A CCD based machine vision system for real-time text detection

Shihua ZHAO, Lipeng SUN, Gang LI, Yun LIU, Binbing LIU

Front. Optoelectron. ›› 2020, Vol. 13 ›› Issue (4) : 418-424.

PDF(1725 KB)
Front. Optoelectron. All Journals
PDF(1725 KB)
Front. Optoelectron. ›› 2020, Vol. 13 ›› Issue (4) : 418-424. DOI: 10.1007/s12200-019-0854-0
RESEARCH ARTICLE
RESEARCH ARTICLE

A CCD based machine vision system for real-time text detection

Author information +
History +

Abstract

Text detection and recognition is a hot topic in computer vision, which is considered to be the further development of the traditional optical character recognition (OCR) technology. With the rapid development of machine vision system and the wide application of deep learning algorithms, text recognition has achieved excellent performance. In contrast, detecting text block from complex natural scenes is still a challenging task. At present, many advanced natural scene text detection algorithms have been proposed, but most of them run slow due to the complexity of the detection pipeline and cannot be applied to industrial scenes. In this paper, we proposed a CCD based machine vision system for real-time text detection in invoice images. In this system, we applied optimizations from several aspects including the optical system, the hardware architecture, and the deep learning algorithm to improve the speed performance of the machine vision system. The experimental data confirms that the optimization methods can significantly improve the running speed of the machine vision system and make it meeting the real-time text detection requirements in industrial scenarios.

Keywords

machine vision / text detection / optical character recognition (OCR) / deep learning

Cite this article

Download citation ▾
Shihua ZHAO, Lipeng SUN, Gang LI, Yun LIU, Binbing LIU. A CCD based machine vision system for real-time text detection. Front. Optoelectron., 2020, 13(4): 418‒424 https://doi.org/10.1007/s12200-019-0854-0

1 Introduction

As an important II-VI semiconductor with wide band gap, ZnS has been attracting extensive attention and has been applied in photocatalysis, optoelectronic devices, and sensors due to its unique optical and electrical properties [1-3]. Various methods had been used to synthesize nanostructural ZnS. For example, ZnS nanoporous nanoparticles were prepared by solution-phase thermal decomposition route using zinc acetate and thiourea as Zn and S sources, respectively [4]. ZnS nanotubes were self-assembled via a thermochemistry process using ZnS powers as precursors [5]. Owing to the fast generation of electron-hole pairs in photocatalysis reaction, ZnS nanocrystals have been used as photocatalyst in photodegadation of organic pollutions [4]. However, a problem is that the ZnS nanocrystals incline to agglomerate in solution and then affect the photocatalysis effection. Therefore, TiO2, CuS and carbon nanotubes had been used as a support to enhance the photocatalytic activity [6-8]. Recently, we have prepared carboxyl modified amorphous carbon nanotubes (ACNTs) by a simple solvothermal method. Thanks to its large surface area, ACNTs were a good candidate for the support to keep the nanoparticles from agglomeration.
Due to the efficient energy of photons provided by the light, the photochemical route is a convenient, mild and environmental friendly method for synthesis of metal and sulfide nanomaterials [9-11]. However, the previous work using photochemical method were focused on preparing nanoparticles, while the synthesis of nanocomposite through photochemical reaction is scarcely reported [12]. Herein, we use carboxyl modified ACNTs as a support and the ultra-violet (UV) lamp as light source to synthesize ACNTs/ZnS nanocomposites. The as-synthesized ACNTs/ZnS nanocomposites exhibit excellent photocatalytic activity to dye molecule under UV light.

2 Materials and method

2.1 Materials

Ferrocenecarboxylic acid, carbon disulfide (CS2), toluene, carbon tetrachloride, Zn(Ac)2·2H2O, thioacetamide (TAA), eosin, methyl blue (MB) and methyl red (MR) are all analytic-grade chemical reagents (Shanghai Chemical Co.) and used without purification.

2.2 Synthesis of ACNTs

In a typical procedure, 0.23 g ferrocenecarboxylic acid was dissolved in 34 mL of toluene, and then, 0.2 mL of carbon tetrachloride and 6 mL of carbon disulfide (CS2) were added into the above solution. The solution was sealed in a Teflon-lined stainless steel autoclave and maintained at 200°C for 12 h. The sponge-like black products were collected and washed with alcohol and distilled water.

