Firmness Detection for Apples Based on Hyperspectral Imaging Technology Combined with Stack Autoencoder-Extreme Learning Machine Method

Libo Rao; Tao Pang; Ranshi Ji; Xiaoyan Chen; Jie Zhang

doi:10.3788/LOP56.113001

Laser & Optoelectronics Progress, Volume. 56, Issue 11, 113001(2019)

Firmness Detection for Apples Based on Hyperspectral Imaging Technology Combined with Stack Autoencoder-Extreme Learning Machine Method

Libo Rao¹, Tao Pang¹, Ranshi Ji¹, Xiaoyan Chen^2,3、*, and Jie Zhang²

Author Affiliations

¹ College of Mechanical and Electrical Engineering, Sichuan Agricultural University, Yaan, Sichuan 625014, China

² College of Information Engineering, Sichuan Agricultural University, Yaan, Sichuan 625014, China

³ Sichuan Provincial Key Laboratory of Agricultural Information Engineering, Sichuan Agricultural University, Yaan, Sichuan 625014, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(8)

Fig. 1. Structural diagram of hyperspectral imaging system

Download full size

Fig. 2. Original spectrum of pixel points

Download full size

Fig. 3. Basic network structure of ELM

Download full size

Fig. 4. Network structures of basic auto-encoder and SAE-ELM. (a) Basic auto-encoder; (b) SAE-ELM

Download full size

Fig. 5. Flow chart of training in SAE-ELM

Download full size

Fig. 6. Scatter plots of prediction and measurements for samples in calibration set and prediction set.(a) SAE-ELM (18); (b) ELM

Download full size

Table 1. Statistics of apple firmness in calibration set and prediction set
View table
Table 1. Statistics of apple firmness in calibration set and prediction set
Set Number ofsamples Min /N Max /N Mean /N Bias
Total 126 28.975 55.900 36.438 4.147
Calibration 90 28.975 55.900 36.720 4.579
Prediction 36 30.800 44.500 35.733 2.717

Table 2. Calibration and prediction results of apple firmness using ELM and SAE-ELM

View table

Table 2. Calibration and prediction results of apple firmness using ELM and SAE-ELM

Index	Index	Calibration		Prediction
Index	Index	$\begin{matrix} R_{C}^{2} \end{matrix}$	R_RMSEC	$\begin{matrix} R_{P}^{2} \end{matrix}$	R_RMSEP	R_RPD
ELM	Full spectral	0.8442	1.7998	0.7099	1.4427	1.883
ELM	Optimal spectral	0.8940	1.4445	0.7345	1.6297	1.968
SAE-ELM(15)	Full spectral	0.8347	1.8517	0.7335	1.3828	1.965
SAE-ELM(16)	Full spectral	0.8285	1.8860	0.6866	1.4994	1.812
SAE-ELM(17)	Full spectral	0.8075	1.9984	0.6252	1.6398	1.657
SAE-ELM (18)	Full spectral	0.8399	1.8222	0.7703	1.2837	2.116
SAE-ELM (19)	Full spectral	0.8633	1.6836	0.7097	1.4432	1.882
SAE-ELM (20)	Full spectral	0.8362	1.8431	0.6601	1.5618	1.740
SAE-ELM(21)	Full spectral	0.8674	1.6284	0.7440	1.3553	2.004
SAE-ELM(22)	Full spectral	0.8538	1.7412	0.7225	1.4111	1.925

Tools

Get Citation

Copy Citation Text

Libo Rao, Tao Pang, Ranshi Ji, Xiaoyan Chen, Jie Zhang. Firmness Detection for Apples Based on Hyperspectral Imaging Technology Combined with Stack Autoencoder-Extreme Learning Machine Method[J]. Laser & Optoelectronics Progress, 2019, 56(11): 113001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites