Tech Talk Breakfast: F-751 Kiwifruit Quality Meter.

Hunter Weber

April 25, 2023 at 3:41 pm | Updated May 10, 2023 at 10:05 pm | 18 min read

We recently hosted a successful tech breakfast in partnership with Start Afresh – The Horticultural Innovators, where we explored the F-751 Kiwifruit Quality Meter. This innovative device utilizes NIR technology for non-destructive and rapid maturity testing. We are thrilled to share the latest results of our three-plus-year collaboration with our New Zealand partners in the kiwifruit sector.

We invite you to watch the video and learn more about this technology. Don’t miss this opportunity to discover how the F-751 Kiwifruit Quality Meter, developed with the expertise of our New Zealand collaborators, can help you enhance your productivity and quality control.

 

Video Transcription

Dave: Welcome to the first start of fresh breakfast this is uh taken us a little
while to get going we had planned for this just after about the middle of February but a cyclone that hit the
country disrupted that significantly it’s uh great to have Gail and George and Bob aleo here with us today they’re
going to tell us a little bit about their nir technology before I do that I would first want to say thank you to
Tyler for pulling this together for me so thanks and for starter first so thanks Tyler and and also the team at
the Tauranga Yacht Club for for the breakfast this morning the concept of the breakfast was just as
I say an opportunity to showcase new tech um one thing I want to say to you is that it’s not an endorsement I have to
say that but I want to say that because start a fresh and independent research provider we help companies to get their
technology operating in our industry but we’re not here to endorse the technology today and
we don’t make it any commercial uh gain out of this of course we’ve done some r d and in this case that r d has been
both paid for by Felix and by uh zesbury but it’s not a commercial game we don’t
stand up here and make a polar money out of giving Galen the opportunity he’s going to sell his Wares to you himself
um before I hand over to these guys and get
them to to tell you all about what they’ve what they’ve done um this is a a concept it’s the first
one we plan to have about three to four of these a year um hopefully I won’t have them in the
middle of the harvest season and there might be a few more people next time um although we don’t want to have too
many more people it costs a fortune to feed pranoli and all of those other people you knew I was going to pick on you
um but if you’re your guests of ours today if you know of technology that
you’d love others to hear about we’ve got some other ideas then we’re keen to hear about it too if we can create the
opportunity to showcase um then we’re keen to do it I think out our next one we’re looking to have a
couple of booze so you’ll be a key speaker but there’ll also be some other technology for for people to wander
around and have a look at so that’s start a fresh breakfast too so without
further Ado I’m going to ask Galen uh George the director of applied science at CID biosciences
is going to shape the story of there and share the story of their nir device
development and he’s going to be joined by Matt uh sorry Bob after uh after he’s
spoken Bob comes from mat Solutions and they’re the local distributor of the
technology so Galen welcome to New Zealand great to have you here
Galen: uh well thanks everyone for joining us this morning as Dave already said uh we really appreciate you guys coming taking
your time out of your busy schedules I know it’s middle of harvest but uh we’re really excited to talk to you about this
technology because we really think it’s going to make a big impact uh and how we do quality assessment in this industry
and so just some introductions first I want to talk about myself really quick and then we’ll talk about our company my
name is Galen I’m the director of applied science at Felix instruments which is a sister company to CID
bioscience we have over the past 30 years now really honed in on making
non-destructive instruments scientific instruments for plant research and now
also for commercial agriculture applications my background’s in food science I’ve worked in a lot of labs my
whole life doing a lot of quality assessment and safety assessment in the agriculture finished foods and cannabis
sectors so a little bit about our company first
uh you some of you might not be aware of who we are we are a small company based
out of Camas Washington in the United States that’s in the Pacific Northwest and we are a company of about 25 people
that manufacture engineer develop do all the r d for commercialize and
create Technology Solutions so we do all of that in-house we build
we manufacture we assemble we do the engineering all in-house at our facility
in campus Washington now how does a company of 25 people actually get all those instruments out
to users worldwide is we rely on a distribution Network and so that’s why
we have Bob with us today who will talk after this but we rely on these distribution Partners to help us have
better support and service and sales in a localized region and so Bob is based
out of New Zealand and so to better address our New Zealand customer base he’s our point of contact so you have
someone who’s local and can actually help service and and sell you
instruments right in your own country um we are utilized worldwide as I
mentioned um so we have customers in over 500 or sorry 100 different countries right now
all sorts of different growing regions for all sorts of different Commodities so we don’t just make kiwi fruit
instrumentation we make instrumentation for a lot of different Industries