2.3 Synthesis of ACNTs/ZnS nanocomposites via photochemical reaction

In the Synthesis of ACNTs/ZnS nanocomposites, 2 mg of ACNTs was dispersed in 15 mL of distilled water with strong ultrasonic. Then, 0.2 mmol of Zn(Ac)2·2H2O and 0.2 mmol of TAA were added to the above solution to form a homogeneous solution by continuous stirring. To obtain the ACNTs/ZnS nanocomposites, we exposed the mixed solution under a 300 W high-pressure mercury lamp (wavelength centered at 365 nm) for 1 h. Magnetic stirring was applied throughout the entire synthesis. The products (named sample D) were filtrated and washed by alcohol and distilled water.
We found that the size and distribution of ZnS nanoparticles which were deposited on ACNTs changed with different experimental parameters. To study the effect of experimental parameters, we experience a series of contrastive experiments by altering the initial amount reactants and reaction time. In the above contrastive experiments, the amounts of ACNTs were kept as 2 mg. The detailed experimental parameters are listed in Table 1.
Tab.1 Summary of experimental parameters
sampleABCDEF
Zn (Ac)2·2H2O/mmol0.80.60.40.20.20.2
TAA/mmol0.80.60.40.20.20.2
reaction time1 h1 h1 h1 h30 min10 min

2.4 Photocatalytic testing

The photocatalytic activity of the ACNTs/ZnS nanocomposites was evaluated by degrading eosin, MB and MR aqueous solution (50 mL) with 10 mg as-prepared ACNTs/ZnS nanocomposites under UV irradiation at ambient temperature. Experimental details were as follows: 10 mg of the ACNTs/ZnS nanocomposites was dispersed in a 50 mL of the dye solution (0.01 g/L) in a 100 mL beaker. The solution was stirred in dark for 6 h and then irradiated for 80 min with a 300 W UV lamp (Yaming Lighting Co., Ltd., 10 cm above the solution) with wavelengths centered at 365 nm. 5 mL solution was taken out of the system at 10 min time interval and centrifuged at 9000 min for 3 min. The concentration of the dye was determined with a Shimadzu U-3010 UV-vis spectrophotometer.

2.5 Characterization

The X-ray powder diffraction patterns were obtained by a Shimadzu X-ray diffractometer (XRD, Shimadzu, XRD-6000). Scanning electron microscopy (SEM, Hitachi, S-4800), high resolution transmission electron microscopy (HRTEM, JEOL-2010) and UV-vis absorption spectroscopy (Shimadzu, U-3010) were used to measure the morphology, structure, size and optical response of the product. The Raman spectrum was recorded at ambient temperature on a LABRAM-HR confocal laser micro-raman spectrometer with an argon-ion laser at an excitation wavelength of 514.5 nm. The infrared spectrum was collected on a Shimadzu IR prestige-21 Fourier transform-infrared (FT-IR) spectrum in the wavelength range from 400 cm-1 to 4000 cm-1. Photoluminescence (PL) spectrum of the as-obtained samples was measured on an Edinburgh 920 fluorescence spectrophotometer using a 325 nm excitation line at room temperature.

3 Results and discussion

SEM and TEM were used to determine the morphology and microstructure of the ACNTs. Figure 1(a) is the SEM image of the as-prepared ACNTs. It reveals that we obtained a large quantity of ACNTs with uniform size. The TEM image of ACNTs, the dark wall and light inner of products, is shown in Fig. 1(b), giving the direct evidence of tubular structure. From the TEM image, we observed that the diameter of carbon nanotubes was about 100 nm and the wall thickness was about 25 nm. It should be noted that as-obtained carbon nanotubes had smooth surface. HRTEM images of a small area of nanotube were shown in Fig. 1(c), indicating that the CNTs were amorphous.
Fig.1 Images of ACNTs. (a) SEM; (b) TEM; (c) HRTEM

