Application of image processing technique to extract morphological characteristics of weedy rice seeds variants for Malaysian seed industry.
Over the past few years, machine vision is gaining popularity and widely used as a tool for quality control in seed production industry. With the advancement of electronic and computing power, image processing technique has proved to be more accurate and consuming less processing time compared to traditional  and non-destructive method. In recent studies, the extraction of seed morphological characteristics using image processing technique were carried out to identify and classify seed varieties by machine vision have been reported in the literature [3,4, 5,6,7].Majumdar & Jayas  applied image processing technique to extracts morphological features and classify different grains using machine vision. The images of non-touching individuals kernel were acquired on a black background and were converted from rectangular pixel images into square pixel images. Then, the images were thresholded using an automatic thresholding technique. The holes inside the thresholded images of individual kernels were filled with a hole filling algorithm. The algorithms to extract morphological characteristics of the cereal grain seeds were developed in C++ programming language. According to Majumdar & Jayas , the results of classification accuracies ranged from 81.6% up to 99.9%.
Bakash et al.  studied the shape, size and weight of 15 annual weed seeds buried in soil seed bank in Pakistan for future development of weed seeds identification, classification and management. The results of their studies suggested that the single characteristic is not enough to distinguish weed seeds as some seed specieses were found to have the same mean value howeverthe consideration of all three characteristics was found to be unique. Thus, they considered these characteristics may be sufficient as a convenient method for identification and classification of weed seeds species buried in soil seed bank.
Weedy rice (Oryza sativa complex) locally known as 'padi angin' is the main competitors to cultivated rice . Species crosses in the field resulted in the development of various weedy rice variants and appears to have intermediate morphological characteristics between cultivated and wild rice . The evolution of weedy rice seeds variants that mimics cultivated rice seeds makes the mechanical separation of weedy rice seeds variants from cultivated rice seeds in seed processing plant become more difficult. Thus, the separation performance of rice seeds in seed processing plant can be improved through the incorporation of an additional separation process that use seed morphological characteristics information extracted from image processing technique. The main purpose of this study was to evaluate selected morphological characteristics of weedy rice seeds variants samples using image processing technique. These information are important in evaluating the variation of weedy rice seeds variants for future development of cultivated rice seeds classification system to separate weedy rice from cultivated rice seeds using image processing and analysis technique.
MATERIALS AND METHODS
Weedy rice seeds variants, collected from several commercial farms in Kedah were labelled as C1 (close panicle), C2 (partly short awned, open panicle), C3 (close panicle), C4 (partly short awned, close panicle) and C5 (partly long awned, close panicle). Each weedy rice seeds variants were collected from a single matured plant by the officers from Rice Seed Testing Laboratory, Department of Agriculture at Teluk Chengai, Kedah to ensure its purity. All weedy rice seeds were air dried to 13-14% moisture content and cleaned to preserve its shelf life.
A CCD colour camera (Basler acA1600-20gc) was used for image acquisition. The camera fitted with 12 mm focal length lens was mounted on top of black box (60.8 cm length x 30.55 cm breadth x 60.0 cm height) perpendicular to the seed plate position. Two light-emitting diode (LED) light bulbs with 6400 K colour temperature was placed on both sides of the seed plate for even illumination. The field of view of the camera were set to cover two rows of seed holes which contained a total of 8 holes in an image frame. The distance between camera lens and the seeds was 7.4 cm. The seeds were placed on a fluorescent green surface with elliptical holesseed plate.
Six images of each weedy rice seeds variants were taken for image processing and analysis. The total number of weedy rice seeds variants samples used in this study was 48 seeds per variants. All images were acquired in Red, Green and Blue (RGB) colour model and stored in Portable Network Graphics (PNG) format with 2-megapixel resolutions (1628 x 1236 pixels). The acquired images were loaded in LabVIEW for image processing and analysis.
