Grain Counting Method based on Artificial Neural Network

In traditional grain counting methods, they often rely on manual visual or manual counting, which leads to subjectivity and instability of the results. Artificial neural networks were used to develop an automated and accurate grain counting method, reducing the subjectivity of grain counting and improving the stability of grain counting results. A large number of grain images were collected through high-definition camera equipment, and the collected grain images were preprocessed to improve image quality. The Residual Network (ResNet) model was used to extract feature information from grain images and output the number of grains through a fully connected layer. The experimental results showed that the grain counting method using the ResNet-50 model had an average accuracy of 98.8%. Therefore, the ResNet-50 model can effectively improve the accuracy of grain counting and provide a new method for grain counting.