NON-DESTRUCTIVE DETECTION OF FROST DAMAGE IN SWEET LEMON USING IMAGE PROCESSING AND ULTRAVIOLET RADIATION

Aedi Firouzjaei et al RJLBPCS 18 www.rjlps.om Life Siene Informatis Puliations Original Researh Artile DOI: 1.679/18..17 NON-DESTRUCTIVE DETECTION OF FROST DAMAGE IN SWEET LEMON USING IMAGE PROCESSING AND ULTRAVIOLET RADIATION Rouhallah Aedi Firouzjaei, Saeid Minaei *, Baak Beheshti Department of Agriultural Systems Engineering, Siene and Researh Branh, Islami Azad University, Tehran, Iran. ABSTRACT: Citrus peel is more resistant to frost damage ompared to the interellular flesh. Thus, it may e possile that the internal tissues of apparently healthy fruits are damaged and devoid of juie. Frost damage under UV radiation appears as light green spots on the fruit skin. The purpose of this study was to propose a fast, non-destrutive method for sweet lemon frost damage detetion and to evaluate the extent of frost damage using image proessing tehnique and UV radiation. This study was done on 135 sweet lemons. The samples at 3 levels of temperature (, 7 and -7 C) were exposed to old treatment (6, 1 and hours). Sample images were aptured under UV light 1, and 36 hours post treatment. The images analyzed using MATLAB software. Green Spot Index or GSI was used to show the extent of fruit frost damage. Results were ompared with the onventional visual method of frost damage assessment y a team of experts. Results showed that the perentage of GSI is signifiant at the 1% level onsidering main, doule and triple interation effets of all treatments. GSI inreased signifiantly y dereasing the temperature to -7 C. By inreasing the duration of old treatment from 6 to hours and inreasing the waiting period from 1 to 36 hours, the green spot index inreased dramatially. The detetion auray of frost damage ased on the perentage of green spots on the lemon fruit skin was aquired to e equal to 9.6% ompared to manual assessment. KEYWORDS: sweet lemon, freezing temperatures, UV radiation, omputer vision. *Corresponding Author: Dr. Saeid Minaei* Ph.D. Department of Agriultural Systems Engineering, Siene and Researh Branh, Islami Azad University, Tehran, Iran. Email Address: minaee@modares.a.ir 18 Life Siene Informatis Puliation All rights reserved 18 July - August RJLBPCS () Page No.1

