Final Report Summary - FRUITGRADING (A LOW COST SORTING SOLUTION FOR THE FRUIT SECTOR BASED ON THE EVALUATION OF INTERNAL FRUIT QUALITY)
Executive Summary:
The Fresh Fruit and Vegetable (FFV) sector plays a very important role in the economy of Europe in terms of labour and output value, as well as in the development of rural areas. In value terms, FFV accounts for a 6.8% share of the EU-27 overall agricultural industry output (more than €23.4b at producer prices) and employment, in full time equivalents, is estimated at 1.3 million workers. Fruit and vegetable holdings in all Member States are typically small and medium farms (SMEs), on average less than 10 hectares.
The popularity of Fresh Fruit and Vegetables has increased in recent years as healthy and nutritious foods. Producers, retailers and distributors of fresh fruit need to anticipate consumer desires particularly with regard to sensory quality and convenience. Consumers consider good quality fruits to be those that look good, are firm and offer good flavour and nutritional value. However, consumers are increasingly interested in qualities that are not necessarily related to external appearance. Consumer acceptance of top quality class fruit (Extra class) is manifested by the fact that over 60% of the total fruit produced by the SME-AG proposers corresponds to First and Extra Class fruit , a segment that reaches a significant price difference at origin (up to 100% more than class II).
While good harvesting practices result in better quality fruit, it is clear that fruit is highly perishable and susceptible to damage from unforeseen climatic conditions. Consequently, fruit farmers often deal with very heterogeneous products in terms of quality. Under these conditions, segmentation of the production is an effective approach to increase the overall value of the production. In this sense, fruit sorting is already used to maximize production income. Nevertheless, fruit sorting is currently performed based on morphological and basic physical characteristics (size, shape, and weight), and external appearance (colour, presence of surface defects, etc).
Unfortunately, these parameters are not totally representative of the produce quality, which encompasses not only external appearance, but also other sensory properties (taste and aroma), nutritional value, chemical constituents, mechanical properties, functional properties and defects. This makes current fruit sorting technology unsuitable for properly classifying fruit pieces according to quality criteria closer to those perceived by European consumers.
The key objective of this project was to develop a low cost fruit sorting prototype module able to classify fruit on an individual basis, based on internal quality criteria, achieving sorting speeds of 5 pieces per second. The module was integrated, calibrated and tested in a customized commercial sorting line. The technology was tested for those cases with a more significant impact among the SME-AG associates: apple, pear, peach, and kiwi.
Project Context and Objectives:
During the first reporting period (M1-M15), efforts have been oriented towards the characterization of fruits at the laboratory level, the development of correlation models for the determination of internal quality parameters of fruit (sugar, acidity, and firmness), and the definition of the specifications for the development of the fruit sorting prototype.
Systematic optical and electromagnetic characterization experiments were conducted at the laboratory using sets of different varieties of apples (bravo, fuji, gala, idared, ligol, sampion), pears (abat, blanquilla, conference, limonera, louise, moretini, rocha, Williams), peaches (albared, bigtop, luciana, miraflores, summer rich, tardibell, rommer star, sweet dream, venus), and kiwis (earligreen, hayward, soreli). The total number of samples analyzed was 378 (peaches), 236 (pears), 135 (kiwi), and 350 (apples). For each sample, reference laboratory methods were used for the accurate determination of target parameters, namely refractometry (SCC), titrimetry (acidity), and penetrometry in the flesh after peeling (firmness).
Based on the experimental results obtained, correlation models were built by means of different statistical and spectral processing techniques, including Partial Least Squares (PLS), k-Nearest Neighbors (KNN), and Radial Basis Function (RBF). In order to validate the correlation models developed, the initial data was randomly divided into three subsets, which were used for the calibration (50%), validation (25%), and test (25%) of the models.
It was found that it is possible to accurately predict sugar (“Soluble Solid Content”, SCC), and acidity (“Titrable Acidity”) based on the reflectance spectra obtained. Root Mean Square Errors (RMSE) obtained from the test of the models were typically below 1º BRIX (sugar content), and 1 meq/100g (acidity). In the case of firmness, being a mechanical parameter relatively independent of fruit composition, the correlation obtained was moderate. Still, good results were obtained in the case of apples.
Interestingly, the correlations found for each fruit variety could be consistently applied when considering together different varieties of the same fruit. In the case of firmness, it is unclear the correlation between this parameter and the spectra obtained. The correlation analysis performed has allowed optimizing the design of the optical instrument that will be developed during the second phase of the project.
In addition to optical characterization, a novel contactless electromagnetic technique was also used to investigate the properties of fruits. Two specific laboratory set-ups, together with the required signal processing and acquisition electronics, have been developed for this purpose. In order to enhance the understanding of the electromagnetic response of fruit tissue, as well as to optimize the experimental configuration used, Finite-Element Modeling and simulations were performed.
