Comparison of Maturity Detection of Ataulfo Mangoes Using Thermal Imaging and NIR

Transcription

1 Comparison of Maturity Detection of Ataulfo Mangoes Using Thermal Imaging and NIR Federico Hahn, Guadalupe Hernandez Universidad Autónoma Chapingo, Chapingo, México POBox 66, km 38.5 Carr México Texcoco, Chapingo, Texcoco Mexico Mexico has become the main exporter of mangoes in the world as fresh fruit, and the newest variety is Ataulfo, having the advantage of a large shelf life. It is an advantage during fruit managing, but can be a disadvantage when fruit is being processed. The fruit gets a yellow colour, but stays green inside for 15 days. A study was carried out to detect the different maturity stages using an infrared imaging camera and a spectral system. The later worked well and could predict the maturity stage with an accuracy of 93%. The infrared camera was not as accurate with fruits collected daily as the fruit temperature was similar and changed only at the first and last days. Keywords: Spectral measurements, infrared imaging, mango maturity 1. Introduction Mexico grows mangoes in more than 164,582 hectares throughout all the country, producing 1, 118, 253 tons every year. In 1995, Mexico exported 29,009,784 ten pound boxes mainly to the United States of America (Ontiveros et al., 1996). Quality export fruit must meet special quality factors. Therefore, mangos packed within the same box should have a similar ripeness time, so that their shelf life would be similar. Gunasekaran et al. (1985) reviewed nondestructive optical sensors for agricultural applications. Surface color has been a very useful maturity indicator and has been used as an important criterion for grading and harvesting fruits as tomato, peaches, citrus and bananas (Chen et al., 1990; Davis and Gardner, 1994). Spectral imaging involves taking images of diffusely reflected light at one or more wavelengths with relatively narrow bandpass filters (Chao et al., 1999). NIR (near infrared reflectance) radiation scarcely penetrates fruits with thick skin, so internal quality prediction becomes difficult, Bellon (1990). Lu and Chen (1998) reviewed the technical features of hyper spectral imaging and its potential use in food quality and safety inspection. Choi et al., (1995) developed a maturity index for classifying tomato in six maturity stages using a vision system. Maturity tomato sorting by hyper spectral imaging presented a reduction on the classification error of individual pixels over standard RGB imaging (Polder et al., 2000). Multispectral imaging could predict tomatoes which would never mature on packinghouses (Hahn, 2002). Maturity quality indicators of mango fruits are size, pulp and peel color, firmness, total soluble solid content (SSC) and titratable acidity, (Báez and Bringas, 1995). Reduction on firmness and titratable acidity are present during fruit ripening, as well as an increase on total soluble solids. Pulp and peel color change is one of the most important characteristic related to maturity changing from green toward yellow and red. Mango sweetness was determined nondestructively using near infrared reflectance (NIR) by Yantarasri, et al. (1996). Hahn (2000) carried out the first attempt for predicting maturity in mangoes correlating hyper spectral data with firmness and color Thermal imaging has not been used broadly in the food industry to detect quality. It has been used as a size indicator in computerized packinghouses as the fruit and transporting band temperature difference is huge. Thermal sizing avoids the problems caused by illumination shading in machine vision systems. This work studied the spectral reflectance signatures of Ataulfo mangoes during maturity and correlated the spectral data with the soluble solid contents SSC of the mango pulp. Another new approach using a thermal imaging camera was tested using two different approaches. A success rate of 93% was obtained in detecting 229

