unripe, mid-ripe and ripe strawberries. Wendel et al. (2018)
applied a hyperspectral camera and navigation system
mounted to a ground vehicle to estimate and map mango
maturity in an orchard, and promising results were determi-
nation coefficient of cross-validation (R
2
cv
) of 0.64 and root
mean square error of cross-validation (RMSECV) of 1.08 %w/w
for dry matter content prediction in fruit on tree. However, to
the best of our knowledge, few researches have been reported
respect to in-field and non-destructive determination of in-
ternal quality and ripeness stages of peaches using hyper-
spectral imaging.
Therefore, the objective of this work was to investigate the
feasibility of portable hyperspectral imager for in-field deter-
mining the internal quality and ripeness stages of Feicheng
peach combined with chemometrics. To this end, hyper-
spectral images were in-field captured from mid-ripe and ripe
Feicheng peaches, and regression models were established
and compared with respect to their monitoring performance
for quality attributes.
2. Materials and methods
2.1. Sample preparation
Field sampling was conducted at a commercial orchard in
Feicheng City, Shandong, China (36.2003
N, 117.0876
E) on
September 14th, 2019. Mid-ripe and ripe Feicheng peach on 25
trees were randomly selected for spectral acquisition by the
experienced growers, and tagged afterwards to ensured that a
specific internal quality measurement was related to a specific
fruit in the image (Fig. 1). Then Feicheng peaches were picked,
wrapped and placed in a cooler and transported to the labo-
ratory on the same day. In this study, 80 mid-ripe and 80 ripe
fruits without leaves barrier and skin defects were collected
for quality monitoring of Feicheng peach.
2.2. Hyperspectral image acquisition
Figure 2 showed in-field image acquisition and processing. A
portable hyperspectral imager (GaiaField-V10E, Jiangsu Dualix
Spectral Imaging Technology Co., Ltd, Wuxi, Jiangsu, China),
mounted on a tripod, was employed to capture the Feicheng
peach images with a size of 1394 1040 pixels. It had a
spectral range of 400e1000 nm, a spectral resolution of 2.8 nm
and a field of view of 22
. To obtain a clear and undistorted
image, exposure time was set to 15.2 ms by scanning the
whole rectangular Teflon plate (Jiangsu Dualix Spectral Im-
aging Technology Co., Ltd, Wuxi, Jiangsu, China) placed on the
tripod. Hyperspectral images were acquired in natural illu-
mination under cloudy conditions from 11:00 am and 2:00 pm,
and the distance was 1.5 m between the lens and the pe-
riphery of tree scenes.
The raw hyperspectral image I
r
was calibrated to reduce
the effects of the uneven illumination and dark current noise
on the image using the following equation:
I
c
¼
I
r
I
d
I
w
I
d
(1)
where, I
c
was the calibrated image; I
d
was the dark reference
image, obtained by covering the lens with an opaque cap; I
w
was the white reference image, acquired from scanning the
rectangular Teflon plate used above. It was mounted to the
tripod and placed in the same scenes with fruits. To correctly
acquire hyperspectral images, for some obscured peaches, the
leaves were push aside and fixed with clips to expose the fruit.
A64 64 pixel region of interest (ROI) was extracted manually
from each fruit using software ENVI5.1 (Environment for
Visualizing Images, Research Systems Inc., Boulder, CO, USA),
covering major part of scanned fruit surface. The raw average
reflectance was calculated from all the pixels in ROI.
2.3. Measurement of internal quality attributes
Following in-field image acquisition, SSC and firmness of
Feicheng peach were measured in the postharvest engineer-
ing laboratory at Shandong Agricultural University, Tai'an,
Shandong, China. Measurements were performed at upper,
middle and lower fruit parts after peeling 1 cm
2
of a skin.
Firmness was first measured by a firmness metre (GY-1,
Thorpe Ltd, Zhejiang, China). A 3.5 mm probe punctured into
the peach about 10 mm at the speed of 0.5 mm/s, and firmness
was averaged at the three measured sections. Then, a digital
refractometer (PAL-1, Atago Co, Tokyo, Japan) was used to
Fig. 1 e Mid-ripe (a) and ripe (b) Feicheng peach on trees.
biosystems engineering 212 (2021) 115e125 117