Research On Identification Technology Of Farmland Seedling Belt Based On Machine Vision

The identification technology of field seedling belt is an important technology in precision agriculture research.With the rapid development of machine vision technology,it is widely used in field weed identification,variable spraying and field navigation of agricultural vehicles.The identification technology can not only improve the efficiency of farming,but also improve the precision of agricultural machinery.Therefore,the search for a real-time and robust crop row extraction algorithm based on machine vision technology is of great significance for accurately identifying crop rows in the field seedling belt and improving the rational utilization of agricultural resources.In this paper,based on the above background,the seedling maize crops were selected as the research object,and the seedling belt identification technology was further studied.The main research contents and results are as follows:(1)Pre-process the collected images.Firstly,the improved ultra-green method was used to gray-scale the color images of the collected crops.When the green component G was 1.27,the processing effect was the best.Then the adaptive Otsu method with automatic threshold value is used for image binarization.In view of the fact that the collected images are susceptible to noise interference,the median filtering method with strong suppression effect on salt and pepper noise is selected for image denoising.Finally,in order to meet the real-time requirements of subsequent field operations,the left and right edge center line method with fast processing speed and clear extraction effect was selected to extract the feature points of crop rows.(2)Aiming at the problem that the clustering number of traditional k-means clustering algorithm needs to be manually specified and the algorithm is easily trapped into the local optimization by the random initial clustering center,an improved k-means clustering algorithm is proposed,which greatly improves the classification accuracy of the clustering algorithm.An improved least square method was proposed for line fitting in crop line centerline detection.The experimental results show that the line fitted by the improved algorithm deviates less from the actual crop line and the fitting precision is higher.(3)The field seedling belt identification system based on MATLAB GUI platform is constructed,and the functions of image acquisition,image preprocessing,crop line detection and so on are realized.(4)The results of experiments in the research field of Jilin Agricultural University show that the crop row recognition rate of the field seedling belt recognition system is 93%.The mean errors of Hough Transform algorithm,Random Hough Transform algorithm and improved k-means clustering algorithm are 2.1035°?1.6374°?1.1411°,respectively.The results show that the improved detection line algorithm meets the real-time requirement and has higher precision.The results of spraying experiments show that the relative error of the system can be controlled within 5%when the working speed of agricultural machinery is 1.8km/h and the spraying quantity is 59.7kg/hm~2?387.3kg/hm~2.The results show that the decision release of the recognition system of the upper computer is accurate,and it is of practical significance to the field operation system of agricultural machinery and the field of agricultural machinery spraying.