Design And Experiment Of High-throughput Automatic Analysis Equipment For Wheat Grains

As one of the important organs of wheat,wheat grain characters are closely related to the yield and quality of wheat,which is an important basis for variety selection,germplasm resources evaluation and wheat quality evaluation.As the key link of wheat breeding,wheat seed analysis needs to measure the parameters such as number of grains,grain length,grain width,length-width ratio,and quantity of 1000-grain weight.The efficiency and precision of seed test directly affect the breeding results of varieties.However,the current wheat grain analysis equipment has a low degree of automation,single function and can not achieve a large number of different varieties of wheat grains batch analysis,which seriously restricts the breeding efficiency.Aiming at the problems of low automation of the existing wheat grain analysis equipment,the inability to achieve batch wheat grain analysis of a large number of different varieties,the poor recognition ability of wheat grains,which lead to obtaining the poor accuracy of grain testing parameters,this study carried out the design and research of highthroughput automatic wheat grain analysis equipment for the purpose of obtaining wheat grain parameter indicators.In the research,the overall structure design,software design and system control design were carried out in combination with hardware equipment and threedimensional software.Through the deep learning network and image processing methods,the parameters such as wheat grain qualification,quantity,grain length,grain width,lengthwidth ratio and so on were accurately obtained,and the batch,full-automatic and highthroughput high-precision parameter acquisition process of different varieties of wheat grain was realized.The specific contents and main conclusions of this study are as follows:(1)Through the analysis of the requirements of wheat seed testing parameters,the existing problems and requirements of the equipment,Creo software was used to design the overall structure of the wheat grain analysis equipment and the components of the cap opening device,the cap closing device,the cap conveying device,the image acquisition platform,the weighing platform,the initial limit device,and the code scanning device,and create a three-dimensional model of the equipment.Design and study the equipment operation process,determine the equipment work execution process under the two working modes of automatic and manual a wheat grain nalysis,and meet the requirements of batch and automatic grain analysis for a large number of different varieties of wheat.(2)From the perspective of overall control,through the analysis of hardware requirements,the selection of control hardware and executive components is carried out according to the needs of the system,and the overall control scheme and the control methods of stepping motor,DC motor,steering gear,flexible vibration plate,electronic balance,code scanner and other components are designed and studied,and the control system is built.According to the system seed examination process,the control strategy is designed,the logic control program under the automatic and manual seed examination mode is written,and the system software interface is written based on the Py Qt5 software,which provides the user operation and the real-time display visual interface of the seed examination parameters,so as to realize the stable,fast and efficient automatic operation of the high-throughput automatic seed examination equipment for wheat seeds.(3)Aiming at the problem of poor accuracy in obtaining parameter indexes and achieving grain counting due to the poor ability of recognition of unqualified wheat grains and impurities in the common wheat grain analysis equipment,this study establishes a normal,broken and moldy wheat grain detection model through the Yolov5 s deep learning network to achieve accurate identification and counting of wheat grains.After 300 iterations of training,the model training location loss and classification loss are stable at about 0.021 and 0.001,respectively,the precision is 98.6%,the recall is 98.4%,and the mean average precision(m AP)is 99.2%,indicating that the model has good detection effect.According to the results of the grain detection model,obtain the position information of normal wheat grains in the image.After intercepting the normal wheat grain images,perform graying,Gaussian filtering,and Canny edge detection on it,establish the minimum outer rectangle of wheat grains,and calculate the grain length,grain width,and length-width ratio parameters of wheat grains based on the rectangular length and width,to achieve rapid and accurate acquisition of wheat grain parameters based on image information.(4)Through the system operation test,the wheat grain analysis efficiency test,the counting accuracy test and the precision test of wheat grain parameter acquisition were carried out.The results obtained were as follows: under the automatic and manual seed testing mode,the analysis time for a wheat grain variety was 2:45 to 3:05 seconds and 40 seconds respectively;The accuracy rate of grain counting is more than 99%,and the R~2is0.99,RMSE is 0.68,MAPE is.13%;The measured error of average particle length and particle width is within 0.03 mm,and the error of average aspect ratio is within 0.2.In terms of R~2、MAPE and RMSE,the results of the average grain length correlation curve were0.9,2.7% and 0.22 respectively,the results of the average grain width correlation curve were 0.8,4.5% and 0.18 respectively,and the results of the average length-width ratio correlation curve were 0.6,3.4% and 0.08 respectively.The above test results show that the high-throughput automatic wheat grain analysis equipment designed by the research institute can obtain a variety of high-precision seed testing parameters efficiently,quickly and accurately,and apply it to the wheat seed testing work.