1. Introduction

Non-destructive food testing plays an important role in food sieving and to guarantee high quality of food production. Most of the existing X-ray sieving machines^[1] for vegetables and fruits use high X-ray absorption contrast to distinguish the foods and foreign matters, like stones and metals. However, X-ray sieving machines cannot function if foreign matters and foods have similar X-ray attenuation^[2]. Owing to high evolution in X-ray imaging techniques during the past decade, as an X-ray grating-based imaging technique^[3-8], Talbot and Talbot-Lau interferometry offers refraction and scattering signals to provide useful information on internal structures for understanding characteristics of the samples, becoming an important technique for food science^[9]. Furthermore, with the use of source grating, this method can be operated with conventional X-ray source, instead of relying on brilliant synchrotron X-ray source. Thus, it is quite suitable for medical imaging and industrial non-destructive testing^[10-15].

In this paper, we use a Talbot-Lau interferometer to obtain absorption, refraction and scattering contrast images of a fresh cherry tomato and a dry umeboshi. As for some other non-destructive methods, like VIS/NIR spectroscopy^[16] and hyperspectral imaging^[17], they mainly focused on fruit quality assessment using surface layer information but internal structure information of the fruits. Using penetration power of X-rays, the three complementary images provide internal structures information of the samples, hence the capability of this method for non-destructive fruit testing.

2. Experiment and Methods

Experiments were carried out at the Institute of Multidisciplinary Research for Advanced Materials, Tohoku University, Japan, using a rotating anode X-ray tube in Talbot-Lau interferometer configuration, which is shown schematically in Fig. 1 ^[18]. A micro-focus X-ray generator with source grating (G0) is used for partially coherent illumination, in its acceleration voltage of 40 kV, the filament current of 45 mA, and the X-ray spot size of 0.3 mm ×0.3 mm. The G0 grating was in 22.7-μm periods, the π/2 phase shift grating (G1) was in 4.36-μm periods, and analyzer grating (G2) was in 5.4-μm periods. For the Talbot interferometer ^[19] the G0–G1 and G1–G2 distances were 106.9 cm and 25.6 cm, respectively. The gratings were optimized at the design photon energy of 27 keV. The projection images were recorded by a CCD detector (Special Instrument Inc.) with an effective receiving area of 68.4 mm ×68.4 mm in pixel size of 18 μm×18 μm.

Fig.1.Full-screen View

Schematics of the X-ray grating interferometer (Talbot-Lau interferometer). A micro-focus X-ray generator is used with source grating (G0). G1 is for phase grating and G2 is for amplitude grating.

A fresh cherry tomato and a dried umeboshi were placed in a sample stage at upstream side of the phase grating G1. The measurements were carried out with air as a reference. Phase stepping^[4] was used in this setup. The absorption grating G2 was scanned along the transverse direction X (Fig. 1) over one period of the grating. Five phase steps were used in both sample and reference measurements. At each phase step an image was taken with an exposure time of 10 seconds. For the two samples, the experimental parameters were the same, at similar temperature and air humidity.

The shifting curve visibility of V=29.9% was calculated using the mean intensity in the five steps in Fig.2(A). A Moire fringe always formed as two reasons, one is the superposition between the self-image of G1 and G2 pattern in the plane of the X-ray detector, and another is the mechanical errors of the gratings. The Moire fringe in Fig.2(B) tend to be caused by mechanical errors of the gratings, this problem can be solved through background subtraction in the data post-processing. The absorption A(x,y), refraction θ(x,y) and scattering V(x,y) signals of the sample were retrieved by Eqs.(1)–(3)^[20]:

Fig.2.Full-screen View

Matched phase step curve (A) with mean intensity value (the red dots) in five steps, and intensity image of Step 1 (B).

where, k is the number of steps during the phase stepping scan in one period of G2; I_k^s(x,y) and I_k^b(x,y) are the gray value of pixels at each step of the scan with and without the specimen, respectively; d₂ is the period of grating G2; and Z is the distance between grating G1 and G2.

3. Results and Discussion

Figure 3 shows the photograph and grating-based X-ray images of the fresh cherry tomato. Structures of the cherry tomato, from the outer to the inner, are epicarp, mesocarp, endocarp and seeds. Epicarp is a botanical term for the outermost layer of the fruit, endocarp is an interior layer which directly surrounds the seeds and mesocarp is the fleshy middle layer between epicarp and endocarp. In the absorption image (Fig.3B), along with the indistinct pedicel, the steady thickness increase from epicarp to the sample inner replies to the increase of X-ray absorption, with the gray values taking on degressive tendency. In the refraction image (Fig.3C), the outer part of fruit stem, epicarp and some of the inner seeds can be distinguished, although the shape and the distribution of the seeds are not clear. Remarkably, not only the outer part of fruit stem but also inner part of stem can be observed clearly. Thanks to the large difference of gray value between mesocarp/endocarp and seeds, the size, position and direction of each seed can be distinguished, and the overlap place among seeds help to understand stratification distribution of seeds in the scattering image (Fig.3D).