Based on the developed set-ups, the electromagnetic response of different fruit samples was analyzed, showing that the radiofrequency spectrum of fruits evolves during the ripening process of fruit pieces. Moreover, changes in fruit firmness resulting from post-harvest maturation could also be related to changes in the electromagnetic spectrum. The most sensitive spectral region to monitor these changes was also identified. Nevertheless, additional efforts are still required in order to optimize the existing instrument, and to improve the correlation models obtained. To achieve this goal, it is necessary to further reduce sensor noise, and to minimize the influence of error sources.
On the other hand, in parallel with technical tasks, exploitation, dissemination, and training actions have been also encouraged since the beginning of the project
During the second reporting period (M16 – M36), efforts have been focussed on the calibration of the MIS and the NIR sensors and its integration into a fruit sorting line. Dissemination activities were taken by the SME-AGs in order to promote the project and its results among its associates. Training activities were performed by the RTDs for a transfer of technology among the SME-AGs. Next tasks provided a more detailed explanation of the work performed during the second period (M16-M36):
• New ruggedized MIS system for on-line use. During the second period of the FRUITGRADING project, UNIMAN had to develop a complete dedicated MIS machine from scratch in order to overcome the limitations of the two earlier laboratory systems. The system could operate a conveyor speeds with a single pass of the fruit with sufficient signal to noise ratio to resolve well the conductivity spectra over the range 160 kHz to 2.5 MHz.
• Review and Calibration of the LEDA-DRS sensor by the NIR sensors. ATEKNEA performed a review of the LEDA-DRS concept design including modification such as additional optics, temperature stabilization and InGaAS PDs. For the sensor calibration, two sets of measurements were performed: December 2013 and January 2014. During the first tests, 54 Fuji apples were analyzed and during the second one 66 conference pears.
• Introduction and calibration of the NIR sensor. In December 2013, ATEKNEA replaced the LEDA-DRS sensor by the continuous NIR sensor, which is a very fast measuring device. The new sensor is able to acquire the whole data necessary to predict the sugar content in 1 ms. During December 2013, a set of 54 apples were used for calibration of the continuous spectrum NIR sensor. The apples were processed and their sugar content was measured using a refractometer ATAGO α-101.
• Integration of the MIS and NIR sensors into an industrial fruit sorting line. Once the industrial prototypes were calibrated, STC and ATEKNEA lead the task of integration in a fruit sorting line in order to establish the sorting capabilities of these new measurements systems.
• Test and industrial validation of the MIS and NIR sensors. Validation tests were performed in OWOC and SOTO premises. During the tests the MIS and NIR sensors were calibrated for obtaining better fruit properties correlations.
• Training activities. A training plan to ensure that the lead-users, end-user SMEs and SME-AGs assimilate the project result including the technical, socio-economic and commercial benefits of the FRUITGRADING technology were created. All members of the consortium were actively involved under the coordination of the AJAP association and assisted by ATEKNEA and the representatives from the other SME-AGs
• Dissemination activities. The project website (www.fruitgrading.eu) was created and it will serve as vehicle of communication and dissemination of the project and its results. All consortium members have been actively involved in dissemination activities, like articles in magazines, newsletters to different private and public entities, and participation in conferences and workshops. All members of the consortium also created a Plan for Use and Dissemination of Foreground generated during the project.
Project Results:
The main objective of this project was to develop a low cost fruit sorting prototype module able to classify fruit on an individual basis, based on internal quality criteria, achieving sorting speeds of 5 pieces per second. This module was integrated, calibrated and tested in a commercial sorting line. The technology was validated for those cases with a more significant impact among the SME-AG associates: apple, pear, peach, and kiwi. To achieve this overall goal, the project aims to achieve different technical and operational objectives, which were achieved within the 36-month-time-frame.
The project list of objectives is presented per WP:
• WP1.- Characterization of fruits by Magnetic Induction Spectroscopy. During the task of characterizing the dielectric properties of the different fruits the next S&T results were obtained:
o Frequency-dependent electrical and magnetic properties of fruits obtained.
o No published data currently jeopardizes the exploitation of the technologies proposed in the project.
o The combined simulation tests on the axial gradiometer indicate that the conductivity of the fruit (which is derived from the measurement of the imaginary part of the mutual coupling) over the full spectrum tested can be extracted. Fruit permittivity and permeability, both affecting the real part of the complex mutual coupling, cannot be separated. However, simulations indicate that the diamagnetic properties of fruit dominate the response of the in-phase component of the spectrum over the complete frequency range. Based on this result, the MIS spectrum is found to be sensitive to changes in fruit firmness, which are linked to changes to the cellular structure of fruits, thus leading to measurable changes of their frequency dependent electrical conductivity.