and the main thing that we do at Felix instruments is portable quality sensors and these sensors use nir
spectroscopy as a the base technology and with this technology it is possible
to do rapid non-destructive measurements that give you multiple different quality traits for the internal quality of a
commodity so right now we have four different uh
commercial I guess uh products for our nir line of instruments we call these
the f-751s and we have an avocado version we have a kiwifruit version and
we have a mango version and we actually are just releasing a melon version as well
we also have an F750 which is kind of a generalized research instrument that can
be used for any of these Commodities all of the models work across all the
instruments so you can really it’s basically all the same Hardware just a
different model and so we’ve spent a lot of time
building up these sensors so that they can provide the end user with a reliable
means to take rapid measurements non-destructively so you don’t have to actually destroy your fruit use them
wherever you need to use them use them out in the field they’re robust enough to handle field conditions and field
varying field temperatures and lighting conditions they’re accurate enough to give you
confidence in the results and they’re also able to provide you even more insights than what normal testing
schemes can provide you so the ability to collect even more data to give you
bigger Data Insights is another key component of this technology
now I’m not just going to say you know we have customers everywhere and keep it that vague we do have customers that
have had great success implementing this technology in their specific
businesses or commodity sectors so our one really big probably most well-known
success story is with the Australian mango industry Association they’ve got upwards of 100 units right
now that they use across their entire industry to help them monitor maturity
throughout the entire growing season and help inform when they need to harvest at the optimal time in varying regions
throughout Australia and this is goes across multiple different mango cultivars
and so they’ve been using this now for a long time and using it quite successfully to map out all this data
and actually plan for their harvest in a more I guess Big Data informed way that
allows them to make sure that they’re getting the optimal quality from their fruit and harvesting exactly when they
should be another success story coming out of Australia actually is the Costa group
they are avocado Growers well they have a lot of orchards and they also are in
charge of packing they’ve got multiple units deployed across multiple Orchards they’re measuring Hass maluma Hass and
Shepherd avocados and they’re doing something similar to Australia where they’re trying to in the field you know
map out their quality and actually monitor the maturity of the crop but
then what they’re doing is just dumping that data into their own proprietary Solutions or current business data
stream and doing all that mapping in-house and so this instrument is meant
to be this kind of open Source instrument where you can do with it whatever you need to in order to get
that data into whatever platform you need to get it into and to do whatever
you need to do with the technology it’s trying to be as open of a solution as possible so that you and and transparent
as possible so you have access to all the data you can see everything that happens within the instrument
so in late 2019 uh we actually had a meeting with Dave
at starter fresh and we learned about this initiative that zespre wanted to
undertake where we were going to essentially collect data over multiple seasons do all these kinds of validation
trials with our our current model that we had in 2019 and they were doing this with our
technology alongside all sorts of other nir Technologies just to assess how well
this technology could fit into your existing systems and so what we did is we worked closely
with Dave and Sean to collect a whole bunch of data test it against our models
add that data into the model run new iterations and test and validate over multiple seasons over multiple
conditions rain different rainfall conditions different temperature conditions and lighting conditions and
and all sorts of different variables because all this variability needs to be built into these models and so what we
did is over the last three seasons we’ve been collecting that data validating and now we’re at the stage where we look
at this model and we have a pretty good degree of confidence in in what the
model can actually do and so I wanted to actually show you some data and I have full validation
reports here for anyone that wants to take this back this data back look at it analyze it assess it for yourself this
is just going to be a a quick little summary of exactly what we’ve done for
these models but we have the full robust reports on this table here printed out ready for you
um as I mentioned we have two different instruments we have this F750 and we have this f751
the F750 is a Research Unit the 751 is meant for more commercial applications
and what I wanted to do was to show you the differences between the two
instruments so that can help better inform your decision on which instrument is the right fit for your company
or your operations and so what we’ve done what I’ve done
here is essentially take out a hundred data points of F750 data 100 data points
of f751 data randomized across all seasons that we’ve collected data from
New Zealand this is a New Zealand only data and I’ve compared how the F750 performs
versus the f751 and there’s a lot of Statistics underneath as well to help summarize that but essentially what the