2 mangoes with different maturity using the NIR. The mangoes tested with the thermal imager presented also a high discriminant value. 2 Materials and methods Ataulfo mangoes (Mangifera indica) were sampled randomly from a commercial packinghouse at San Luis la Loma, Guerrero, Mexico. In order to study reflectance signature changes fruits were stored during thirteen days at 12 o C and analyzed every daily. Two different tests were carried out: 2.1 NIR analysis NIR (near infrared spectroscopy) thermal imaging. An Ocean Optic PC 1000 computerized spectrophotometer working in the nm range was used to obtain the mango spectral reflectance signatures with a resolution of 5 nm. Four 100 W incandescent lamps distributed at the bottom of the plate, were used as light source and a lens collected the reflected radiation from the fruit. The radiation was concentrated towards an optical fiber, which sent the information towards the monochromator. The portable spectrometer acquired five spectral reflectance measurements around the mango tissue at its equatorial zone and five measurements at the fruit apex or fruit nose with the use of an optic fiber. The data was averaged to obtain the final spectral reflectance signature, which represented each zone for each fruit. Fig.1. Mango spectral measurement SSC BRIX measurements were obtained using a manual TECHNIKA refractometer and classified in three groups for further analysis: low SSC ( BRIX), medium SSC (16-18 BRIX) and high SSC ( BRIX). Statistical analysis used the NCSS (North Carolina Statistical Software) to analyze the data through discriminant analysis in order to provide the wavelength or set of wavelengths which best predicted the maturity status of the fruit. Additionally a total of 300 spectra were used to classify the discriminant trainer, composed of 3 groups with 100 spectra each. Each of these groups corresponded to a different maturity status as specified by the BRIX measurements. A training set of 150 spectra was used to validate the trainer accuracy. The maturity index (accuracy) indicated the percentage of proper sorting mangoes having different storage days. Discriminant analysis was also carried out to study the effect of peel color changes on firmness and internal SSC. 2.2 Thermal imaging 230

3 Thermal imaging is a non-destructive infrared sensing technique that measures infrared energy emitted from the object surface. The detected energy is converted by the camera into a thermal map called a thermogram. Unlike NIR, thermal imaging does not require an illumination source for spectral reflectance, which can be affected by the varied skin color of mango fruits or by the applied illumination. Each mango was then thermally imaged using a radiometric 500D infrared camera (Raytheon Commercial Infrared, Dallas, USA) using two different heating treatments. The infrared camera detector presents a pixel array of 320X240 with a 7-14 µm spectral band. The camera emissivity was fixed to 0.85 and the temperature range from o to 32ºC. Prior to the experiment the mangoes were refrigerated (3 C) for at least 3 hours. Each fruit was imaged for 3 minutes and the images saved every 30 seconds while heat was applied to the ambient air at 35 C, circulating through the wind tunnel before being extracted (Fig. 2). A hole needs to be present on the wind tunnel so that the infrared camera can obtain the radiated heat. Figure 2 shows the experiment assembly and the hypothesis to be tested was that green fruits and ripe fruits present a different heating constant. The experiment was carried with unripe and ripe fruits corresponding to low and high SSC, respectively. A total of 10 experiments were carried out. Two different analyses were carried out; the first one began from three degrees and was heated continuously for three minutes. The temperature increase was then analyzed studying the colors present at the peel and bone. Although the camera was positioned always at the same distance (55 cm) it is necessary to measure the mango width and length as well as the bone parameters. In this way the measurement can be made nondependent on the fruit size. The second analysis heated the fruit until 26ºC and then every fifteen seconds a measurement was taken while it was getting cool. extractor movable door wind tunnel Heater focused area infrared camara Fig.2. Infrared camera setup for heating the samples. 3 Results The averaged results (Fig. 3) presented a reflectance difference at 660 nm as the tissue at the apex was firmer and greener than the tissue at the equatorial zone. Differences can be attributed to the closeness of the hard shell to the apex, where an average of only 1 cm of pulp is present, as compared with the 3 cm found at the equatorial zone. The reflectance magnitude throughout all the spectra was higher at the equatorial zone than at the apex. The signature pattern showed differences in shape, having the apex two reflectance peaks registered at 715 nm and 560 nm, while the equatorial zone spectrum showed three peaks at 715 nm, 625 nm and 580 nm. Mango apex spectral signatures acquired every two days after being picked from the packinghouse presented in 715 nm its best discriminant wavelength (Fig. 4). The apex spectral reflectance decreased daily; although at 715 nm was not completely linear, showing the highest differences between the first and the second day. An average relative reflectance reduction of 6.43 was noted for mangoes kept at 12 C. Discriminant analysis showed that the best discriminant analysis were 715, 580 and 625 nm, Table.1. The SSC accuracy was best for all the cases and a 231