o Simulation models have been also used to define optimized geometrical parameters for the implementation of a gradiometer system to inspect fruit. Based on these models, technical recommendations for the implementation of the laboratory and prototype sorting system were provided. In addition to geometrical parameters, important aspects to be taken into account during the implementation of the design are (1) the need for electric field screening of the coils, to prevent capacitive coupling between the emitting and receiving elements; (2) the sensitivity of the response of the instrument to temperature changes in the fruit, for which it is recommended to take into account this parameter in the model; and (3) in the case of anisotropic fruits (particularly kiwi), orientation of the sample needs to be taken into account.
o Laboratory set-ups for investigating the relation between the MIS signal and internal quality parameters of fruit samples have been successfully produced. These systems have proved sensitive enough to sense variations in the electrical properties of fruits related to substantial changes in their internal properties. In particular, the resistive component of fruits tends to increase during post-harvest maturation. Likewise, the resistive component of fruits changes notably after frozen/unfrozen process, which modifies fruit firmness, and its internal cellular tissue structure.
o Correlation models between MIS spectra and fruit parameters (firmness, TA, and SCC) were produced based on the experimental results
• WP2.- Characterization of fruits by Diffuse Reflectance Spectroscopy. During the tasks of investigating the correlation between the reflectance spectra of different fruits over a wide spectral range with their physic-chemical characteristics, the next S&T results were obtained:
o Spectral dataset obtained to be used in the development of correlation models between the spectra and fruit internal quality parameters. The spectral dataset obtained was found to be large enough to develop reliable correlation models between the LEDA-DRS spectra, and fruit internal quality parameters. Moreover, the spectral dataset was also adequate for the modeling of the response of a virtual multiwavelength instrument, and to define its technical specifications.
o It was found that it is possible to accurately predict sugar (“Soluble Solid Content”, SCC), and in certain fruit varieties acidity (“Titrable Acidity”) based on the LEDA-DRS reflectance spectra obtained. Root Mean Square Errors (RMSE) obtained from the test of the models were typically below 1º BRIX (sugar content), and 1 meq/100g (acidity).
o Good correlation values were found for all fruit varieties analyzed in the case of the determination of sugar content (SCC), with adjusted coefficient of determination (Adj. R2) values typically in the range of 0.8.
o The prediction of TA was found to be possible indirectly in varieties in which there is an internal correlation between the evolution of acidity and sugar content, such as in the case of pears. In other cases, in which the relation is not so obvious (ex: peaches), the correlation obtained was less satisfactory.
o In the case of firmness, being a mechanical parameter relatively independent of fruit composition, the correlation obtained was moderate. Unexpectedly, good results were obtained in the case of firmness prediction in apples. This is a promising result (Adj.R2 = 0.82 and 0.76 for SCC and firmness, respectively), given the economic relevance of this fruit, and the interest of simultaneously predict sugar content and firmness – the most relevant parameters for end-users, by means of a single sensing unit based on diffuse reflectance spectroscopy.
o Optimum set of central wavelengths was identified. Simulated response of the virtual instrument proves that the expected accuracy of the virtual instrument with at least 7 LEDs is comparable to that of a continuous laboratory instrument. For example, SCC prediction in apples with a 7-LED virtual instrument is nearly the same (or even better) that the one achieved at the laboratory using a 7-parameters PLS model (Adj. R2 = 0.80).
• WP3.- Correlation models and definition of specifications. During the task of improving the quality prediction modules based on the combination of the electrical and optical information of the MIS and NIR sensors, the next S&T results were obtained:
o The inclusion of MIS spectral data has not improved the accuracy of the calibration models based exclusively on LEDA-DRS spectra. This is attributed to the low correlation already found in the MIS spectra dataset.
o In relation to the calibration of the MIS, it has been proposed a calibration transfer procedure based on the normalization of the signals to the values obtained with the laboratory instruments. Calibration transfer procedures have been defined, and are available for the implementation of the calibration models in the industrial prototypes.
o Key specifications for the design of the pilot sorting line, including the geometrical arrangement and dimensions of the coils in the gradiometer, set of operational frequencies (MIS module), minimum distance between fruit pieces to avoid interferences between consecutive MIS measurements, dataset of wavelengths and layout of the LEDA-DRS sensor, the use of auxiliary sensing elements, and the characteristics of the conveying system (dimensions and speed).
• WP4.- Industrial prototype development. During the task of designing a FRUITGRADING prototype that includes both the MIS and the NIR sensors, the next S&T results were obtained:
o Integration of the MIS and NIR hardware modules into a fruit sorting line.
o End of the development of the fruit quality inspection system based on the LEDA-DRS spectroscopy. The LEDA-DRS integrated was able to perform 5 measurements per second.
o End of the development of the Magnetic Induction System (MIS) spectroscopy. The MIS sensor was improved until it reached 10 measurements per second.
o Prototypes (MIS and NIR) integration into an industrial sorting line including the electronics, mechanics and opto-mechanics (for the LEDA-DRS sensor).
o Replacement of the LEDA-DRS sensor by a NIR sensor provided by ATEKNEA which showed better results for sugar content detection in fruits.
o Calibration tests of the MIS and NIR sensors at the laboratory level.
