Abstract
This work focuses on the quality assessment of agricultural product based on microscopic image, generated by Foldscope. Microscopic image-based food quality assessment always be an efficient method, but its system complexity, costly, bulk size and requirement of special expertise confines it usability. To encounter such issues, Foldscope which is small, lightweight, cheap and easy to use has been considered to verify its usability as food quality assessment device. In this purpose, measuring starch of potato has been selected to check its compatibility and microscopic images are taken from two image modalities—conventional microscope and Foldscope and the results have been compared. The image processing techniques including morphological filtering followed by Otsu’s method has been employed to detect starch efficiently. In total, 20 images from each of the system have been captured. Following the experiment, the presence of starch (in %) estimated based on the image taken from microscope and Foldscope are 23.50 ± 0.79 and 24.29 ± 0.73 respectively, which is consistent. Such results reveal that the Foldscope can be used in food quality assessment system, which could make such devices simple, portable and handy.
You have full access to this open access chapter, Download conference paper PDF
Similar content being viewed by others
Keywords
1 Introduction
Food quality assessment is always being an important part at the time of purchasing or consuming any sort of foods. Now-a-days, it becomes more evident as fertilizer and pesticides in crops are being used enormously without maintaining the nutrition value to meet the huge demand of food for rapidly growing population. In addition, harmful chemicals are also being used in vegetables and fruits for immoral purposes; for instance, Copper Sulphate (CuSO4) is used to look the vegetable fresh; Calcium Carbide (CaC2) is used to ripe fruits artificially and so on. Besides, several synthetic colors are applied on vegetables to maintain their freshness. Such malpractices make the food quality assessment more evident. Generally, in market the quality of the agricultural products are evaluated based on their shape, size, colour etc. which are performed by means of vision, touch, smell, odour, taste, flavour etc. Indeed, all these methods are tedious, time consuming, very, much subjective, but not even very reliable. Therefore, it is very significant to develop a rapid, reliable, easy to use system to examine the agricultural products.
In this direction, comparing to other aforementioned manual techniques to assess the food quality, the visual perceptions (usually bare eyes) dominate the other means, which is not only inefficient but also unreliable due to subjectivity, inconsistency etc. However, such obscures can be tackled by the computer vision technology that corresponds to the human vision, in inspecting quality of fruits and vegetables. It captures images electronically, then interpret and recognize the characters/information to examine the quality, and grading of agricultural products. In this direction, several attempts have been made for grading of food including apples [1, 2], mango [2] sweet pepper [3], papaya [4], tomato [5, 6] etc. However, all the methods consider the traditional camera images, taken from the surface of the vegetables and fruits, which are failed to provide inside details such as nutrients contents of a products. Such details are the key components to quantify the food quality or grading.
Certainly, the microscopic images (resolution of 200 nm or smaller) provide insight information at cell level which could be appropriate to examine the food quality more precisely than the use of conventional camera images. Few works have been conducted in this direction based on such microscopic images to evaluate the quality of vegetable and fruits such as evaluation of potato and its starch [7], browning of apple due to storage [8], and effect of freezing in blueberries [9]. However, these images are usually generated by microscopes which are expensive, complex, bulky and need some sort of expertise to operate. In this regard, recently, a newly developed light weight (~9 gm), low cost (~Rupees 250), and 140X magnification powered origami based paper microscopic called Foldscope [10] could be a good solution to generate the microscopic images. Besides, the generated images can be captured directly by simple mobile camera. Thus, a combining Foldscope with mobile camera for capturing microscopic images and further employing computer vision and image processing techniques to analyze those images could be a very proper food quality assessment system. Therefore, in this work, microscopic images of potato has been considered to verify the usability of such device. The potato has been used as one of the most important crop consumed in India after rice, wheat and maize [11]. Usually, it contains water (80%) and dry matter (20%) which is well-known as starch [11]. Inevitability, these starch amount defines the quality of a potato [12].
In a microscopic image of potato, usually starches are glided on the surface of the cells which makes it challenging to distinguish from regular cells. In this purpose, staining method enhances either cell boundaries or starch, could be useful in such detection process. But, it requires not only involvement of additional arrangement (chemical process etc.), but also high-end microscope (like electron microscopes) [13]. Another way, use of image processing techniques such as morphological image enhancement on microscopic image directly could be very convenient to sidestep staining process [14]. Certainly, a well-established, simple and accurate image segmentation techniques such as Otsu’s segmentation method [15] can be employed. It works with intensity of an image and considers thresholding technique to discriminate two set of classes of an image. Thus, it could be applicable to distinct starch from cell background. Nevertheless, such thresholding sometimes leads to improper segmentation that can be resolved easily by morphological filtering. Therefore, measuring the starch content using microscopic images which will be captured by above mentioned setup and deploying image processing technique is the main objective of this work.
The paper has been organized as follows: Sect. 2 provides a system overview of the proposed methods. In Sects. 3 and 4, the methodology and experiment have been elaborated respectively. Results and discussion have summarized in Sect. 5. Finally, conclusions have been drawn in Sect. 6.
