Research On Identification And Prediction Of Sugarcane Nodes In Black Box Of Intelligent Pre Cutting Workstation

At present,the pre cutting of sugarcane is an important process under the sugarcane planting mode.The identification of sugarcane nodes is the key technology of intelligence.Only when the nodes are identified can the intelligent cutting be carried out better,which can not only increase the enthusiasm of workers,reduce the intensity of operation,but also save costs and improve the yield of sugarcane.In this paper,a large number of sugarcane images are collected on the sugarcane pre cutting seed machine,the SURF Algorithm is proposed to be applied to the sugarcane node recognition,and three improved BP Neural Network models are proposed to predict the unrecognized sugarcane nodes at the end of sugarcane,the design of the sugarcane node prediction system reduces the operation difficulty,and provides a theoretical basis for the intellectualization of sugarcane.This paper is from the following aspects:Firstly,after the preliminary investigation,it is concluded that the target of sugarcane pre cutting seed machine is half a year old sugarcane seed in Guangxi.Theoretical analysis and research are carried out on machine vision and image processing technology,image pre-processing is carried out on sugarcane,contour and interested area of sugarcane are extracted,recognition rate and error recognition rate are taken as the evaluation criteria by using statistical knowledgeSecondly,The diameter and width of sugarcane are measured statistically,and the SURF Algorithm is applied to the identification and matching of sugarcane knots.The feature points of sugarcane are extracted,the collected sugarcane images are tested experimentally,the recognition rate and error recognition rate are calculated,the coordinate distance of each sugarcane knot and the unrecognized coordinate distance of sugarcane knots are obtained and counted,through experiments,the recognition rate of SURF Algorithm is 76%,the error recognition rate is9%,the response time is 87 ms,the recognition rate of improved SURF Algorithm is 92%,the error recognition rate is 2%,the response time is52 ms,and the recognition effect is improved.Thirdly,according to the requirements of double sugarcane nodes,the first seven sugarcane nodes of a whole sugarcane are used to predict the coordinate information of unrecognized sugarcane nodes at the end due to the cover of sugarcane leaves or soil.and the average absolute error is calculated Difference,average relative error,The results show that the average absolute error of BP Neural Network is 15 mm and the average relative error is 1.05%.The average absolute error of BP Neural Network is 10.7 mm and the average relative error is 0.76% after the improvement of L-M Algorithm.The Genetic Algorithm is added to L-M-BP model to predict the unrecognized sugarcane data,The absolute error of GA-L-MBP model is 8.3mm,and the average relative error is 0.59%.By comparing the prediction results of the three methods,it is concluded that the best prediction model GA-L-M-BP is better than BP and L-M-BP model in the prediction of sugarcane knots.This method has reference value for the prediction of sugarcane knots in the pre cutting seed machine.The optimal algorithm combination is used to predict the sugarcane end patching cutter.Design the prediction system of sugarcane saving to facilitate the followup study.Finally,the experiments show that the improved SURF Algorithm can significantly improve the identification of sugarcane nodes,and the prediction accuracy of the improved BP Neural Network meets the predetermined requirements.This paper provides a reference for the identification of sugarcane pitch,the prediction of sugarcane pitch data and the prediction of supplementary cutter of sugarcane pre cutting seeder,which has a certain practical significance.