• WP5.- Integration in a fruit sorting line. During the tasks of integrating the MIS and NIR sensors into the FRUITGRADING prototype, the next S&T results were obtained:
o The NIR sensor was integrated into the fruit sorting line provided by STC after electrical, mechanical, optical and software modifications.
o A user interface made by ATEKNEA with the assistant of STC and using LABVIEW V2013 which shows a graph containing the shape of the acquired signal when a fruit is present.
o Deliverable D5.1 that summarizes the work performed for the integration.
• WP6.- Industrial validation. During the task of industrial validation of the FRUITGRADING prototype, the next S&T results were obtained:
o Improvements of the prediction algorithm for producing better correlations between signal measure and fruit properties for the MIS and NIR sensors.
o The NIR sensor provided promising results for measuring the sugar content in pears.
o The MIS sensor provided promising results for measuring the acidity in peaches.
• WP7.- Training. During the task of training, the next results were obtained:
o A training plan was created in order to ensure that the lead-users, end-user SMEs and SME-AGs assimilate the project result including the technical, socio-economic and commercial benefits of the FRUITGRADING technology. The training plan describes the road map performed and all training actions within the FRUITGRADING project. All members of the consortium were actively involved under the coordination of the AJAP association and assisted by ATEKNEA and the representatives from the other SME-AGs.
• WP8.- Dissemination. During the task of dissemination, the next results were obtained:
o The consortium as a result of their actions could clearly disseminate the FRUITGRADING project to the Target Audience, making use of a strong campaign, through various means of promoting technology used. It was reached the goal in order to convey the real benefits of the product, achieving captivate potential interested in learning more about the prototype. The project was disseminated in various trade fairs, as well as the several publication of articles related to FRUITGRADING.
• WP9.- IPR. During the task of protection of the Intellectual Property Rights of the project, a PUDF (Plan for Using and Dissemination of the Foreground) and IPR (Intellectual Property Rights) documents have been created.
Potential Impact:
Fruit distribution in Europe is performed mostly at a regional or national level. Prices are conditioned by different variables related to the quality of the product, and in some cases, by stationary fluctuations of the demand in relation to the product supply. Independently of this, the oligopolistic structure of the fruit market has a strong influence on the fruit prices at origin. In 2002, the market share of the 5 largest retailers (discounts, hypermarkets, etc) in EU-15 reached 69%. The buying power of large retailers is therefore significant, and even more important when considering that large retailers often form buying groups.
In spite of this complex context, there is an increasing market demand for top quality fruit within Europe and the United States, which reflects the growing interest of consumers for high quality fruit. Interestingly, in spite of the decrease in the fruit consumption over the last years, the fresh fruit gourmet market has steadily increased. In the U.S. retail sales of gourmet/premium products increased 55% ($15,000 million) during the 2001-2005 period, with an average rate of 8.9% during the last five years.
Nevertheless, the divergence between the sorting criteria currently used in industry, based on external quality indicators, and the quality criteria perceived by consumers, has an impact on the price of the fruit. SME-AG proposers have estimated that the regular price of the extra class fruit could be further increased between 20%-40%, if the fruit producers could ensure that the fruit meets certain minimum “internal quality” criteria.
The availability of FRUITGRADING technology would lead to a significant economic benefit for over 2,500 European fruit farmers represented by the SME-AG proposers. In particular, the following key competitive advantages may be highlighted:
• The possibility of assigning part of their production to new “premium grade” segments, granting minimum quality requirements according to objective and measurable internal quality indicators. In contrast with existing sorting methods based on the external appearance of fruit, FRUITGRADING would provide an effective means for assessing fruit quality, an sorting fruit accordingly.
• Being a non-destructive analytical method, FRUITGRADING would allow for a substantial reduction of discarded fruit resulting from current practices (laboratory analysis)
• Enabling fruit farmers to design and implement pre-sorting strategies before fruit storage in maturation chambers.
Regarding the dissemination activities, the consortium created a dissemination plan in spreading the FRUITGRADING through Publications, Magazines, Media, Advertising, workshops and fairs and events. As chairman of the Exploitation Board, Urszula Makosz (TRSK), coordinated all the activities encompassing any transfer of knowledge outside the Consortium. All partners took an active role in discussing and deciding the content of the material to be published to ensure that initially no confidential information was disclosed on potentially marketable aspect of the project. Furthermore, aspects dealing with the design of the MIS submodule, LEDA-DRS spectrometer, as well as quality prediction models, was regarded as particularly sensitive, and therefore excluded from any dissemination outside the consortium. However, all partners promoted the widespread diffusion of the generic results in terms of the performance and industrial applicability of the developed technology at a European level, through the channels specified in Deliverable 8.2. The consortium carried out mailing of information about the project to government authorities and interested parties. Information and exhibition stands were set up and manned at the various different identified and selected events. The consortium carried out specific dissemination actions of the FRUITGRADING project in the media and in some events related to the agricultural sector.
