|
|
Abstract To improve the automation level of corn sorting and solve the problems of the complicated process of feature modeling by traditional methods and the high requirement of deployment of convolution neural network, the corn quality sorting and grading device was designed based on electromagnetic vibration and convolution neural network with several parts of corn kernel community feeding unit, electromagnetic feeding unit, control unit, sorting and collecting unit and constant light intensity visual single. The automatic separation of corn kernels, automatic recognition and sorting of corn kernels could be realized in the designed device. The results of model and prototype tests show that the model size is only 5.83 MB, which requires low computer hardware. The model mAP is 88.03%, and the overall classification and detection performance of the model is good. The model has strong recognition ability for excellent corn kernels, and P, R, FPR and F1 values are 98.75%, 94.84%, 3.78% and 96.85%, respectively. The actual detection accuracy of corn kernel is increased to 96.50% by the prototype, and the actual effective sorting rate is 97.51%.
|
|
|
|
|
|
|
|
