Neural network application for thermal image recognition of low-resolution objects

Applications of neural networks in human shape visual perception

Advances in optical and electronic technology can immensely reduce noise in images and greatly enhance human visual recognition. However, it is still difficult for human eyes to identify low-resolution thermal images, due to the limits imposed by psychological and physiological factors. In addition, changes in monitor brightness and lens resolution may also interfere with visual recognition abilities. To overcome these limitations, we devised a suitable and effective recognition method which may help the military in revising the shape parameters of long-range targets. The modulation transfer function was used as a basis to extend the visual characteristics of the human visual model and a new model was produced through the incorporation of new shape parameters. The new human visual model was next used in combination with a backpropagation neural network for better recognition of low-resolution thermal images. The new model was then tested in experiments and the results showed that the accuracy rate of recognition steadily rose by over 95%.

Application of artificial neural network (ANN) and partial least-squares regression (PLSR) to predict the changes of anthocyanins, ascorbic acid, Total phenols, flavonoids, and antioxidant activity during storage of red bayberry juice based on fractal analysis and red, green, and blue (RGB) intensity values

Artificial neural network (ANN) and partial least-squares regression (PLSR) models were developed to predict the changes of anthocyanin (AC), ascorbic acid (AA), total phenols (TP), total flavonoid (TF), and DPPH radical scavenging activity (SA) in bayberry juice during storage based on fractal analysis (FA) and red, green, and blue (RGB) intensity values. The results show the root mean squared error (RMSE) of ANN-FA decreased 2.44 and 12.45% for AC (RMSE = 18.673 mg/100 mL, R(2) = 0.939) and AA (RMSE = 8.694 mg/100 mL, R(2) = 0.935) compared with PLSR-RGB, respectively. In addition, PLSR-FA (RMSE = 5.966%, R(2) = 0.958) showed a 12.01% decrease in the RMSE compared with PLSR-RGB for predicting SA. For the prediction of TP and TF, however, both models showed poor performances based on FA and RGB. Therefore, ANN and PLSR combined with FA may be a potential method for quality evaluation of bayberry juice during processing, storage, and distribution, but the selection of the most adequate model is of great importance to predict different nutritional components.

Using neural networks to estimate the losses of ascorbic acid, total phenols, flavonoid, and antioxidant activity in asparagus during thermal treatments

Artificial neural networks (ANNs) with back-propagation algorithm were developed to predict the percentage loss of ascorbic acid, total phenols, flavonoid, and antioxidant activity in different segments of asparagus during water blanching at temperatures ranging from 65 to 95 degrees C as a function of blanching time and temperature. In this study, the one-hidden-layer ANNs are used, and the number of neurons in the hidden layer were chosen by trial and error. Optimized ANN models were developed for predicting nutrient losses in bud, upper, middle, and butt segments of asparagus. ANN models were then tested against an independent data set. Our results showed that the predicted values of the correlation coefficients between experimental and ANNs ranged from 0.8166 to 0.9868. Therefore, ANNs could be potential tools for the prediction of nutrient losses in vegetables during thermal treatments.