List of Websites:
http://www.fruitgrading.eu(si apre in una nuova finestra)
The Fresh Fruit and Vegetable (FFV) sector plays a very important role in the economy of Europe in terms of labour and output value, as well as in the development of rural areas. In value terms, FFV accounts for a 6.8% share of the EU-27 overall agricultural industry output (more than €23.4b at producer prices) and employment, in full time equivalents, is estimated at 1.3 million workers. Fruit and vegetable holdings in all Member States are typically small and medium farms (SMEs), on average less than 10 hectares.
The popularity of Fresh Fruit and Vegetables has increased in recent years as healthy and nutritious foods. Producers, retailers and distributors of fresh fruit need to anticipate consumer desires particularly with regard to sensory quality and convenience. Consumers consider good quality fruits to be those that look good, are firm and offer good flavour and nutritional value. However, consumers are increasingly interested in qualities that are not necessarily related to external appearance. Consumer acceptance of top quality class fruit (Extra class) is manifested by the fact that over 60% of the total fruit produced by the SME-AG proposers corresponds to First and Extra Class fruit , a segment that reaches a significant price difference at origin (up to 100% more than class II).
While good harvesting practices result in better quality fruit, it is clear that fruit is highly perishable and susceptible to damage from unforeseen climatic conditions. Consequently, fruit farmers often deal with very heterogeneous products in terms of quality. Under these conditions, segmentation of the production is an effective approach to increase the overall value of the production. In this sense, fruit sorting is already used to maximize production income. Nevertheless, fruit sorting is currently performed based on morphological and basic physical characteristics (size, shape, and weight), and external appearance (colour, presence of surface defects, etc).
Unfortunately, these parameters are not totally representative of the produce quality, which encompasses not only external appearance, but also other sensory properties (taste and aroma), nutritional value, chemical constituents, mechanical properties, functional properties and defects. This makes current fruit sorting technology unsuitable for properly classifying fruit pieces according to quality criteria closer to those perceived by European consumers.
The key objective of this project was to develop a low cost fruit sorting prototype module able to classify fruit on an individual basis, based on internal quality criteria, achieving sorting speeds of 5 pieces per second. The module was integrated, calibrated and tested in a customized commercial sorting line. The technology was tested for those cases with a more significant impact among the SME-AG associates: apple, pear, peach, and kiwi.
Project Context and Objectives:
During the first reporting period (M1-M15), efforts have been oriented towards the characterization of fruits at the laboratory level, the development of correlation models for the determination of internal quality parameters of fruit (sugar, acidity, and firmness), and the definition of the specifications for the development of the fruit sorting prototype.
Systematic optical and electromagnetic characterization experiments were conducted at the laboratory using sets of different varieties of apples (bravo, fuji, gala, idared, ligol, sampion), pears (abat, blanquilla, conference, limonera, louise, moretini, rocha, Williams), peaches (albared, bigtop, luciana, miraflores, summer rich, tardibell, rommer star, sweet dream, venus), and kiwis (earligreen, hayward, soreli). The total number of samples analyzed was 378 (peaches), 236 (pears), 135 (kiwi), and 350 (apples). For each sample, reference laboratory methods were used for the accurate determination of target parameters, namely refractometry (SCC), titrimetry (acidity), and penetrometry in the flesh after peeling (firmness).
Based on the experimental results obtained, correlation models were built by means of different statistical and spectral processing techniques, including Partial Least Squares (PLS), k-Nearest Neighbors (KNN), and Radial Basis Function (RBF). In order to validate the correlation models developed, the initial data was randomly divided into three subsets, which were used for the calibration (50%), validation (25%), and test (25%) of the models.
It was found that it is possible to accurately predict sugar (“Soluble Solid Content”, SCC), and acidity (“Titrable Acidity”) based on the reflectance spectra obtained. Root Mean Square Errors (RMSE) obtained from the test of the models were typically below 1º BRIX (sugar content), and 1 meq/100g (acidity). In the case of firmness, being a mechanical parameter relatively independent of fruit composition, the correlation obtained was moderate. Still, good results were obtained in the case of apples.
Interestingly, the correlations found for each fruit variety could be consistently applied when considering together different varieties of the same fruit. In the case of firmness, it is unclear the correlation between this parameter and the spectra obtained. The correlation analysis performed has allowed optimizing the design of the optical instrument that will be developed during the second phase of the project.
In addition to optical characterization, a novel contactless electromagnetic technique was also used to investigate the properties of fruits. Two specific laboratory set-ups, together with the required signal processing and acquisition electronics, have been developed for this purpose. In order to enhance the understanding of the electromagnetic response of fruit tissue, as well as to optimize the experimental configuration used, Finite-Element Modeling and simulations were performed.