Full size|PPT slide

Figure 2(a) shows the XRD pattern of the ACNTs, only a broad and weak peak at around 20°, indicating that the as-prepared carbon nanotubes is amorphous. Two peaks are observed in Raman patterns. One peak is observed at around 1586 cm-1 named G-band, which is related to the vibration of sp2-bonded carbon atoms in the hexagonal graphitic plane. The other peak is at 1370 cm-1 named D-band, which corresponds to the vibrations of carbon atoms with dangling bonds in disorder graphite planes [13,14]. Figure 2(c) shows the IR patterns of the ACNTs. The weak peak at 1606 cm-1 is the C=O stretching vibrations of carbonyl groups; the peak at around 1702 cm-1 is assigned to the -COOH stretch; the peak at around 2941 cm-1 is the O-H stretching vibration of carbonyl groups [15]. The peak at 2350 cm-1 indicates that CO2 is adsorbed on the surface of ACNTs. The present results indicate that as-obtained ACNTs were modified with carboxylic groups during preparation without further treating.
Fig.2 Patterns of ACNTs. (a) XRD; (b) Raman; (c) IR

Full size|PPT slide

Figure 3(a) is the XRD pattern of ACNTs/ZnS nanocomposites. Three apparent peaks correspond respectively to the crystal planes of the (111), (220), and (311) of face-centered cubic ZnS, consistenting with the standard diffraction date (JCPDS Card File No. 80-0020). The peaks are broad, indicating that the size of ZnS nanoparticles is small. The SEM image of ACNTs/ZnS nanocomposites is shown in Fig. 3(b), in which the roughness surface of the products reveals that ZnS nanoparticles are deposited on ACNTs successfully. From the high-magnification TEM image of ACNTs/ZnS nanocomposites (sample D) in Fig. 3(c), we observed that ZnS nanoparticles were well dispersed on ACNTs with uniform sharp along the axis of ACNTs. The size of ZnS nanoparticles is about 50 nm. The HRTEM image (Fig. 3(d)) of ZnS nanoparticles shows that the lattice fringes on the crystal face have a spacing of 0.3 nm, corresponding to the (111) plane of cubic ZnS. A selected area electron diffraction pattern of ZnS nanoparticles is presented in the inset of Fig. 3(d). The concentric rings of electron diffraction pattern indicate that the ZnS nanoparticles are polycrystalline.
Fig.3 Images of ACNTs/ZnS nanocomposites. (a) XRD; (b) SEM; (c) TEM; (d) HRTEM (insert is electron diffraction pattern)

Full size|PPT slide

We investigated the influence of initial amount of Zn(Ac)2·2H2O and TAA by varying the initial amount as 0.8, 0.6, and 0.4 mmol. The corresponding images are shown in Figs. 4(a)-(c). When the initial amount of Zn(Ac)2·2H2O was decreased from 0.8 mmol to 0.4 mmol , the size of ZnS nanoparticles was decreased from 500 to 100 nm. Meanwhile, the percentage of coverage and the quantity of ACNTs/ZnS nanocomposites were improved. When 0.8 mmol Zn(Ac)2·2H2O was used, only few ZnS nanoparticles are aggregated and deposited on ACNTs with large size (500 nm, Fig. 4(a)), and a large number of ZnS nanoparticles exist in the products. It indicates that the excess ZnS may not favor the homogenous growth of ACNTs/ZnS nanocomposites. When only 0.4 mmol of Zn(Ac)2·2H2O was added, the deposited ZnS nanoparticles are uniform with sizes about 100 nm (Fig. 4(c)). One reasonable explanation is that, along with the decrease of Zn and S source, the amount of ZnS nuclei decreases, leading to the decrease of the size of ZnS nanoparticles. The results show that the amounts of reactant play an important role in the synthesis of ACNTs/ZnS nanocomposites. It has to be noted that ACNTs became shorter after the deposit of ZnS nanoparticles. The phenomenon can be explained as follows. A large quantity of ZnS nanoparticles were deposited on ACNTs through chemical bond, leading to the reduction of the carbon-carbon bond energy of ACNTs. Thus, UV irradiation may provide enough energy to cutting off the ACNTs.
Fig.4 SEM images of ACNTs/ZnS nanocomposites corresponding to samples A-C in Table 1. (a) A; (b) B; (c) C