Image Processing and Morphological Characteristics Extraction:
The software applications used in this study was LabVIEW 2012 development environment. First, the image was defined to be extracted by red colour plane. The image was then converted to grey level for further processing. An edge enhancement was performed by using Laplacian filters. Then, the image was thresholded to separate the seed grains from the background image. Morphological image processing technique was then carried out to enhance the desired image for morphological characteristics extraction. Dilation operation was carried out to close the countours of every seed grains images . Next, small particles were removed to eliminate noisy particles and unwanted particles formed on the background image. Then, hole filling operation was employed to fill the holes found in seed grain images with pixels value 1. The final operation was done by copying the image source that corresponded to a non-zero pixel value in the mask image to display the processed image. The extraction of individual morphological characteristics of weedy rice seeds variants were carried out to extract individual seed length, width, rectangular aspect ratio, major axis length, minor axis length and aspect ratio [3, 4].All the obtained values were multiplied by the calibration factor (mm/pixel) to convert pixels unit to milimeter (mm).
Weedy rice seeds length, width, rectangular aspect ratio, major axis length, minor axis length, and aspect ratio were analyzedusing one way analysis of variance (ANOVA) by comparing the means of each morphological characteristics between weedy rice seeds variants samples. Tukey-Kramer procedure was then used to compare all pairs of weedy rice seeds variants for each morphological characteristicsto identify which pairs are insignificantly differentin term of means value. The significance level of the analysis was established at 95%.
RESULTS AND DISCUSSIONS
The one way ANOVA results from Table 1 indicated that there were significant difference in term of mean values of seeds length, width, rectangular aspect ratio, major axis length, minor axis length, and aspect ratio among the weedy rice seeds variants. The findings indicated that each weedy rice seeds variants can be distinguished according to the selected morphological characteristics.
The comparison between variants pair were further analyzed using Tukey-Kramer procedure. The mean values for all morphological characteristics and comparison results by Tukey Kramer were summarized as in Table 2. It was found that C1-C4 pair have the most insignificant difference in term of mean seed length, rectangular aspect ratio, major axis length and aspect ratio. The second most insignificant difference pairs in term of mean values were C3-C5 pair for rectangular aspect ratio and minor axis length and C2-C4 pair for minor axis length and aspect ratio. Meanwhile, C1-C5 and C2-C3 pairs were the found to have insignificant difference mean values of seed width.
The insignificant difference in mean values of some morphological characteristics indicated that the single morphological characteristic is not sufficient to distinguish individual weedy rice seeds by its variants as there were more than one weedy rice seed variants possess similar mean value. This finding wasalso reported by Cheng et al. . Thus, the combination of more than one morphological characteristics for further classification work need to be considered to achieve higher classification accuracy.
The morphological characteristics of seed length, width, rectangular aspect ratio, major axis length, minor axis length and aspect ratio were obtained by the extraction of processed image in LabVIEW environment of the acquired weedy rice seeds variants images. The comparison results indicated that weedy rice seed variants C1 and C4 samples have the closest similarity in morphological characteristics reflected from the insignificant difference in term of mean values of seed length, rectangular aspect ratio, major axis length and aspect ratio. The mean values of the extracted morphological characteristics of other seeds variants pairs were also found to be insignificantly difference. However, the combination of the extracted morphological characteristics were found to be unique and can be considered for further seed identification and classification work. As a recommendation, more morphological characteristics need to be extracted using image processing to obtain sufficient information to achieve higher seed classification accuracy.
Received 25 September 2014
Received in revised form 26 October 2014
Accepted 25 November 2014
Available online 31 December 2014
We would like to thank Ministry of Education for funding this research under the Fundamental Research Grant Scheme (FRGS) 9003-00388.
 Vibhute, A., S.K. Bodhe, 2013. Applications of Image Processing in Agriculture: A Survey. International Journal of Computer Applications, 52-2: 34-40.
 Varma, V.S., K.K. Durga, K. Kheshavulu, 2013. Seed Image Analysis: Its Applications in Seed Science Research. International Research Journal of Agricultural Sciences, 1-2: 30-36.
 Majumdar, S., D.S. Jayas, 2000a. Classification of Cereal Grains Using Machine Vision: I. Morphology Models. Transaction of the ASAE, 43-6: 1669-1675.