Based on the developed set-ups, the electromagnetic response of different fruit samples was analyzed, showing that the radiofrequency spectrum of fruits evolves during the ripening process of fruit pieces. Moreover, changes in fruit firmness resulting from post-harvest maturation could also be related to changes in the electromagnetic spectrum. The most sensitive spectral region to monitor these changes was also identified. Nevertheless, additional efforts are still required in order to optimize the existing instrument, and to improve the correlation models obtained. To achieve this goal, it is necessary to further reduce sensor noise, and to minimize the influence of error sources.
On the other hand, in parallel with technical tasks, exploitation, dissemination, and training actions have been also encouraged since the beginning of the project
During the second reporting period (M16 – M36), efforts have been focussed on the calibration of the MIS and the NIR sensors and its integration into a fruit sorting line. Dissemination activities were taken by the SME-AGs in order to promote the project and its results among its associates. Training activities were performed by the RTDs for a transfer of technology among the SME-AGs. Next tasks provided a more detailed explanation of the work performed during the second period (M16-M36):
• New ruggedized MIS system for on-line use. During the second period of the FRUITGRADING project, UNIMAN had to develop a complete dedicated MIS machine from scratch in order to overcome the limitations of the two earlier laboratory systems. The system could operate a conveyor speeds with a single pass of the fruit with sufficient signal to noise ratio to resolve well the conductivity spectra over the range 160 kHz to 2.5 MHz.
• Review and Calibration of the LEDA-DRS sensor by the NIR sensors. ATEKNEA performed a review of the LEDA-DRS concept design including modification such as additional optics, temperature stabilization and InGaAS PDs. For the sensor calibration, two sets of measurements were performed: December 2013 and January 2014. During the first tests, 54 Fuji apples were analyzed and during the second one 66 conference pears.
• Introduction and calibration of the NIR sensor. In December 2013, ATEKNEA replaced the LEDA-DRS sensor by the continuous NIR sensor, which is a very fast measuring device. The new sensor is able to acquire the whole data necessary to predict the sugar content in 1 ms. During December 2013, a set of 54 apples were used for calibration of the continuous spectrum NIR sensor. The apples were processed and their sugar content was measured using a refractometer ATAGO α-101.
• Integration of the MIS and NIR sensors into an industrial fruit sorting line. Once the industrial prototypes were calibrated, STC and ATEKNEA lead the task of integration in a fruit sorting line in order to establish the sorting capabilities of these new measurements systems.
• Test and industrial validation of the MIS and NIR sensors. Validation tests were performed in OWOC and SOTO premises. During the tests the MIS and NIR sensors were calibrated for obtaining better fruit properties correlations.
• Training activities. A training plan to ensure that the lead-users, end-user SMEs and SME-AGs assimilate the project result including the technical, socio-economic and commercial benefits of the FRUITGRADING technology were created. All members of the consortium were actively involved under the coordination of the AJAP association and assisted by ATEKNEA and the representatives from the other SME-AGs
• Dissemination activities. The project website (www.fruitgrading.eu) was created and it will serve as vehicle of communication and dissemination of the project and its results. All consortium members have been actively involved in dissemination activities, like articles in magazines, newsletters to different private and public entities, and participation in conferences and workshops. All members of the consortium also created a Plan for Use and Dissemination of Foreground generated during the project.
Project Results:
The main objective of this project was to develop a low cost fruit sorting prototype module able to classify fruit on an individual basis, based on internal quality criteria, achieving sorting speeds of 5 pieces per second. This module was integrated, calibrated and tested in a commercial sorting line. The technology was validated for those cases with a more significant impact among the SME-AG associates: apple, pear, peach, and kiwi. To achieve this overall goal, the project aims to achieve different technical and operational objectives, which were achieved within the 36-month-time-frame.
The project list of objectives is presented per WP:
• WP1.- Characterization of fruits by Magnetic Induction Spectroscopy. During the task of characterizing the dielectric properties of the different fruits the next S&T results were obtained:
o Frequency-dependent electrical and magnetic properties of fruits obtained.
o No published data currently jeopardizes the exploitation of the technologies proposed in the project.
o The combined simulation tests on the axial gradiometer indicate that the conductivity of the fruit (which is derived from the measurement of the imaginary part of the mutual coupling) over the full spectrum tested can be extracted. Fruit permittivity and permeability, both affecting the real part of the complex mutual coupling, cannot be separated. However, simulations indicate that the diamagnetic properties of fruit dominate the response of the in-phase component of the spectrum over the complete frequency range. Based on this result, the MIS spectrum is found to be sensitive to changes in fruit firmness, which are linked to changes to the cellular structure of fruits, thus leading to measurable changes of their frequency dependent electrical conductivity.
o Simulation models have been also used to define optimized geometrical parameters for the implementation of a gradiometer system to inspect fruit. Based on these models, technical recommendations for the implementation of the laboratory and prototype sorting system were provided. In addition to geometrical parameters, important aspects to be taken into account during the implementation of the design are (1) the need for electric field screening of the coils, to prevent capacitive coupling between the emitting and receiving elements; (2) the sensitivity of the response of the instrument to temperature changes in the fruit, for which it is recommended to take into account this parameter in the model; and (3) in the case of anisotropic fruits (particularly kiwi), orientation of the sample needs to be taken into account.
