Title: Apple multi-index fusion classification based on improved LSTM

Authors: Shuhui Bi; Lisha Chen; Lei Wang; Xuehua Yan; Liyao Ma

Addresses: School of Electrical Engineering, University of Jinan, Jinan 250022, Shandong, China ' School of Electrical Engineering, University of Jinan, Jinan 250022, Shandong, China ' Shandong Guige Intelligent Technology Co., Ltd., Jinan 250000, Shandong China ' School of Electrical Engineering, University of Jinan, Jinan 250022, Shandong, China ' School of Electrical Engineering, University of Jinan, Jinan 250022, Shandong, China

Abstract: Apples non-destructive testing (NDT) based on near infrared (NIR) spectroscopy technology can predict the internal quality, but it mainly focuses on single internal index detection. In order to better reflect the internal quality of apple, multi-index fusion classification based on improved long and short-term memory (LSTM) neural network will be considered in this paper. Firstly, apples' soluble solids content (SSC) and moisture content (MC) are selected as the key indexes, which are the main factors affecting the quality and taste of apple, and the correlation analysis of spectra with apples' SSC and MC is carried out, and the correlation thermal maps are drawn. Secondly, multiple scattering correction (MSC) is used for pre-processing to correct the baseline translation and drift of spectral data. Moreover, sparse principal component analysis (SPCA) is combined with to reduce the spectral data dimension, enhancing the models predictive precision. Then, the improved particle swarm optimisation (IPSO) algorithm is employed to optimise the LSTM multi-index prediction model. Finally, the simulation results indicate that the SSC and MC classification accuracy by the proposed IPSO-LSTM model substantially surpasses that by LSTM model.

Keywords: near infrared spectroscopy technology; multi-index; long and short-term memory neural network; apple classification.

DOI: 10.1504/IJAMECHS.2025.144587

International Journal of Advanced Mechatronic Systems, 2025 Vol.12 No.1, pp.1 - 11

Received: 24 Jun 2024
Accepted: 29 Sep 2024
Published online: 23 Feb 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article