beard cutting process does not have a good automatic equipment to match it,so it is still necessary to manually remove the garlic roots.Therefore,it is of great significance to develop automatic whisker cutting equipment suitable for the further processing of garlic.In order to realize automatic and precise beard cutting,reduce the damage to the garlic fruit parts,and obtain the garlic root position shape and size information accurately and quickly becomes a key process.In this paper,combined with the actual cutting technology requirements of garlic deep processing,the typical shape and size parameters of garlic collected in real time are used to predict the shape and size of garlic roots.The research content involves the image acquisition method,image preprocessing,image segmentation,feature extraction and prediction recognition algorithm of garlic samples.Firstly,by analyzing the color space model of the collected image,this paper selects the R channel image in the RGB color space for image segmentation.Then,based on the characteristics of the garlic image,a threshold interpolation segmentation algorithm and a circular segmentation algorithm based on morphological and circularity are designed to realize the garlic fruit.Division of the area.Extract the 12 absolute shape features of the fruit area of all garlic sample images to establish the garlic feature training set,and establish the regression prediction model to achieve the prediction of garlic root size,and the experimental test is completed.The test results show that the average absolute percentage error between the predicted value and the actual value of the established BP neural network model is 3.59%,and the average absolute percentage error obtained by the SVR model is 4.62%.Two types of prediction models can be used to predict the circumference of garlic roots and garlic,indicating that these shape features are effective as predictors.The research results can provide data support for the precise automatic cutting of garlic.