Fig.3.Full-screen View

A fresh cherry to (A) and its X-ray images of absorption (B), refraction (C) and scattering (D). Scale bar: 3 mm.

Figure 4 provides the photograph and grating-based X-ray images of the umeboshi, one kind of dried fruit. Structures of umeboshi, from the outer to the inner, are epicarp, mesocarp, endocarp and seed. The boundary of endocarp and seed can be distinguished from the sample in absorption image (Fig.4B), because the lignified endocarp is of lower density than mesocarp and the endocarp-seed interspace produces a low gray value. By contrast, wrinkles on epicarpand section thickness of endocarp can be observed clearly in Fig.4C for the refraction contrast. It is worth noting that in the scattering image (Fig.4D) the hard lignified endocarp becomes fuzzy as low scattering signals, which exaggerate the seed obviously. Also, scattering signals from the endocarp are a little higher than those from the mesocarp, and this helps to find its position. It is exciting to find that the refraction image reveals details of the lignified endocarp while the scattering image highlights characteristics of the seed, including the shape, size and some other details. Therefore, the three X-ray images of different contrast provide complementary structure information of the dried umeboshi.

Fig.4.Full-screen View

A dried umeboshi (A) and its X-ray images of absorption (B), refraction (C) and scattering (D). Scale bar: 2 mm.

The traditional X-ray non-destructive imaging can only obtain the absorption information, which is strongly related to the sample thickness and density. The absorption value is larger with a thicker sample, and the inner signal is drown in the whole signal group if the internal structures are of similar density. That is why seed structures of the cherry tomato cannot be seen in Fig.3B, but the umeboshi endocarp and seed with density difference can be observed in Fig.4B.

Using the grating-based X-ray imaging, refraction and scattering images can be retrieved from the same set of projection data. They reproduce abundant information of surface and internal structures (Figs.3C, 3D, 4C and 4D). The refraction image is sensitive to the density and refraction angles. This helps to reveal the seeds in the fresh cherry tomato and wrinkles on epicarp and the details of lignified endocarp in the dried umeboshi. The scattering signal is sensitive to the density and the reflection at the internal or external interfaces^[21], which shows the seeds clearly in both fresh and dried fruits. Thus, all the internal structures that reflect the quality of fruits (fresh or dry) can be retrieved distinctly. It demonstrates that the grating-based X-ray imaging has high potential for non-destructive food testing.

Furthermore, the contrast-to-noise ratio (CNR)^[22] can quantify the contrast of the three kinds of images as follows:

where μ is the mean value and σ² is the variance of selected region of interest (ROI) in the absorption, refraction and scattering images. The variances are weighted with the factor ω which is given by the ratio of the number of pixels in the matrix. The subscripts 1 and 2 mean the area of boxes by the white dotted lines and the yellow lines, respectively, as given in Figs. 3 and 4. A high CNR value means a high contrast to noise, whereas a low value represents a low contrast. As shown in Table 1, the scattering images of the fresh cherry tomato and dried umeboshi have the largest CNR value, hence the highest contrast of the three kinds of images. Therefore, scattering imaging helps to distinguish different structures and materials more clearly than absorption and refraction imaging.

4. Conclusion

In this work, the grating-based X-ray imaging by Talbot-Lau interferometer was used to investigate the internal structures of fresh and dried fruits non-destructively. Together with the traditional absorption image, the refraction and scattering images reproduce abundant information of the surface and interface. The refraction information is always clear at the boundary of different structures as the existence of refraction angle. The scattering signals are clearly sensitive to status of a fruit, with the highest contrast-to-noise ratio of the three kinds of images. The information in scattering images reflects the quality of fruits on the morphology, indicating the high potential of this imaging technique to conduct non-destructive testing on fruits. Moreover, the 3D tomography of fruit research can be operated when scattering images in a series of different angles are taken. However, grating-based X-ray imaging is time-consuming now. Further work on grating interferometry scanning system with a number of line detectors, and sample-maneuvering mechanism, will make this non-destructive technique to be a practical implementation in the near future.