2 System Overview
A system overview of the proposed method has been displayed in Fig. 1 which involves several stages including sample preparation, image acquisition, image processing, starch detection and estimation. Firstly, conventional methods for acquiring microscopic images such as sectioning of samples etc. are conducted to acquire microscopic image of potato. Next, images are generated using Foldscope as well as conventional microscopic. Further, image processing techniques such as morphological filtering followed by segmentation are conducted to detect the starch from those two sets of images. Finally, the percentage of starch has been estimated and the results are compared.
3 Methodology
Aim of the proposed work is to measure the starch content in potato directly (avoiding traditional staining process) from microscopic image which are acquired from two image acquisition modalities—Foldscope and traditional microscopic. Generated images from both of the modalities consist of starch spreading over regular cells and its boundaries. However, the intensity difference between them is relatively small which confines to employ simple thresholding technique to discriminate starch from regular cells—leads to under segmentation. Certainly, a popular segmentation method known as Otsu’s method which performs the thresholding, based on statistical measures can be used. Nonetheless, direct use of Otsu’s method sometimes results over segmentation which can be overcome by a morphological filtering. Therefore, a morphological filtering has been conducted followed by Otsu’s algorithm to improve the system performances by eradicating uneven illumination and the cell boundaries. Further, to estimate the starch quantity efficiently binarization and morphological operation have been performed.
3.1 Morphological Filtering
Morphological operation analyses the geometrical structure of an image. It performs dilation, erosion, opening, closing operation with an image of definite shape and size (structuring element). Morphological filter can be constructed on the basis of morphological operations. It offers better result than the linear filtering as it deforms the image geometry. The morphological filter has been designed based on two morphological operations called top-hat and bottom hat transform by opening and closing operations respectively. The whole operation performs on grayscale image. The brightest parts of the images are enhanced by top-hat transform; whereas, the bottom-hat transform does the reverse process. Mathematically, for a grayscale image \( g(x,y) \) with structuring element \( a(x,y) \),
Top-hat transform followed by bottom-hat transform increases the contrast between cell boundaries and starch which further benefits to discriminate the starch distinctly. The contrast enhanced image (h) is obtained by adding the image itself with its top-hat transform as shown in (3); whereas, the starch boundaries (i) are further enhanced by subtracting the bottom-hat transformed image from contrast enhanced image (h) as displayed in (4).
Furthermore, the starches have been discriminated from (3) and (4). These steps define morphological filtering of gray scale images.
3.2 Starch Detection
As mentioned, the segmentation operation has been performed by Otsu’s method [15] which estimated the threshold values based on statistical measures. It separates the image into two classes based on threshold value t. Further, t can be optimized by either minimizing the weighted sum of within-class variance (\( \sigma_{w} \)) or maximizing the between-class variance (\( \sigma_{b} \)).
Two class probabilities \( q_{c1} (t) \) and \( q_{c2} (t) \) are calculated in terms of intensity probability P(k) for different intensity k, from I bins of histogram expressed as,
and class means \( \mu_{c1} (t) \) and \( \mu_{c2} (t) \) expressed as
Therefore, the total class variance,\( \sigma_{total(cv)} \) for any given threshold t in terms of class probabilities (qc) and class mean \( \sigma \) can be expressed as
And between-class variance
The algorithm has been worked as follows:

4 Experiment
4.1 Experimental Setup
The microscopic images are generated and captured by two modalities as mentioned earlier. One is the Foldscope which is an origami based optical microscope which can be assembled from a printed A4 size paper in 10 min. The dimension of the device is 70 × 20 × 2 mm3 and of weight 8 gm. The images are generated by 140 × magnifications with 2micron resolution. Another, a traditional microscope Olympus CH20i using 10X objective lens has been used to generate images. Two set of images of same sample has been generated and captured by a smart phone camera (MotoG4 Play, 1280X720 HD, 294 ppi, 5 inches diagonal display, 8 MP). The simulation has been conducted using Matlab tool.
4.2 Data
For experimental validation, potato starch has been taken into account. Among different varieties of potatoes, the ‘Jyoti’ variety is been selected as it is largely consumed in India. The specification of the ‘Jyoti’ variety has been briefed in Table 1. To generate the microscopic images, thin sections of potato are placed under the Foldscope as well as the traditional microscope and consequently the images are captured. The images display the cell boundaries along with starch as indicated in Fig. 3(a) and (b) which are captured by traditional microscopic and Foldscope respectively. In total 20 images from both of the modalities are taken.
5 Results and Discussions
The whole experiment performed on grayscale image and after morphological filtering the enhanced images has been displayed in Fig. 4(a) and (b) in which few cell boundaries can be observed; whereas starches are very prominent.
Next, the starches have been detected by employing Otsu’s method as illustrated in Fig. 5, in which the cell boundaries are totally removed .