Aedi Firouzjaei et al RJLBPCS 18 www.rjlps.om Life Siene Informatis Puliations 1.INTRODUCTION Citrus is a native of Southeast Asia [1]. The origin of that region is etween India, South China and Indonesia []. Lemon elongs to the itrus family and is generally produed in moderate as well as in dry, hot limates. Ranked after orange and tangerine, sweet lemon (Citrus limetta) is the third important itrus fruit [3]. Frost damage in itrus fruits is one of the major reasons for degradation of fruits in the itrus industry. Assessment of frost damage in itrus fruits is very hallenging sine it annot e deteted visually with the naked eye [, 5, 6]. Chilling and old onditions ause damage in itrus (the tree as well as the fruits). Damaged itrus fruits are not suitale for onsumption due to their poor taste and low juie ontent as frost damage results in fruit degradation [7, 8]. The destrutive effets of hilling have een oserved in itrus growing ountries suh as Spain, Italy, Greee, Turkey, Ameria and Japan [9]. During the first weeks after the hilling inidene, the evaluation of damage level to fruits is done y inspetors through areful investigation of flesh segments using visual and segment ut method [1]. This method is destrutive and time-onsuming. A wide variety of instruments has een tested to detet the frost damage in itrus fruits. For instane appliation of hromatography and mass spetrometry for the non-destrutive measurement of greenhouse volatile gases emitted from healthy and damaged oranges for one hour, shows that the ethanol, ethylene otanoate, methyl hexanuate and ethyl otanoate emissions inrease in ompletely frozen Navel oranges [11]. A gas sensor was used for frost damage detetion in oranges. In this method, the level of ethanol gas in the headspae of the sample preservation ontainer was used for the assessment. The auray of the mentioned method was 1 and 37 perent for healthy and partially frozen fruits, respetively [1]. In order to inrease the auray of this method, a handheld (manual) ethanol sensor with higher sensitivity ompared to the one used in Tan et al, [13] study was applied for measuring the ethanol level in oranges with the aim of frost damage detetion [1, 15].In another researh, Nulear Magneti Resonane (NMR) spetrometer was utilized for frost damage investigation in the flesh segment of itrus [5, 16]. Making use of NMR is very omplex and expensive and requires extensive studies. A smell sensor has een used for quality evaluation of oranges. This sensor is sensitive to a wide range of volatile materials suh as aromati mixtures, alohols and aldehydes [17]. Making use of ultraviolet radiation researhers have found a non-destrutive method for frost damage detetion. This kind of damage has its effets on itrus peel oil in the form of oservale ompounds on the skin surfae. The results showed that the visual appearane of frozen oranges is different and a vivid pattern of yellow spots in the range of 1- mm width an e oserved when plaed under UV light (365nm) [18,19]. This experiment was onduted on California Navel Oranges. A researh was performed to find the est exitation wavelength for sattering the fluoresene of itrus peel under UV light (3-7nm) on 15 speies with the aim of frost damage detetion in itrus fruit. Results showed that the appropriate exitation wavelength to ahieve the est sattering of fluoresene is in the ranges of 35-38 nm and 9-18 Life Siene Informatis Puliation All rights reserved 18 July - August RJLBPCS () Page No.11

Aedi Firouzjaei et al RJLBPCS 18 www.rjlps.om Life Siene Informatis Puliations 5nm [18, ]. Overtime, frost damage leads to fruit spoilage and if not identified in time, it an spread and ause the spoilage of adjaent fruits. Thus, the itrus industry requires a fast and nondestrutive method in order to distinguish frost damage in itrus fruits. The goal of this study is to employ a non-destrutive and relatively fast proedure to identify frost damage in sweet lemon fruits and evaluate the extent of damage utilizing image proessing tehniques.. MATERIALS AND METHODS In order to ondut the experiments, 135 sweet lemon fruits were harvested on Novemer 1 from an orhard loated in Ghaemshahr, Mazandaran Provine in Iran. The samples were stored in a refrigerator for 8 hours at 1 C efore the experiments. Fruits whih had diameters of around 6m were not hemially or physially treated postharvest. Freezing onditions for sweet lemon was simulated in the laoratory using a ooled inuator (Model Panasoni MIR-15-PE). Statistial evaluation of the results was performed using a ompletely randomized fatorial design with 5 repliations for eah treatment. The independent variales inluded temperature, hill duration and tempering period (holding time at room temperature 5± C after treatment) the levels of whih are shown in Tale 1. Tale 1: Levels of independent variales in frost damage investigation. Treatment 1 3 Temperature ( C) A 7-7 Chill duration (hours) B 6 1 Tempering time (hours) C 1 36 Fig.1. (A) light green spot pattern visile in freeze-damaged sweet lemons 1 hours after thawing when viewed under 365nm illumination, (B) Automatially segmented image of sweet lemon in part showing light green spot extration (green spot differentiation y image proessing). In order to investigate frost damage using the mahine vision system, sweet lemon fruits were digitally imaged aording to Tale 1. To do this, an imaging platform and an illuminating system were employed. The fruit was loated in the middle of the light ox diretly under amera. The illuminating system inludes two ultraviolet radiation lamps (365 nm, Blak light, model TL 8W 18 Life Siene Informatis Puliation All rights reserved 18 July - August RJLBPCS () Page No.1