 Cheng, F., Y.B. Ying, Z.Y. Liu, 2005. Rice Seeds Information System Based on Artificial Neural Network. Proceedings of SPIE 5999: Intelligent Systems in Design and Manufacturing VI, 599910.
 Rad, S.J.M., F.A. Tab, K. Mollazade, 2012. Design of an Expert System for Rice Kernel Identification using Optimal Morphological Features and Back Propagation Neural Network. International Journal of Applied Information Systems (IJAIS), 3-2: 2249-0868.
 Hernandez, F.G., J.G. Gil, 2011. A Machine Vision System for Classification of Wheatand Barley Grain Kernels. Spanish Journal of Agricultural Research, 9-3: 672-680.
 Paliwal, J., N.S. Visen, D.V. Jayas, N.D.G. White, 2003.Comparison of a Neural Network and a Non- parametric Classifier for Grain Kernel Identification. Biosystems Engineering, 85-4: 405-413.
 Bakash, A., A.A. Dasti, A. Munir, I. Khaliq, M.A. Din, M.S. Akhtar, 2006. Studies on Size, Shape and Weight of Certain Weed Seeds Buried in the Soil Seed Bank. Pakistan Journal of Weed Science Research, 12: 79-82.
 Zainudin, H., M. Azmi, A.H. Othman, 2010. Morphological Study of the Relationships between Weedy Rice Accessions (Oryza sativa complex) and Commercial Rice Varieties in Pulau Pinang Rice Granary Area. Tropical Life Sciences Research, 21-2: 27-40.
 Cao, Q, B.R. Lu, H. Xia, J. Rong, F. Sala, A. Spada, F. Grassi, 2006. Genetic diversity and origin of weedy rice (Oryza sativa f. spontanea) populations found in North-eastern China revealed by simple sequence repeat (SSR) markers. Annals of Botany, 98-6: 1241-1252.
 Ghadge, P.N., K. Prasad, 2012. Some physical properties of rice kernels: Variety PR-106. J. Food Process Technology, 3-8: 175.
 IMAQ Vision for LabVIEW User Manual, National Instruments Corp., 2000.
(1) A.A. Aznan, (1) I.H. Rukunudin, (2) A.Y.M. Shakaff, (1) R. Ruslan, (2) A. Zakaria and (2) F.S.A. Saad
(1) School of Bioprocess Engineering, Universiti Malaysia Perlis, Kompleks Pusat Pegajian Jejawi 3, 02600 Arau, Perlis, Malaysia.
(2) Centre of Excellence for Advanced Sensor Technology, Universiti Malaysia Perlis ,Pusat Pengajian Jejawi 2, Jalan Jejawi Permatang, 02600 Jejawi, Arau, Perlis, Malaysia.
Corresponding Author: A.A. Aznan, School of Bioprocess Engineering, Universiti Malaysia Perlis, KompleksPusatPegajianJejawi 3, 02600 Arau, Perlis, Malaysia.