point I want to get across with these graphs is that we have very similar performance and that’s because we’ve
done is our we’ve done due diligence to ensure that the data within the model is
equivalent from both types of instruments and so when you do that the neural network we use can actually learn
which type of instrument is being used and it can more accurately predict based on the instrument type
and so another thing I wanted to just kind of point out is the actual accuracy numbers
so this is averaged over 100 data points you can see that the average between the
reference method which is for dry matter the oven drying method versus our F750
or f751 average across 100 devices or across 100 measurements is within point
zero 8.0 you know 0.1 of each other that’s an
average of 100 fruits so that’s easy to have averages of large fruit um you know be really close to each
other but if you look on a single scan basis our rmses are below one which
means that on average when you take a measurement with a device if you scanned
any random kiwi fruit this is for gold so if you scan any random gold kiwi fruit
and you then take that measure that same kiwi fruit you put it in the oven and
you dry it those two measurements on average are going to be within 0.64 for
the F750 or 0.79 for the s751. of each
other and so that is uh the kind of level of accuracy that we wanted to see
we wanted to bring it we obviously you want to bring it down as low as possible with compounding errors it is impossible
to get this to be an exact one-to-one relationship because you have air present in the human operator of the
oven how they’re cutting the kiwi fruit all those compounding errors uh combined
with the fact that we’re actually modeling based on those techniques that have air this is uh really
um really good performance for including all those compounding errors for green we have something similar I
also did not remove any data from these data sets so I included there are bad predictions or outliers those are all
included so you can see you know for the 750 there’s a couple fruit that predicted too high based on the
destructive measurements only around 13 or 14 but it’s predicting almost 17 or
18. so there are those times where you’ll get a scan That Is Random a
randomly too high scanner randomly too low prediction scan but that’s where the power of averaging comes in because if
you are taking scans of multiple fruit and using the average then what you do
is you smooth out that error and you still get really good predictions and so with the Hayward green this is still dry
matter we’re seeing again rmses below one and the f751 actually outperforming
this F750 so they’re very equivalent Technologies
and with the red we have a little bit above one for the F750 and just around
one for the f751. one thing I want to say is that this
isn’t in I think that there’s a perception with this technology that we create it and then we wash our hands of
it we’re done we give it out and that’s that that’s not how this technology works and we know that we know that this requires
updating these models require updating with new data and every season you know adding new data in revalidating making
sure that the entire range that you guys need to be able to predict in is represented in our data sets
so we make a very concerted effort to make
sure that this technology works for you the way you want it to work for you we don’t just make it and abandon it and
let you guys deal with it the accuracy and that’s that so we’re going to
continue collect data we’ve got more data from this season that we’ll be adding into the model
filling in some gaps in the range to make sure that everything’s evenly distributed as far as all of our
reference data and just continuing to make sure that it remains temperature independent that it remains independent
of all these other variables that you’re going to encounter weather variables and on all this stuff so that is how this
technology works and that is our philosophy on how this technology works is that we aren’t
going to just make it and abandon it for you guys to just deal with so we’re always going to want to cooperate with
you to make sure that it works the way you need it to work and just a has a little bit of a wrap up
the as I’m already mentioned you know these instruments can be used anywhere really along the supply chain and I
think that’s going to be a pretty instrumental key to this technology is that we can use the same exact Tech in
the field at the pack house along the supply chain to make sure that the quality is staying the way we need it to
because at the end of the day it is all about what the consumer is getting and the quality of the consumer is is
perceiving from your fruit and with all of the challenges this industry faces as far as climate change
whether varying weather conditions extreme weather conditions labor shortages all sorts of all sorts of
other challenges that are going to be new challenges that we’ll see in the future with all that this kind of
technology nir technology the f751s and
other online nir Technologies and all are going to be very beneficial for the
industry as far as quality control is concerned but also on saving up time on
on on doing measurements saving fruit waste so reducing food waste reducing labor costs reducing
inaccuracies between different operators of instruments and and and quality
assurance technicians that are doing you know maybe they’re doing different techniques so all this stuff we’re just
looking for consistency non-destructive ability and the ability to do it wherever along
the supply chain and so this kind of Technology provides all that and I just