Aedi Firouzjaei et al RJLBPCS 18 www.rjlps.om Life Siene Informatis Puliations BLB 1FM, Philips Poland) whih are loated at orners symmetrially. The images were aptured using a digital amera (Canon powershot-sx3 IS, 1.1 Megapixel sensor with DIGIC proessor) and sent to a personal omputer for proessing and reording the data. Sweet lemons were then evaluated for freeze damage using the onventional segment ut method y 3 inspetors and lassified into quality groups [1]. At the time of inspetion, the inspetors were not aware of the type of hill treatment that lemons were sujeted to. In this study, investigation of the digital image features and optimization of their quantities were done using the image proessing tool in MATLAB analytial software. Then, the image of eah sweet lemon fruit was investigated y MATLAB software and the damaged fruits were separated from the healthy ones ased on the presene or asene of green spots on fruit skin. In addition, the extent of damage was designated y alulating a new indiator alled the Green Spot Index or GSI (ratio of the green spot area to the total surfae area) of the image. The system auray and its proper funtioning were determined ased on the apaility of preise disrimination etween healthy and freeze-damaged fruits. Comparison of the means was performed using Dunan s multiple range test (at 99% onfidene level). All the analyses were arried out using SAS ver. 9.1. 3. RESULTS AND DISCUSSION Using image proessing tehnique, GSI was alulated after separating the ojet from the akground image and deteting the light green spots on the sweet lemon surfae (Fig.). Fig.. Results of image proessing using MATLAB software Aording to the results of UV radiation and image proessing, all samples were lassified ased on visual surfae damage sore and GSI given y mahine vision automatially into groups: 1- no damage, - slight damage, 3- moderate damage and - severe damage. The -hour freeze-treated sweet lemons at -7 C and tempering time of 36 hours were lassified as frozen (severe damage) aording to USDA segment ut method. Moreover, almost no green spots were oserved on sweet lemons refrigerated at 7 C for whih GSI was equal to zero. In order to investigate the effets of 18 Life Siene Informatis Puliation All rights reserved 18 July - August RJLBPCS () Page No.13

Aedi Firouzjaei et al RJLBPCS 18 www.rjlps.om Life Siene Informatis Puliations temperature, hilling duration and tempering time on GSI, the data aquired y image proessing were analyzed ased on ompletely randomized fatorial design. Tale : Results of analysis of variane for green spot perentage on sweet lemon as affeted surfae y the independent variales. Soure variations Temperature (A) Chill duration (B) Tempering time (C) interation (A*B) interation (A*C) interation (B*C) interation (A*B*C) Error DF 8 18 Mean Square 65.398** 1619.859** 67.99** 167.789** 671.79** 138.956** 137.99** 15.6 **: signifiant at the%1 level Results of the analysis of variane (Tale ) showed that the perentage of green spots on the fruit surfae is signifiant for all the main effets, doule and triple interation effets (α = %1). Comparison of the means for GSI due to doule and triple interation effets are shown in figures 3,, 5 and 6. Here, different levels of temperature (A), hill duration (B) and tempering period (C) are aording to Tale 1. Comparison of the means using Dunan s multiple range tests showed that y dereasing the temperature from 7 to -7 C, the perentage of green spots on fruit skin representing the extent of frost damage would inrease dramatially (Tale 3). Tale 3: Comparison of the means for GSI on sweet lemon surfae (Dunan s multiple range test). Temperature ( C) 7-7 % green spots.39.11 a Chill duration(hours) 6 1.7 3.8 13.61 a Tempering time (hours) 1 36 3. 6.7 1.86 a Mean values with the same letter are not signifiantly different (α =.1). 18 Life Siene Informatis Puliation All rights reserved 18 July - August RJLBPCS () Page No.1

% green spots % green spots % green spots Aedi Firouzjaei et al RJLBPCS 18 www.rjlps.om Life Siene Informatis Puliations 5 a 35 3 5 15 1 5 Fig.3. Comparison of the means for interation effets of temperature and hill duration (A B). 35 a 3 5 15 1 5 d d d d d d Fig.. Comparison of the means for interation effets of temperature and tempering time (A C). 5 a 15 1 5 d d d d Fig.5. Comparison of the means for interation effets of hill duration and tempering time (B C). Freezing auses oil glands (and eventually the itrus peel to rupture) leading to the formation of watery spots on the fruit peel surfae. These spots spread overtime. The presene of fluoresene material in the peel s oil is responsile for appearane of green spots under UV radiation. At 7 C, 18 Life Siene Informatis Puliation All rights reserved 18 July - August RJLBPCS () Page No.15

