Design And Experiment Of Lint-free Cottonseed Sorting Device Based On Machine Vision

Cotton,a crucial economic crop in China,is also the pillar industry in the Xinjiang region.With the advancement of agricultural production levels and the development of modern agriculture,the demand for high-quality cottonseed selection has been growing.Cottonseeds are prone to damage and mold during processing,transportation,and storage,which reduces germination rates and impacts yield.Therefore,it is essential to screen for superior cottonseeds before planting.Traditional manual and mechanical sorting methods face issues of low efficiency and inconsistent quality.Consequently,the development of a precise cottonseed sorting technology and equipment is of utmost importance.In this study,we designed a lint-free cottonseed sorting device based on machine vision.Utilizing dual cameras to simultaneously capture images of falling cotton seeds and integrating the YOLOv5 classification model,the device can rapidly remove substandard cottonseeds while retaining the undamaged ones.The specific research content is as follows:(1)Research of the visual discrimination model for lint-free cottonseed appearance quality.This study used Xinlu Medium 49 cottonseeds as test samples,characterized by their brownish-ellipsoid shape,8-10 mm in length,and 4-6mm in width.We collected 300 images each of intact,damaged,and moldy seeds,performed background removal preprocessing on the images,and used image enhancement techniques to expand the samples to 3,600,preparing for model training.We extracted and analyzed the external features of the cottonseeds.We trained cottonseed images using object classification models such as VGG and Mobile Net,finding that Dense Net yielded the best classification results with an accuracy of 95.1%.To explore better training outcomes,we further established the YOLOv5 s model,and after training for 2,500 iterations,the average precision reached 99.16%,the accuracy was 98.92%,and the recall rate was 99.24%.(2)Lint-free cottonseed sorting device design.The sorting device mainly includes a feeding structure unit,an image acquisition unit,and a cottonseed sorting unit.The overall framework of the hardware device is designed,and key structural unit designs and hardware selections are determined to verify the feasibility of the device’s overall operation.In this setup,the feeding mechanism uses a stepper motor to drive the turntable for single-grain cottonseed feeding.The image acquisition unit employs two industrial cameras to capture images of falling cottonseeds and uses the YOLOv5 s model to discriminate the seeds.The sorting mechanism uses a stepper motor and baffle to eliminate inferior seeds.(3)Research on the control system of the lint-free cottonseed sorting device.The control system primarily uses the stm32 microcontroller control board as the main controller,which controls the feeding module,sorting mechanism,and communication with the upper computer.Additionally,an improved inter-frame difference method is used in conjunction with a soft trigger method to capture images of falling cottonseeds.A humanmachine interaction interface is designed for monitoring and controlling cottonseed sorting.(4)Testing and experimentation of the lint-free cottonseed sorting device.Experiments are conducted to evaluate the performance of the feeding device,camera acquisition,and model detection.After testing,the overall sorting performance of the device is stable,with an average cottonseed sorting accuracy of 92.91% and a sorting efficiency of 388 seeds/min.