Design And Experimental Analysison Device Of Corn Seeds Double-sided Damage Detection Based On Machine Vision

Corn seeds will be damaged in the process of growth,harvesting,threshing,storage and transportation due to growth defects,mechanical gnawing and other factors,which will affect the germination rate and the yield of corn.Therefore,pre-sowing detection of maize seeds is very important.Aiming at the problems of low mechanization,low detection accuracy and only one-sided detection in the process of corn seeds detection,this paper applies machine vision technology to the appearance quality detection of corn seeds,proposes a corn seeds damage detection method based on machine vision,and designs a corn seeds double-sided damage detection device.In this paper,a machine vision-based detection method for corn seeds damage is proposed,which takes "Denghai 605" summer maize seed as the research object.After obtaining the corn seeds images,the image noise is analyzed by subtraction method,and the impulse noise is identified as the noise type.The median filtering method is used to denoise the images,and threshold segmentation method and morphological processing are used to segment the corn seeds and background in the image after denoising.According to the morphological characteristics of corn seeds,6 geometric parameters,including perimeter,area,perimeter-area ratio,long-axis length,short-axis length and length-width ratio,3 shape characteristic parameters,rectangularity,roundness and compactness,and Hu moment invariant characteristic parameters,were extracted respectively.Based on Support Vector Machine(SVM),a corn seeds damage detection model was constructed.Sixteen eigenvectors were input,and the optimal values of penalty factor and kernel function parameters were determined by grid search and cross-validation.The recognition model was tested,and the test results showed that the recognition accuracy of the SVM recognition model for corn seed damage could reach 95%.The overall design scheme of double-sided damage detection device for corn seeds was determined,and the key mechanisms such as feeding mechanism,conveying mechanism,removal and recovery mechanism were designed.The dimension parameters and installation position of key components were determined through experimental analysis and theoretical calculation.In order to determine the installation angle of seed guide tube and the opening angle of bottom lids,the sliding friction angle of corn seeds with different moisture content on acrylic material was studied.It was found that the higher the moisture content of corn seeds,the bigger the sliding friction angle.When the maximum moisture content is 12.8%,the sliding friction angle is 26.8 degrees.In order to analyze the opening conditions of the bottom lids,the maximum attraction and minimum attraction distance of NdFeB wafers were tested.The image acquisition and processing system is designed,the double-sided image acquisition scheme is determined,and the power control system of the device is designed based on Raspberry Pi and Arduino Uno.The performance indexes of the device were tested,and the experimental results show that the accuracy of system detection is 98.4%,the classification accuracy of device is 95.2%,the metering accuracy of metering device is 97.7%,the feeding rate of seed is 98.1%,the rejection accuracy is 99.2%,the opening accuracy of bottom lids is 100%,and the closing accuracy of bottom cover is 99.7%.Among them,the accuracy of device classification is greatly affected by the accuracy of metering device and seed feeding rate.This research realizes the double-sided detection of corn seeds,and provides a reference for crop seed detection based on machine vision technology.