really think that this is a great time in this development of this technology for us to start thinking about how this
technology is going to shape our our industry here and also how it’s going to
be able to fit into your current practices and that’s really all I wanted to say
again uh afterwards if you guys want we can actually show you some Hands-On demo
with the instruments you can play with them you can you can see it what they’re all about we have as I mentioned
validation reports up here as well and uh I will have my business cards out as
well for you guys to take but I will pass the torch on to Bob and let him talk a little bit about what kind of
service he provides locally for you guys so up
Bob: well um thanks for the opportunity and for coming out well relatively early for
most people thanks Dave for the opportunity as well to be a part of this um how many people have used an nir or
come across an nir oh great brilliant I don’t need to babble on about it then all right
um I perceive nir as near infrared which means it gives you nearness it gets you
close to the result but then the closer you are is determined by how much background data is in what’s in the
model how robust is the model the r squared value like we’ve just seen so there’s a lot of work that’s gone into
this and standing here very excited after a long three years of waiting that we actually have a model that works and
that’s the not the device itself but the the banking data so anyway
enough about that so um Bobble air Matt Solutions so we are based in
Christchurch but do cover the entire country for services so we have a
business that started um as part of another business uh started 1987 so we’ve been around a
while a family-owned business um and we have now two distinct businesses out of that one of them ismat
Solutions so we Supply instruments support them and we
also have a testing laboratory which gives us really great insight into applications uh specifically enabling us
to be able to Hands-On work with customers on different challenges
um so for Agriculture and environment we have of
course nir we have guest analysis uh so is an environment things like your soil sensors and capacitance as well as
weather stations and the like from meter plant growth Chambers it’s a recent
addition you can think of it as a fancy incubator they can do CO2 control ethylene control as well as temperature
and humidity of course in light and then we also have which is not applicable in this room necessarily but grain handling
and sampling Solutions as well on the food side we have a range of instruments
again kind of goes across with the modified atmosphere packaging SO gas
analysis oxygen carbon dioxide ethylene water activity how dry is something nir
again leak detection and so we’ve got texture in at the
bottom there we’ve got temperature and humidity control but one thing that’s really cool that we
do is shelf life analysis because um we have an entire side of the business that’s focused on shelf life
and shelf life Improvement Solutions um now I bring this here because this is
the basis of the lab that we’ve got uh that then allows us to do some some
fancy stuff nothing to to the level that starter Fresh Starts might add that’s a
next level but our operation is is based on informing a customer definitively on
what their products going to behave like in different environments in different
uh you can call it we’ve subjected to different challenges to see the resulting resulting effect on shelf life
so here today uh in partnership with Felix since 2017 we’ve been working with
Felix instruments and we look after their Solutions in New Zealand
um so we import them Supply them locally and we also relentlessly support them so
without question anything happens we’re your local protocol so from the nir to
to the gas analysis Solutions as well um so we have other partners we work
with so the hunter labels for color we have meter environment that do the soil and environmental Monitoring Solutions
um all through that we um we are not the source of everything however so we’re not a catalog company I
wouldn’t give you a fat catalog and expect you to choose what you want however we prefer for you to talk to us
so we like to think of ourselves like the name says met Solutions we’re a solution based company so we listen to
what your challenges and then we’re trying to find a solution that fits that challenge however if we don’t have it
then chances are we know where to find it or where we can refer you to so we do
not expect you to just pick something from the table and take it away and use it well in this case whatever is on the
table is ready to use however we work with you to get the best solution for
the job um so our premium service uh I think the name
says it all it’s a service and that’s why we have a business that’s been growing double digits for the last I
don’t know 15 years and this is because our service is based on a guarantee on
offering you support so everything we sell and we do we have a a wee bit of
experience with nir so we do sell some pretty fancy nir some of them hundreds
of thousands of dollars and all of them are supported locally so we have fully trained in-house technicians that
either do it in-house in Christchurch or travel the country so our most northern
customer at the moment is in mango Toronto and the southernmost is Bluff and we cover everything in between and
we do have customers and feed mills and farm and food manufacturing and the like
so the key part is I wanted to highlight is
for example with this device if something happens you buy one and something happens to it you drop it if
you run over it with the with the truck then that’s a different story but you know something within reason happens to