Full size|PPT slide

The effect of reaction time was also investigated, the corresponding images were shown in Fig. 5.When the reaction time was increased from 10 min to 1 h, the size of ZnS nanoparticles is increased from 20 nm to 50 nm .When the system only reacted for 10 min, only few ZnS nanoparticles are deposited on ACNTs with the size of around 20 nm (Fig. 5(a)). Along with the reaction proceeding, the size and coverage of ZnS nanoparticles are increased and at last the uniform ZnS nanoparticles are deposited on ACNTs surface. On the basis of our experiment process of synthesizing ACNTs/ZnS nanocomposites, a probable formation mechanism is proposed. Amorphous carbon nanotubes were used as substrate for deposition ZnS nanoparticles. At first, the positive metal ions Zn2+ were adsorbed onto the surfaces of ACNTs by electrostatic attraction. And then, S2- ions were released from TAA due to the UV irradiation. The S2- ions will be heterogeneously nucleated with Zn2+ ions to form ZnS nanoparticles on the surfaces of ACNTs. UV irradiation improved the decomposition of TAA and nucleation rate of ZnS nanoparticles. The main formation mechanism is depicted in Fig. 6.
Fig.5 TEM images of ACNTs/ZnS nanocomposites prepared with different reaction time. (a)10 min; (b) 30 min; (c) 1 h

Full size|PPT slide

Fig.6 Formation mechanism of ACNTs/ZnS nanocomposites

Full size|PPT slide

The photoluminescence spectrum of ZnS materials was complicated, with different structures having different spectra. Figure 7 depicts the photoluminescence spectrum of ACNTs/ZnS nanocomposites (sample D) and two emission bands are observed in the spectrum. One band at around 430 nm, which is attributed to the surface defect states. The other band is at around 474 nm, which is related to the self-activated defect centers [16].
Fig.7 Photoluminescence spectrum of ACNTs/ZnS nanocomposites

Full size|PPT slide

As the ZnS nanoparticles are homogenously deposited on ACNTs with small size, the ACNTs/ZnS nanocomposites may have an improved photocatalytic activity. The photocatalytic activity of the as-prepared ACNTs/ZnS nanocomposites (sample D) was evaluated by monitoring the absorption spectrum of MB, eosin and MR. In the case of MB and eosin, the dye was decolorized completely after about 50 min (Figs. 8(a) and 8(b)). As for MR, it needs 80 min for the entire degradation (Fig. 8(c)). Our results show that the as-prepared ACNTs/ZnS nanocomposites exhibit an efficient photocatalytic activity for different organic dye molecule, although the efficiency of the nanocomposites is different (Fig. 8(d)). The efficient degradation of the dye should be ascribed to that ZnS nanoparticles deposited on the surfaces of ACNTs had elevated specific surface area. Under the irradiation of UV, electrons and holes are generated in ZnS nanoparticles, meanwhile, the elevated specific surface area affords more active sites for photochemical reaction and thus enhance the photocatalytic activity.
Fig.8 Absorption spectra of different dyes degraded by ACNTs/ZnS nanocomposites. (a) MB; (b) eosin; (c) MR; (d) photodegradation rate of different dyes

Full size|PPT slide

4 Conclusions

An efficient photochemical route to synthesize ACNTs/ZnS nanocomposites has been proposed. For the abundant carboxyl groups on the ACNTs surface, ZnS nanoparticles can be easily deposited on it. The as-prepared ACNTs/ZnS nanocomposites contain ACNTs with a diameter about 100 nm and ZnS nanoparticles with sizes ranging from 50-500 nm. Photocatalysis study indicates the ACNTs/ZnS nanocomposites exhibit good activity for photodegradation of the organic dye molecule. The nanocomposites synthesized by this photochemical route are expected to find more applications in catalysts and photoelectricity fields.

