Research On Bad Bud Detection And Seed Cutting Parameters Of Sugarcane

Sugarcane seed quality directly affects the yield of newly planted and long-rooted sugarcane,and seed selection is a prerequisite for ensuring the normal emergence of sugarcane seeds.During the process of sugarcane growing,harvesting and planting as sugarcane seeds,the sugarcane buds may experience phenomena such as insect infestation,mechanical damage,decay,and mold.Seed selection requires identifying and removing the bad sprouted sugarcane seeds.At present,the recognition of sugarcane seed bad bud depends on manual recognition,which is labor-intensive,inefficient,and costly.The project designs a sugarcane cutting and seed selection device that automatically recognizes and removes bad buds,and studies the algorithm and cutting mechanism of sugarcane seed bad bud recognition.Based on the improved YOLOv4 sugarcane seed bad bud recognition model,according to the recognition and classification of sugarcane bud features,the sugarcane cutting machine avoids bud cutting and eliminates bad bud sugarcane seeds to achieve automatic seed selection.The main research contents are as follows:(1)Obtaining high-quality images is the prerequisite for successful recognition and classification of cane buds.Build a visual image collection platform,select Guitang 44 as the sugarcane sample,and collect 790 images of sugarcane sample containing individual buds.In order to increase the diversity of sugarcane image datasets,sample augmentation was performed on the images,Based on the criteria for judging the quality of sugarcane buds,annotate and construct a dataset of 1580 sugarcane buds.(2)Improved YOLOv4 algorithm for detecting and recognizing sugarcane sprout datasets.By adding a Lightweight Attention Module(CBAM)to the YOLOv4 backbone network,the network’s ability to extract sugarcane bud features is enhanced and the impact of background noise is reduced;Using the K-means algorithm to recluster the dataset,generate anchor boxes that match the characteristics of sugarcane sprouts,and improve the accuracy of sugarcane seed bad bud recognition;Replacing the existing standard convolutions in path aggregation networks with deep separable convolutions significantly reduces parameters and computational load;The accuracy rate of the improved YOLOv4 algorithm reached 98.54%,the average accuracy reached 97.42%,the recall rate reached 96.18%,and the average recognition time of a single image was 17 ms.The improved YOLOv4 algorithm has achieved rapid and accurate identification of sugarcane seed sprouts,meeting the seed selection needs of sugarcane cutting machines.(3)Study and determine the mathematical model of sugarcane cutting location.After sugarcane bud identification and classification,the specific pixel coordinates of sugarcane bud prediction box are output.Based on the conversion relationship between world coordinates and pixel coordinates,a mathematical model of sugarcane seed cutting location is determined,and the world coordinates corresponding to the pixel coordinates of sugarcane bud prediction box are obtained,with a positioning accuracy of over 96%.Analyze the causes of sugarcane bud location errors,calculate the actual distance between the cutting tool and the sugarcane bud to determine the real-time cutting position,and control the cutting tool to cooperate with the sugarcane transportation speed to achieve avoiding bud cutting.(4)Design a sugarcane seed cutting and seed selection device that automatically identifies and removes bad buds from sugarcane seeds.By using a visual module to identify the bad buds of sugarcane seeds,the sugarcane conveying device transports the entire cane for avoiding buds and cutting seeds.Conduct multi-factor experiments on seed cutting tools,determine the three optimal parameter combinations that affect seed cutting through simulation experiments,analyze the fatigue life of seed cutting tools,and determine the most applicable set of influencing factors as follows: the distance between the incision and the support point is 177.944 mm,the tool landing speed is 1.569m/s,the tool installation offset angle is 5.923 °,the tool front angle is 10.899 °,and the tool back angle is 8.637 °.Ensure that the cutting edge is flat,the sugarcane buds are not damaged by the tool or the pressure of the transport roller during operation.