A Novel Tran_NAS Method for the Identification of Fe- and Mg-Deficient Pear Leaves from N- and P-Deficient Pear Leaf Data

Abstract

Trace element deficiency diagnosis plays a critical role in pear cultivation. However, high-quality diagnostic models are challenging to investigate, making it difficult to collect samples. Therefore, this manuscript developed a novel transfer learning method, named Tran_NAS, with a finetuning neural network that uses a neural architecture search (NAS) to transfer learning from nitrogen (N) and phosphorus (P) to iron (Fe) and magnesium (Mg) to diagnose pear leaf element deficiencies. The best accuracy of the transferred NAS model is 89.12%, which is 11% more than that of the model without the transfer of trace element-deficient samples. Meanwhile, Tran_NAS also has better performance on source datasets after comparing with different proportions of training sets. Finally, this manuscript summarizes the transfer model coincident characteristics, including the methods of batch normalization (BN) and dropout layers, which make the model more generalizable. This manuscript applies a symmetric homogeneous feature-based transfer learning method on NAS that is designed explicitly for near-infrared (NIR) data collected from nutrient-deficient pear leaves. The novel transfer learning method would be more effective for the micro-NIR spectrum of the nondestructive diagnosis.

Simultaneous Rapid Detection of Multiple Physicochemical Properties of Jet Fuel Using Near-Infrared Spectroscopy

“…The mean R p 2 and RPD values of the model increased from 0.7940 and 1.3 to 0.8193 and 1.6, respectively, compared to those of the full-wavelength spectrum model. The improvements in R p 2 and RPD values for the 11 physicochemical property indicator models ranged from 0.11 to 9.87% and from 1.51 to 87.91%, respectively. These results indicate that using the UVE algorithm for feature band selection in full-wavelength spectra can effectively improve the decrease in model measurement accuracy caused by the introduction of redundant information, thereby enhancing the accuracy and reliability of the model.…”

“…Near-infrared (NIR) spectroscopic analysis can reflect the overtone and combination frequency absorption information of hydrogen-containing functional groups in compounds . By combining chemometric algorithms, a rapid detection model can be established between the spectra and target properties, thus enabling the qualitative or quantitative analysis of samples. − NIR spectroscopy is highly efficient, rapid, and nondestructive and can simultaneously determine multiple components. It has been widely applied in various industries, such as chemical engineering, agriculture, and medicine. − In recent years, researchers have applied NIR spectroscopy to the characterization and quality detection of jet fuels.…”