Provide a Method Based on Image Processing and Artificial Neural Network for Using on Automatic Adjustment of Onion Topper

Tractor mounted onion topper is one of the technologies used to remove onion leaves. The position of the blades in this machine plays an important role in the quality of the onion topping. In the case of communication between the physical characteristics of the bulbs and the length of the leaves remaining after the topping, it is possible to provide methods for automatic adjustment of the blades. In this research, a method was proposed to calculate the diameters of the bulbs before topping using image processing. Then the remaining leaf length on onions was estimated in topping process using the Multi-Layer perceptron (MLP) and the bulbs were classified in four groups according to the size of the leaves remaining by using the Learning Vector Quantization (LVQ). The statically parameters such as root mean square error, mean absolute error, specificity, precision, sensitivity and accuracy were used to evaluate the networks. The results showed that the designed neural network predicted leaf cutting height with RMSE and MAE values ​​of 0.025 and 0.01 respectively. Also, the classification of onions was carried out with a total accuracy of 91%. The results of this research can be used to set up automated mechanisms of cutting blades in onion topper.