o Laboratory set-ups for investigating the relation between the MIS signal and internal quality parameters of fruit samples have been successfully produced. These systems have proved sensitive enough to sense variations in the electrical properties of fruits related to substantial changes in their internal properties. In particular, the resistive component of fruits tends to increase during post-harvest maturation. Likewise, the resistive component of fruits changes notably after frozen/unfrozen process, which modifies fruit firmness, and its internal cellular tissue structure.
o Correlation models between MIS spectra and fruit parameters (firmness, TA, and SCC) were produced based on the experimental results
• WP2.- Characterization of fruits by Diffuse Reflectance Spectroscopy. During the tasks of investigating the correlation between the reflectance spectra of different fruits over a wide spectral range with their physic-chemical characteristics, the next S&T results were obtained:
o Spectral dataset obtained to be used in the development of correlation models between the spectra and fruit internal quality parameters. The spectral dataset obtained was found to be large enough to develop reliable correlation models between the LEDA-DRS spectra, and fruit internal quality parameters. Moreover, the spectral dataset was also adequate for the modeling of the response of a virtual multiwavelength instrument, and to define its technical specifications.
o It was found that it is possible to accurately predict sugar (“Soluble Solid Content”, SCC), and in certain fruit varieties acidity (“Titrable Acidity”) based on the LEDA-DRS reflectance spectra obtained. Root Mean Square Errors (RMSE) obtained from the test of the models were typically below 1º BRIX (sugar content), and 1 meq/100g (acidity).
o Good correlation values were found for all fruit varieties analyzed in the case of the determination of sugar content (SCC), with adjusted coefficient of determination (Adj. R2) values typically in the range of 0.8.
o The prediction of TA was found to be possible indirectly in varieties in which there is an internal correlation between the evolution of acidity and sugar content, such as in the case of pears. In other cases, in which the relation is not so obvious (ex: peaches), the correlation obtained was less satisfactory.
o In the case of firmness, being a mechanical parameter relatively independent of fruit composition, the correlation obtained was moderate. Unexpectedly, good results were obtained in the case of firmness prediction in apples. This is a promising result (Adj.R2 = 0.82 and 0.76 for SCC and firmness, respectively), given the economic relevance of this fruit, and the interest of simultaneously predict sugar content and firmness – the most relevant parameters for end-users, by means of a single sensing unit based on diffuse reflectance spectroscopy.
o Optimum set of central wavelengths was identified. Simulated response of the virtual instrument proves that the expected accuracy of the virtual instrument with at least 7 LEDs is comparable to that of a continuous laboratory instrument. For example, SCC prediction in apples with a 7-LED virtual instrument is nearly the same (or even better) that the one achieved at the laboratory using a 7-parameters PLS model (Adj. R2 = 0.80).
• WP3.- Correlation models and definition of specifications. During the task of improving the quality prediction modules based on the combination of the electrical and optical information of the MIS and NIR sensors, the next S&T results were obtained:
o The inclusion of MIS spectral data has not improved the accuracy of the calibration models based exclusively on LEDA-DRS spectra. This is attributed to the low correlation already found in the MIS spectra dataset.
o In relation to the calibration of the MIS, it has been proposed a calibration transfer procedure based on the normalization of the signals to the values obtained with the laboratory instruments. Calibration transfer procedures have been defined, and are available for the implementation of the calibration models in the industrial prototypes.
o Key specifications for the design of the pilot sorting line, including the geometrical arrangement and dimensions of the coils in the gradiometer, set of operational frequencies (MIS module), minimum distance between fruit pieces to avoid interferences between consecutive MIS measurements, dataset of wavelengths and layout of the LEDA-DRS sensor, the use of auxiliary sensing elements, and the characteristics of the conveying system (dimensions and speed).
• WP4.- Industrial prototype development. During the task of designing a FRUITGRADING prototype that includes both the MIS and the NIR sensors, the next S&T results were obtained:
o Integration of the MIS and NIR hardware modules into a fruit sorting line.
o End of the development of the fruit quality inspection system based on the LEDA-DRS spectroscopy. The LEDA-DRS integrated was able to perform 5 measurements per second.
o End of the development of the Magnetic Induction System (MIS) spectroscopy. The MIS sensor was improved until it reached 10 measurements per second.
o Prototypes (MIS and NIR) integration into an industrial sorting line including the electronics, mechanics and opto-mechanics (for the LEDA-DRS sensor).
o Replacement of the LEDA-DRS sensor by a NIR sensor provided by ATEKNEA which showed better results for sugar content detection in fruits.
o Calibration tests of the MIS and NIR sensors at the laboratory level.
