In this work, we propose a metamaterial absorber at microwave frequencies with significant sensitivity and non-destructive sensing capability for grain samples. This absorber is composed of cross-resonators periodically arranged on an ultrathin substrate, a sensing layer filled with grain samples, and a metal ground. The cross-resonator array is fabricated using the printed circuit board process on an FR-4 board. The performance of the proposed metamaterial is demonstrated with both full-wave simulation and measurement results, and the working mechanism is revealed through multi-reflection interference theory. It can serve as a non-contact sensor for food quality control such as adulteration, variety, etc. by detecting shifts in the resonant frequencies. As a direct application, it is shown that the resonant frequency displays a significant blue shift from 7.11 GHz to 7.52 GHz when the mass fraction of stale rice in the mixture of fresh and stale rice is changed from 0% to 100%. In addition, the absorber shows a distinct difference in the resonant absorption frequency for different varieties of grain, which also makes it a candidate for a grain classification sensor. The presented scheme could open up opportunities for microwave metamaterial absorbers to be applied as efficient sensors in the non-destructive evaluation of agricultural and food product quality.

In this paper, the grain loss assessment was studied based on logistic regression, and 5400 samples of 31 provinces in our country in the year 2012-2014 were selected, and the 7 typical provinces among them were respectively tested. Using the logistic regression model to predict the loss rate of grain harvesting step, the prediction result is 86.25%. The stochastic gradient descent algorithm is used to optimize the parameters of the model, when the learning_rate is 0.1, the prediction results of grain harvest losses of up to 92.53%, which further improves the prediction accuracy of grain harvest step loss.

Prediction of protein subcellular location has currently become a hot topic because it has been proven to be useful for understanding both the disease mechanisms and novel drug design. With the rapid development of automated microscopic imaging technology in recent years, classification methods of bioimage-based protein subcellular location have attracted considerable attention for images can describe the protein distribution intuitively and in detail. In the current study, a prediction method of protein subcellular location was proposed based on multi-view image features that are extracted from three different views, including the four texture features of the original image, the global and local features of the protein extracted from the protein channel images after color segmentation, and the global features of DNA extracted from the DNA channel image. Finally, the extracted features were combined together to improve the performance of subcellular localization prediction. From the performance comparison of different combination features under the same classifier, the best ensemble features could be obtained. In this work, a classifier based on Stacked Auto-encoders and the random forest was also put forward. To improve the prediction results, the deep network was combined with the traditional statistical classification methods. Stringent cross-validation and independent validation tests on the benchmark dataset demonstrated the efficacy of the proposed method.