Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Convolution Neural Network Based Automatic Corn Kernel Qualification

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:  2018 ASABE Annual International Meeting  1801859.(doi:10.13031/aim.201801859)
Authors:   Chao Ni, Dongyi Wang, Maxwell Holmes, Robert Vinson, Yang Tao
Keywords:   Corn kernel, Automatic Qualification Machine, Machine vision, Detection, Deep Neural Network

Abstract. Corn qualification is an important and time-consuming task in biosystem area. The human-based inspection strategy needs to be updated gradually with the quick development of corn industry. In this paper, an automatic corn qualification machine is proposed. Compared to related research, our machine integrates several new designs in both hardware and software components. Firstly, a gravity-based dual-side cameras design expands the machine`s field of view to describe corn kernels more completely. Second, touching kernels are preprocessed by a new K-means clustering guided curvature method, which can improve the robustness of our machine. Most importantly, deep convolutional neural network, a novel image processing technique, is embedded into the system to evaluate corn kernels. Compared to some traditional image classifiers, deep convolutional neural network can mimic human ideas, and make decisions based on overall considerations about image features. In our test dataset, deep convolutional neural network can achieve 97.0% test accuracy, and outperforms than current researches. Our corn qualification machine promises a great commercial value in the market, and it is also a good and practical application of artificial intelligent in real world.

(Download PDF)    (Export to EndNotes)