% green spots Aedi Firouzjaei et al RJLBPCS 18 www.rjlps.om Life Siene Informatis Puliations GSI (under all hill durations and tempering periods) was equal to zero and there was no signifiant effet at this temperature (Fig.6). In other words, aove zero temperatures whih do not ause freezing, naturally do not lead to frost damage. In frost ondition, the perentage of green spot on lemon peel inreased with hill duration and tempering period in whih a signifiant effet was oserved (Fig.6). 7 a 6 5 3 d d d 1 e e e e e e e e e e e e e e e e e e e e e Fig.6. Comparison of the means for triple interation effets of temperature, hill duration and tempering time (A B C). Diagnosti analysis was done ased on the perentage of green spots on fruit surfae in order to develop a preditive model from group memerships. Tale : The results of disriminant analysis Original Predited Group Memership Groups 1 3 Total 1 9 9 13 15 3 3 9 1 5 13 18 1 1 1 13.3 86.7 1 3 5 75 1 7.8 7. 1 Aording to Tale, it was found that 1% of the undamaged sweet lemons have een plaed in the no-damage group. Two, three and five lemons with slight, moderate and severe damage were lassified as no-damage, slight and severe damage, respetively. Damage lassifiation auray is 18 Life Siene Informatis Puliation All rights reserved 18 July - August RJLBPCS () Page No.16

Aedi Firouzjaei et al RJLBPCS 18 www.rjlps.om Life Siene Informatis Puliations presented in Tale as a perentage. Finally, y investigating the data and ased on the otained oeffiients, estimation auray of the frost damage level ased on the perentage of green spots deteted y image proessing was determined to e equal to 9.6%. Compared to non-destrutive methods for frost damage detetion suh as hromatography [11] and ethanol sensor for measuring the ethanol level of oranges [1], making use of image proessing and UV illumination with the aility of real-time monitoring is a suitale method for frost damage detetion in sorting and grading lines of itrus pakaging. This is while, hromatography and ethanol sensor methods are only appropriate for the evaluation of small groups. Another advantage of image proessing tehnique and UV illumination (ompared to other methods) is that they are fast and ost-effetive suh that itrus damage detetion monitoring an e implemented in a very short time.. CONCLUSION The light green spot pattern visile on the surfae of sweet lemons when illuminated y long wave UV radiation (365nm) an e utilized to identify slight to severe levels of frost damage. Results of the analysis of variane showed that the perentage of green spots on fruit peel surfae under main, doule and triple interation effets is very signifiant (α = %1). Reduing the temperature to -7 C, aused the numer of green spots to inrease signifiantly. By inreasing the hill duration from 6 to hours and tempering period from 1 to 36 hours, green spot density inreased dramatially. Based on diagnosti analysis, a 9.6% predition auray of frost damage level ased on the perentage of green spots on sweet lemon surfae was otained using UV image proessing tehnique. 5. CONFLICT OF INTEREST There is no onflit of interest REFERENCES 1. Olsen M, Matheron M, MClure M, Xiong Z. Diseases of Citrus in Arizona. Plant Disease Puliations, Cooperative Extension, College of Agriulture & Life Sienes, University of Arizona, Tuson, AZ... Timmer L.W, Garnsey S.M. Broadent P. Diseasesof itrus. In: Diseases of Tropial Fruit Crops (ed. R.C. Ploetz), 3; pp. 197 6. 3. Iglesias D.J, Cerós M, Colmenero-Flores J.M, Naranjo M.A, Ríos G, Carrera E, Ruiz-Rivero O, Lliso I, Morillon R, Tadeo F.R. Talon M. Physiology of itrus fruiting. Braz. J. of Plant Physiol. 7; 19: 333 36.. Gamhir P.N, Choi Y.J, Slaughter D.C, Thompson J.F, MCarthy M.J, Proton spin spin relaxation time of peel and flesh of navel orange varieties exposed to freezing temperature. J. Si. Food Agri. 5; 85: 8 86. 5. Zhang L, MCarthy M.J. NMR Relaxometry Study of Development of Freeze Damage in Mandarin Orange. J Si Food Agri. 16; 96(9):3133-9. 18 Life Siene Informatis Puliation All rights reserved 18 July - August RJLBPCS () Page No.17