4 maximum of 93% was achieved. Best discriminant accuracies were obtained at the apex, being the model capable of predicting the storage day with a success rate of 93 %. Table 1. Accuracy of the fruits according to SSC group or daily maturity for each wavelength set. Wavelength set SSC group discriminant accuracy (%) Daily discriminant accuracy (%) 715, 580, /400, 580/ /400, 625/ Reflectance (%) apex equatorial zone Wavelength (nm) Fig. 3. Difference between apex and equatorial zone reflectance Fig. 4. Changes in spectral reflectance of mangos as affected by ripening at 12 C Fig. 5. Green and ripe Ataulfo mangoes positioned at its front and lateral views before being imaged by the thermal infrared camara. 232

5 (c) (d) Fig. 6. Thermal images acquired from a ripe (right) and a green mango (left) viewed in the frontal position at a delay of 0 minutes, one minute; (c) two minutes and (d) three minutes. (c) (d) Fig. 7. Thermal images acquired from a ripe (right) and a green mango (left) viewed in the lateral position at a delay of 0 minutes, one minute; (c) two minutes and (d) three minutes. The thermal infrared images can be seen on figures 6, 7 and 8. Figure 6 shows the frontal mango images taken at 3ºC (Fig. 6a), after one, two and three minutes. Mango frontal position can be viewed on Fig. 5.a. The bone is always the coolest point of the image, and after three minutes the bone image in not present anymore. After one minute after counting the orange pixels it can be noted that the bone of the unripe mango does not looses heat as the ripe one. The reason is that the pulp is harder and less air spaces are present. Figure 7 shows the 233

6 mango images taken on a lateral view and the image of the unripe fruit on the left (Fig. 7b) behaves similarly than on Fig. 7d, taken 2 minutes later. It is obvious that the frontal experiment works much better. (c) (d) Fig. 8. Thermal images acquired from a ripe (right) and a green mango (left) viewed in the lateral position at a delay of fifteen; thirteen; (c) forty five and (d) sixty seconds after being heated. The second analysis considered that an initial temperature of 26ºC was applied to the mango skin. The fruit was then left without any heating and the fruit began to return to its initial temperature which was of 3ºC. Both on Fig. 8b and Fig. 8c the orange pixels increased quicker on the ripe fruit which losses heat quicker than the green fruit. 4 Conclusions The results withdrawn from the present work are: 1. NIR is the best option for detecting maturity 2. Maturity detected as BRIX level and grouped in three different categories achieved the highest success rate of 93%. 3. Maturity detected by days after harvesting presented accuracies over 80%. 4. Infrared imaging can be used but the parameters have to be calibrated and an imaging software generated. 5 References Báez-Sañudo, R., Bringas-Taddei, E., Mexican fresh mango quality standard grades and application methodology. Proc. Interamer. Soc. Trop. Hort. 39, Bellon, V., Feasibility of new types of non-contact evaluation of internal quality inspection by spectroscopic techniques: Near infrared, nuclear magnetic resonance. Proceedings of International Conference on Agricultural Mechanization. Zaragoza, Spain. pp Chao, K., Park, B., Chen, Y.R., Hruschka, W.R., (1999). Chicken heart diseases characterization by multi-spectral imaging. ASAE Paper No Chen, S., Fon, D., Chang Ch., Universal Color indices for maturity evaluations of fruits. Paper ASAE No Davis, J.M., Gardner, R.G., Harvest maturity affects fruit yield, size and grade of fresh-market tomato cultivars. HortScience, 29(6),

7 Gunasekaran, S., Paulsen, M.R., Shove G.C., Optical methods for nondestructive quality evaluation of agricultural and biological materials. Journal of Agric. Engr. Res. 32, Hahn, F., Multispectral prediction of unripe tomatoes. Biosystems Engineering 81(2), Hahn, F., Hyperspectral mango ripeness detection. ASAE Paper No Lu, R., Chen, Y.R., Hyperspectral imaging for safety inspection of food and agricultural products. In Pathogen detection and remediation for safe eating, Proceedings of the International Society for Optical Engineering, 3544: Ontiveros, N., Avena-B, S., Ponce de Leon, G., Bosquez, M., Cruz, G., Baez-Sañudo, L.A., th. International Mango Symposium. Tel Aviv, Israel (Abstract) pp. 33. Yantarasri, T., Kawano S.P., Chen, Somsrivichai, J., Nondestructive tests for Asian mangoes. 5th. International Mango Symposium. Tel Aviv, Israel (Abstract) pp