• WP5.- Integration in a fruit sorting line. During the tasks of integrating the MIS and NIR sensors into the FRUITGRADING prototype, the next S&T results were obtained:
o The NIR sensor was integrated into the fruit sorting line provided by STC after electrical, mechanical, optical and software modifications.
o A user interface made by ATEKNEA with the assistant of STC and using LABVIEW V2013 which shows a graph containing the shape of the acquired signal when a fruit is present.
o Deliverable D5.1 that summarizes the work performed for the integration.
• WP6.- Industrial validation. During the task of industrial validation of the FRUITGRADING prototype, the next S&T results were obtained:
o Improvements of the prediction algorithm for producing better correlations between signal measure and fruit properties for the MIS and NIR sensors.
o The NIR sensor provided promising results for measuring the sugar content in pears.
o The MIS sensor provided promising results for measuring the acidity in peaches.
• WP7.- Training. During the task of training, the next results were obtained:
o A training plan was created in order to ensure that the lead-users, end-user SMEs and SME-AGs assimilate the project result including the technical, socio-economic and commercial benefits of the FRUITGRADING technology. The training plan describes the road map performed and all training actions within the FRUITGRADING project. All members of the consortium were actively involved under the coordination of the AJAP association and assisted by ATEKNEA and the representatives from the other SME-AGs.
• WP8.- Dissemination. During the task of dissemination, the next results were obtained:
o The consortium as a result of their actions could clearly disseminate the FRUITGRADING project to the Target Audience, making use of a strong campaign, through various means of promoting technology used. It was reached the goal in order to convey the real benefits of the product, achieving captivate potential interested in learning more about the prototype. The project was disseminated in various trade fairs, as well as the several publication of articles related to FRUITGRADING.
• WP9.- IPR. During the task of protection of the Intellectual Property Rights of the project, a PUDF (Plan for Using and Dissemination of the Foreground) and IPR (Intellectual Property Rights) documents have been created.
Potential Impact:
Fruit distribution in Europe is performed mostly at a regional or national level. Prices are conditioned by different variables related to the quality of the product, and in some cases, by stationary fluctuations of the demand in relation to the product supply. Independently of this, the oligopolistic structure of the fruit market has a strong influence on the fruit prices at origin. In 2002, the market share of the 5 largest retailers (discounts, hypermarkets, etc) in EU-15 reached 69%. The buying power of large retailers is therefore significant, and even more important when considering that large retailers often form buying groups.
In spite of this complex context, there is an increasing market demand for top quality fruit within Europe and the United States, which reflects the growing interest of consumers for high quality fruit. Interestingly, in spite of the decrease in the fruit consumption over the last years, the fresh fruit gourmet market has steadily increased. In the U.S. retail sales of gourmet/premium products increased 55% ($15,000 million) during the 2001-2005 period, with an average rate of 8.9% during the last five years.
Nevertheless, the divergence between the sorting criteria currently used in industry, based on external quality indicators, and the quality criteria perceived by consumers, has an impact on the price of the fruit. SME-AG proposers have estimated that the regular price of the extra class fruit could be further increased between 20%-40%, if the fruit producers could ensure that the fruit meets certain minimum “internal quality” criteria.
The availability of FRUITGRADING technology would lead to a significant economic benefit for over 2,500 European fruit farmers represented by the SME-AG proposers. In particular, the following key competitive advantages may be highlighted:
• The possibility of assigning part of their production to new “premium grade” segments, granting minimum quality requirements according to objective and measurable internal quality indicators. In contrast with existing sorting methods based on the external appearance of fruit, FRUITGRADING would provide an effective means for assessing fruit quality, an sorting fruit accordingly.
• Being a non-destructive analytical method, FRUITGRADING would allow for a substantial reduction of discarded fruit resulting from current practices (laboratory analysis)
• Enabling fruit farmers to design and implement pre-sorting strategies before fruit storage in maturation chambers.
Regarding the dissemination activities, the consortium created a dissemination plan in spreading the FRUITGRADING through Publications, Magazines, Media, Advertising, workshops and fairs and events. As chairman of the Exploitation Board, Urszula Makosz (TRSK), coordinated all the activities encompassing any transfer of knowledge outside the Consortium. All partners took an active role in discussing and deciding the content of the material to be published to ensure that initially no confidential information was disclosed on potentially marketable aspect of the project. Furthermore, aspects dealing with the design of the MIS submodule, LEDA-DRS spectrometer, as well as quality prediction models, was regarded as particularly sensitive, and therefore excluded from any dissemination outside the consortium. However, all partners promoted the widespread diffusion of the generic results in terms of the performance and industrial applicability of the developed technology at a European level, through the channels specified in Deliverable 8.2. The consortium carried out mailing of information about the project to government authorities and interested parties. Information and exhibition stands were set up and manned at the various different identified and selected events. The consortium carried out specific dissemination actions of the FRUITGRADING project in the media and in some events related to the agricultural sector.
List of Websites:
http://www.fruitgrading.eu(si apre in una nuova finestra)