Aedi Firouzjaei et al RJLBPCS 18 www.rjlps.om Life Siene Informatis Puliations 6. Pavia D.L, Lampman, G, Kriz G.S, Vayvyan J.R. Introdution to Spetrosopy. Western Washington Univerity. United States of Ameria. 9. 7. Syvertsen J.P. Dehydration of freeze-damaged oranges. Hort Siene. 198; 17: 83 8. 8. Zhao X, Burks T, Qin J, Ritenour M. Digital mirosopi imaging for itrus peel disease lassifiation using olor texture features. Applied Engineering in Agriulture. 9; 5(5):769-776. 9. Davis S, Alrigo G. Citrus. Crop Prodution Siene in Hortiulture Series. CAB. International, Wallingford, UK. 199. 1. USDA. Arizona California Citrus Loss Adjustment Standards Handook. FCIC-5 (11-1999). USDA, Washington, DC. 1999. 11. Oenland D.M, Aung L.H, Bridges D.L, Makey B.E. Volatile emissions of navel oranges as preditors of freeze damage. J. Agri. Food Chem. 3; 51: 3367 3371. 1. Tan E.S, Slaughter D.C, Thompson J.F. Freeze damage detetion in oranges using gas sensors. Postharvest Biol. Tehnol. 5; 35: 177 18. 13. Thompson J.F, Slaughter D.C, MCarthy M.J. Survey of Sensing Methods for Detetion of Freeze Damage-Annual Progress Report for Projet CRB No. 56-17. California Citrus Researh Board, Visalia, CA. 6. 1. Thompson J.F, Slaughter D.C. Survey of Sensing Methods for Detetion of Freeze Damage- Annual Progress Report for Projet CRB No. 56-17. California Citrus Researh Board, Visalia, CA. 5. 15. Gamhir N, Choi J, Slaughter C, Thompson F, MCarthy J. Proton spin spin relaxation time of peel and flesh of navel orange varieties exposed to freezing temperature. J. Si. Food Agri. 5; 85: 8 86. 16. Natale C.D, Maagnano A, Martinelli E, Paolesse R, Proietti E, D Amio A. The evaluation of quality of post-harvest oranges and apples y means of an eletroni nose. Sensors and Atuators B. 1; 78: 6 31. 17. Slaughter D.C, Oenland J.F, Thompson M.L, Arpaia D.A. Non-destrutive freeze damage detetion in oranges using mahine vision and ultraviolet fluoresene. Postharvest Biology and Tehnology. 8; 8(3): 31-36. 18. Momin Adul M, kondo N, Kuramoto M, Ogawa Y, Investigation exitation wavelength for fluoresene emission of itrus peel ased on uv-vis spetra, J. of EAEF. 1; 5():16-13. 19. Kondo N, Kuramoto M, Shimizu H, Ogawa Y, Kurita M, Nishizu T, Chong V.K, Yamamoto K. Identifiation of Fluoresent Sustane in Mandarin Orange Skin for Mahine Vision System to Detet Rotten Citrus Fruits. J. of EAEF. 9; : 5-59.. Aleixos N, Blaso J, Navarron F, Molto E. Multispetral inspetion of itrus in real-time using mahine vision and digital signal proessors, J. of Comput. Eletron. Agri. ; 33: 11-137. 18 Life Siene Informatis Puliation All rights reserved 18 July - August RJLBPCS () Page No.18