Table 1: One way ANOVA. Morphological Source of SS df MS F Characteristics Variation Length Between Groups 416.866 4 104.216 43.575 Within Groups 562.041 235 2.392 Total 978.907 239 Width Between Groups 2.063 4 0.5 16 22.909 Within Groups 5.292 235 0.023 Total 7.355 239 Rectangular Between Groups 69.300 4 17.325 51.832 Aspect Ratio Within Groups 78.549 235 0.334 Total 147.850 239 Major Axis Length Between Groups 593.915 4 148.479 61.984 Within Groups 562.927 235 2.395 Total 1156.841 239 Minor Axis Length Between Groups 5.256 4 1.3 14 25.431 Within Groups 12.144 235 0.052 Total 17.400 239 Aspect Ratio Between Groups 250.798 4 62.699 40.782 Within Groups 361.299 235 1.537 Total 612.096 239 Morphological Source of P-value F critical Characteristics Variation Length Between Groups 2.47E-27 2.410 Within Groups Total Width Between Groups 5.36E-16 2.410 Within Groups Total Rectangular Between Groups 2.98E-31 2.410 Aspect Ratio Within Groups Total Major Axis Length Between Groups 1.07E-35 2.410 Within Groups Total Minor Axis Length Between Groups 1.62E-17 2.410 Within Groups Total Aspect Ratio Between Groups 6.16E-26 2.410 Within Groups Total Table 2: Extracted morphological characteristics of weedy rice seeds variants using image processing technique. Morphological Characteristics Weedy Rice Seeds Variants C1 C2 Length [mm] Mean Value 8.312 9.724 Standard Deviation 0.248 0.673 Insignificant Difference Pairs C1-C4 in Term of Mean Value Width [mm] Mean Value 2.806 2.706 Standard Deviation 0.176 0.173 Insignificant Difference Pairs C1-C5 C2-C3 in Term of Mean Value Rectangular Mean Value 2.977 3.610 aspect ratio Standard Deviation 0.255 0.363 Insignificant Difference Pairs C1-C4 C3-C5 in Term of Mean Value Major axis Mean Value 9.015 10.894 length [mm] Standard Deviation 0.617 0.916 Insignificant Difference Pairs C1-C4 in Term of Mean Value Minor axis Mean Value 2.602 2.450 length [mm] Standard Deviation 0.253 0.153 Insignificant Difference Pairs C2-C4 C3 and C5 in Term of Mean Value Aspect ratio Mean Value 3.494 4.475 Standard Deviation 0.403 0.567 Insignificant Difference Pairs C1-C4 C2-C4 in Term of Mean Value Morphological Characteristics Weedy Rice Seeds Variants C3 C4 Length [mm] Mean Value 10.769 8.448 Standard Deviation 0.362 1.326 Insignificant Difference Pairs C1-C4 in Term of Mean Value Width [mm] Mean Value 2.625 2.895 Standard Deviation 0.131 0.119 Insignificant Difference Pairs C1-C5 C2-C3 in Term of Mean Value Rectangular Mean Value 4.113 2.923 aspect ratio Standard Deviation 0.261 0.486 Insignificant Difference Pairs C1-C4 C3-C5 in Term of Mean Value Major axis Mean Value 12.106 9.501 length [mm] Standard Deviation 0.827 1.476 Insignificant Difference Pairs C1-C4 in Term of Mean Value Minor axis Mean Value 2.290 2.458 length [mm] Standard Deviation 0.141 0.248 Insignificant Difference Pairs C2-C4 C3 and C5 in Term of Mean Value Aspect ratio Mean Value 5.324 3.969 Standard Deviation 0.654 1.150 Insignificant Difference Pairs C1-C4 C2-C4 in Term of Mean Value Morphological Characteristics Weedy Rice Seeds Variants C5 Length [mm] Mean Value 11.723 Standard Deviation 3.050 Insignificant Difference Pairs C1-C4 in Term of Mean Value Width [mm] Mean Value 2.802 Standard Deviation 0.135 Insignificant Difference Pairs C1-C5 C2-C3 in Term of Mean Value Rectangular Mean Value 4.185 aspect ratio Standard Deviation 1.066 Insignificant Difference Pairs C1-C4 C3-C5 in Term of Mean Value Major axis Mean Value 13.228 length [mm] Standard Deviation 2.765 Insignificant Difference Pairs C1-C4 in Term of Mean Value Minor axis Mean Value 2.174 length [mm] Standard Deviation 0.290 Insignificant Difference Pairs C2-C4 C3 and C5 in Term of Mean Value Aspect ratio Mean Value 6.373 Standard Deviation 2.301 Insignificant Difference Pairs C1-C4 C2-C4 in Term of Mean Value
|Printer friendly Cite/link Email Feedback|
|Author:||Aznan, A.A.; Rukunudin, I.H.; Shakaff, A.Y.M.; Ruslan, R.; Zakaria, A.; Saad, F.S.A.|
|Publication:||Advances in Environmental Biology|
|Date:||Nov 1, 2014|
|Previous Article:||Integration of biosystems engineering for mango production in tropical greenhouse.|
|Next Article:||Study of tidal current power generation by vertical axis underwater turbine with diffuser in Costal Peninsular Malaysia.|