Design And Experimental Study Of Positive Bud Garlic Mechanism Based On Visual Recognition

With an annual production of about 20 million tons,China’s garlic exports account for more than 70% of the world’s total.Garlic has made important contributions to farmers’ income and national foreign exchange.As a natural antibiotics,the garlic industry has great development potential.The large and morphological rules of Chinese garlic have obvious competitive advantages in the international market,which is closely related to the agricultural craftsmanship of garlic single-planting.Studies have shown that,compared with the positive bud seeding,reverse seeding and horizontal seeding of garlic require longer seedling emergence time,lower garlic weight,and higher garlic head abnormality rate.The garlic positive bud mechanism is one of the core technologies of the garlic single-grain positive bud seeding,which is of great significance for improving the commodity and international competitiveness of garlic in my country.Most traditional garlic positive bud planting machines use a purely mechanical positive bud mechanism,which relies on the inherent characteristics of garlic seed distribution and bud tip shape to determine and adjust the orientation of the garlic bud tip.However,this approach generally yields better results for garlic varieties with more pronounced features in a specific characteristic.Most traditional purely mechanical positive bud mechanisms exhibit better performance when generating positive buds for garlic seeds with plump shapes and long bud tips,such as those found in the Cangshan garlic variety.When facing with hybrid garlic seeds with slender shapes,as well as short and bent bud tips,the positive bud generation performance is less than ideal.To address the significant differences in positive bud generation performance across different garlic varieties using garlic planting machines,and to enhance the applicability of positive bud mechanisms for garlic seeds,this paper has proposed a machine vision-based method for identifying the orientation of garlic seeds.Related technologies and devices have been developed and studied,with the specific research contents outlined as follows:(1)In order to solve the problem of low efficiency of the existing garlic sowing machine fixed type positive bud mechanism,this study proposes a double-wheel disc drive mechanism based on the principle of parallelogram mechanism as the basic structure of the positive bud mechanism.In conjunction with the visual identity system,the positive bud mechanism positive bud process is established and the functions of the various parts of the positive bud mechanism are clarified.(2)In response to the problem that the existing purely mechanical positive bud mechanism has difficulty in determining the orientation of garlic seeds with inconspicuous characteristics and short and curved sprout tips,a method is proposed to identify the orientation of garlic seeds in an vertical position by machine vision and to use the mechanical structure to sprout the garlic seeds,to design and improve the structure and function of each part of the mechanism and to build a three-dimensional model of the mechanism.The results of the Adams motion simulation of the mechanism were analyzed to achieve the function of carrying the garlic seeds and turning them over for positive bud,and it was found that the mechanism had difficulties in adapting to the positive bud of a multi-row garlic planter.(3)In response to the problem that the vertical posture recognition positive bud mechanism is difficult to realize the positive bud of multi-row garlic planter,a positive bud mechanism that recognizes the orientation of garlic seeds in a horizontal lying posture is improved and designed.The study has designed structure and functions of each part of the mechanism,and established a three-dimensional model of the mechanism.The results of the Adams motion simulation of the mechanism are analyzed which determined the functional feasibility.(4)The research of target detection algorithm based on deep learning convolutional neural network is carried out.Based on YOLOv5 target detection algorithm,the method of identifying the left and right orientation of garlic seeds in horizontal lying posture has been proposed.Using Cang Shan garlic and hybrid garlic seeds as the identification objects,we build image data sets with labels and trained the algorithm network.Analysis of the network training results showed that the algorithm accuracy and recall were stable at 99%,the average precision mean(m AP@0.5)was stable at 99% and the reconciliation precision mean(m AP@0.5:0.95)was stable at 97%,which showed the algorithm recognition is at a high level.(5)The physical prototype of the horizontal lying posture recognition positive bud mechanism has been established,the prototype has been assembled and the test bench has been built.The test bench was used to conduct visual recognition tests to establish the average accuracy of garlic seed orientation recognition of 97% for different varieties,sizes and double wheel disc speeds,which verified the accuracy of the visual recognition algorithm.The positive bud test has been carried out to achieve an average positive bud rate of more than 90% for different varieties and sizes of garlic seeds at low and medium speeds.The results shows this mechanism has overcame the limitations which traditional purely mechanical positive bud mechanism is not ideal for positive bud of different varieties of garlic seeds.