References

[1]
Contes A, Carpenter B, Case C, Satheesh S, Suresh B, Wang T, Wu J D, Ng A Y. Text detection and character recognition in scene images with unsupervised feature learning. In: Proceedings of International Conference on Document Analysis and Recognition. Beijing: IEEE, 2011, 440–445
[2]
Ye Q, Doermann D. Text detection and recognition in imagery: a survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(7): 1480–1500
CrossRef Pubmed Google scholar
[3]
Zhang X, Gao X, Tian C. Text detection in natural scene images based on color prior guided MSER. Neurocomputing, 2018, 307: 61–71
[4]
Smith R. An overview of the tesseract OCR engine. In: Proceedings of International Conference on Document Analysis and Recognition. Parana: IEEE, 2007, 629–633
[5]
Epshtein B, Ofek E, Wexler Y. Detecting text in natural scenes with stroke width transform. In: Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition. San Francisco: IEEE, 2010, 2963–2970
[6]
Jaderberg M, Simonyan K, Vedaldi A, Zisserman A. Reading text in the wild with convolutional neural networks. International Journal of Computer Vision, 2016, 116(1): 1–20
CrossRef Google scholar
[7]
Ren S, He K, Girshick R, Sun J. Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137–1149
CrossRef Pubmed Google scholar
[8]
Liu W, Anguelov D, Erhan D, Szegedy C, Reed S E, Fu C, Berg A C. SSD: single shot MultiBox detector. In: Proceedings of European Conference on Computer Vision. Berlin: Springer, 2016, 21–37
[9]
Redmon J, Divvala S K, Girshick R B, Farhadi A. You only look once: unified, real-time object detection. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016, 779–788
[10]
Tian Z, Huang W, He T, He P, Qiao Y. Detecting text in natural image with connectionist text proposal network. In: Proceedings of European Conference on Computer Vision. Berlin: Springer, 2016, 56–72
[11]
Ma J, Shao W, Ye H, Wang L, Wang H, Zheng Y, Xue X. Arbitrary-oriented scene text detection via rotation proposals. IEEE Transactions on Multimedia, 2018, 20(11): 3111–3122
[12]
Liu Y, Jin L. Deep matching prior network: toward tighter multi-oriented text detection. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017, 3454–3461
[13]
Shi B, Bai X, Belongie S. Detecting oriented text in natural images by linking segments. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 3482–3490
[14]
Liao M, Shi B, Bai X, Wang X, Liu W. TextBoxes: A fast text detector with a single deep neural network. 2016, arXiv:1611.06779
[15]
Liao M, Shi B, Bai X. TextBoxes++: a single-shot oriented scene text detector. IEEE Transactions on Image Processing, 2018, 27(8): 3676–3690
CrossRef Pubmed Google scholar
[16]
Dai Y, Huang Z, Gao Y, Xu Y, Chen K, Guo J, Qiu W. Fused text segmentation networks for multi-oriented scene text detection. 2018, arXiv:1709.03272
[17]
Hu H, Zhang C, Luo Y, Wang Y, Han J, Ding E. WordSup: exploiting word annotations for character based text detection. In: Proceedings of IEEE International Conference on Computer Vision. Venice: IEEE, 2017, 4950–4959
[18]
Deng J, Dong W, Socher R, Li L J, Li K, Li F F. ImageNet: a large-scale hierarchical image database. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Miami: IEEE, 2009, 248–255
[19]
Karatzas D, Shafait F, Uchida S, Iwamura M, Bigorda L G I, Mestre S R, Mas J, Mota D F, Almazàn J A, Heras L P D L. ICDAR 2013 robust reading competition. In: Proceedings of International Conference on Document Analysis and Recognition. Washington, DC: IEEE, 2013, 1484–1493
[20]
Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. 2014, arXiv:1409.1556
[21]
Howard A G, Zhu M, Chen B, Kalenichenko D, Wang W, Weyand T, Andreetto M, Adam H. MobileNets: efficient convolutional neural networks for mobile vision applications. 2017, arXiv:1704.04861
[22]
Zhang X, Zhou X, Lin M, Sun J. ShuffleNet: an extremely efficient convolutional neural network for mobile devices. 2017, arXiv:1707.01083v2
[23]
Iandola F N, Han S, Moskewicz M W, Ashraf K, Dally W J, Keutzer K. SqueezeNet: AlexNet-level accuracy with 50 ´ fewer parameters and<0.5 MB model size. 2016, arXiv:1602.07360v4

RIGHTS & PERMISSIONS

2019 Higher Education Press and Springer-Verlag GmbH Germany, part of Springer Nature
AI Summary AI Mindmap
PDF(1725 KB)

Accesses

Citations

Detail

Sections
Recommended

/