Further, the starch present in potato has been calculated for both sets of images and their results are compared. The presence of starch has been calculated from the final image using (11)
where, \( P_{s} \) and \( BP_{s} \) are the total no of pixels and black pixel (Starch) of the sample image respectively. The presence of starch (in %) estimated based on the image taken from microscope and Foldscope are 23.50 ± 0.79 and 24.29 ± 0.73 respectively and plotted in Fig. 6 in which the horizontal and vertical axes refers to measured starch and image generation modalities respectively. As seen the results are very much consistent.
6 Conclusion
In this wok, two set of microscopic images are generated and captured by conventional microscope and Foldscope which is a low cost, portable, paper microscope. Aiming its usability in food quality assessment system, estimation of starch in potato has been considered. The microscopic images are generated without staining by Foldscope as well as conventional microscope and captured by mobile camera. Next, morphological filtering followed by and Otsu’s method has been employed. The starch of potato for both the image modalities are detected and measured successfully. The results for both of the cases are very consistent which offers that the Foldscope could be used in food quality assessment system. In future, more set of images along with several crops will be verified. In addition, deep learning-based recognition system will be incorporated to improve is efficiently.
References
Cárdenas-Pérez, S., Chanona-Pérez, J., Méndez-Méndez, J.V., Calderón-Domínguez, G., López-Santiago, R., Perea-Flores, M.J.: Evaluation of the ripening stages of apple (Golden Delicious) by means of computer vision system. Biosyst. Eng. 159, 46–58 (2017). https://doi.org/10.1016/j.biosystemseng.2017.04.009
Ali, M.A., Thai, K.W.: Automated fruit grading system. In: IEEE International Symposium in Robotics and Manufacturing Automation (ROMA), Kuala Lumpur, Malaysia, pp. 1–6, September 2017
Sa, I., Ge, Z., Dayoub, F., Upcroft, B., Perez, T., McCool, C.: Deepfruits: a fruit detection system using deep neural networks. Sensors 16, 1222 (2016)
Pereira, L.F.S., Barbon Jr., S., Valous, N.A., Barbin, D.F.: Predicting the ripening of papaya fruit with digital imaging and random forests. Comput. Electron. Agric. 145, 76–82 (2018)
Mehra, T., Kumar, V., Gupta, P.: Maturity and disease detection in tomato using computer vision. In: International Conference on Parallel, Distributed and Grid Computing (PDGC), Solan, India, pp. 399–403, December 2016
Arakeri, M.P.: Computer vision based fruit grading system for quality evaluation of tomato in agriculture industry. Procedia Comput. Sci. 79, 426–433 (2016)
Liu, Q., Donner, E., Tarn, R., Singh, J., Chung, H.-J.: Advanced analytical techniques to evaluate the quality of potato and potato starch. In: Advances in Potato Chemistry and Technology, pp. 221–248. Elsevier (2009)
Cropotova, J., Tylewicz, U., Cocci, E., Romani, S., Dalla Rosa, M.: A novel fluorescence microscopy approach to estimate quality loss of stored fruit fillings as a result of browning. Food Chem. 194, 175–183 (2016)
Allan-Wojtas, P., Goff, H., Stark, R., Carbyn, S.: The effect of freezing method and frozen storage conditions on the microstructure of wild blueberries as observed by cold-stage scanning electron microscopy. Scanning 21, 334–347 (1999)
Foldscope. https://www.foldscope.com
Borah, S., Bowmick, B., Hazarika, C.: Production behaviour of potato in Assam-A critical analysis across zones and size groups of farms. Econ. Affairs 61, 23 (2016)
Lutaladio, N., Castaldi, L.: Potato: the hidden treasure. J. Food Compos. Anal. 22, 491–493 (2009)
Liu, Q., Zhang, L.J., Liu, X.P.: Microscopic image segmentation of Chinese herbal medicine based on region growing algorithm. In: Advanced Materials Research, pp. 4110–4115 (2013)
Ravi, S., Khan, A.: Morphological operations for image processing: understanding and its applications. In: Proceedings of 2nd National Conference on VLSI, Signal Processing & Communications NCVSComs, Guntur, India, December 2013
Otsu, N.: A threshold selection method from gray-level histograms. IEEE Trans. Syst. Man Cybern. 9, 62–66 (1979)
Acknowledgement
The work has been supported by Department of Science and Technology, Govt. of India under IMRPINT-II with number IMP/2018/000538.
Author information
Authors and Affiliations
Corresponding authors
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this paper
Cite this paper
Biswas, S., Barma, S. (2019). Usability of Foldscope in Food Quality Assessment Device. In: Deka, B., Maji, P., Mitra, S., Bhattacharyya, D., Bora, P., Pal, S. (eds) Pattern Recognition and Machine Intelligence. PReMI 2019. Lecture Notes in Computer Science(), vol 11941. Springer, Cham. https://doi.org/10.1007/978-3-030-34869-4_53
Download citation
DOI: https://doi.org/10.1007/978-3-030-34869-4_53
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-34868-7
Online ISBN: 978-3-030-34869-4
eBook Packages: Computer ScienceComputer Science (R0)