it and you need support we’re there to help you now something that we pride ourselves in is a very very interesting
insurance policy which most people appreciate when you buy a device like
this and you have to send it away for example to Christchurch for service every courier company doesn’t matter who
it is we’ll only cover you up to two thousand dollars and most people get nervous uh sending a device like that
when they know what the cover is so we have a service an insurance policy at math solutions that when the device
leaves your door to when it comes back to your door it’s our responsibility for full replacement value and that’s
completely free so you don’t have to pay for any anything else and we also have
free loan machines so we have devices that are ready to go within within a
reason availability of course but we have devices ready to go if something happens to your device we just make sure
that the loan device has the same model as you’ve been using and with no offsets
I must add if there is offsets we’ll apply them first before we give you the device we’ll send you advice overnight
again free of charge you get it you start using it you send us and for as long as we retain your instrument you
can keep ours until you get yours back of course we charge for the service and part of it okay but the loan machine is
free of charge and the the other part of this is
is the model updates now I’ve worked with nir for way a bit of time
um and come across some interesting challenges with modal updates and the negative
perception about performance on nir is based on a underlying misinformation of perfection
so one divided by one okay doesn’t exist you I squared of one doesn’t exist well
nine now I’ve seen nine eight before but this was for Something Completely homogeneous and they had
I don’t know 80 000 samples in the model and it doesn’t change but
we work with Felix who work with starter fresh to collect the data season upon
season upon season and we update that model once we update that model if your instrument comes in for service we
update your instruments model to the latest model free of charge
at the moment okay so it’s free of charge so you get an updated validation
model in your instrument to get better results in the next season and this is a
free service okay so I’m talking everything free free free there is a catch okay so the catchers you need to
buy the device okay and for this um this meeting we do have instruments that
were utilized in the collection of these these models um in the collection of this data sorry
to create the models and so we would like to offer those instruments at a discounted rate
um and I think there’ll be 25 off the uh price uh list price and that so I’m a
Salesman I’m just blunt and unashamedly so so put your hands up if you want to
approach me I’ve got business cards here uh if you want to take this up but it won’t be repeated so there’s how many devices
17 7 16 or 17 devices and after that the 25 is gone okay now
they’re not they didn’t fall off a truck okay so they have been utilized in the
model development but they didn’t fall off a track so what man Solutions is going to do we’re going to get all the
devices back we’re going to run a full service on them if there’s batteries need to replace and whatever accessories
need need to be checked and when you get these devices you’re going to get them with full warranty just like a brand new
instrument okay and if anything happens again we’re just based in Christchurch so it’s not and in the same time zone I
must add so yeah it’s uh it’s it’s an added Advantage so
thank you very much I don’t want to blabber on too much um and any questions please do get in
touch um right here habit help thank you [Applause]
right I know I’m uh keeping you guys from Harvest some of you I just want to
say thank you first of all galen’s come from uh Mid North West is that what you
called it mid Northwest of uh of the US he had a false start getting here in
February and then he’s been out for the avocado conference so I think looking forward to getting home Bob thanks for
coming up from Christchurch I really reiterate any uh any ideas you’ve got
for a starter fresh breakfast we’re happy to hear them any feedback you’ve got if you think that maybe this morning
wasn’t quite what you thought it would be I told you it would be a free breakfast that’s all I told you and you
got that so uh so no com but if the breakfast wasn’t here like and let us know too
um thanks all for coming out uh if you want to catch up with Galen or Bob up here please come up and do that now
um it’s uh it’s your opportunity that 25 percent discount account that they’re offering just one thing I did want to reiterate
Galen had mentioned it is that this piece of work was part of a wider piece of work there were a range of other
Technologies in IR that were were evaluated um we may or may not in the future talk
about the other Technologies depends on those companies but that’s part of our
independence we’ve captured the starter independently and we’ve given that data
unchanged unchallenged to to Galen and his team in fact the way that that
happens is that we capture 100 of a data set and we only give them a portion of the data for them to create the models
on so it is they get a blind half of the data set is blind they don’t get to see
it all until they provide us a prediction when they provide us a prediction we give them back the other
half of the data set and they can’t change the prediction after that so that’s the way that we run these
independent studies to ensure that there is there is independence in there and that we’re not just fudging the numbers
um have a great rest of your day for those of you are there harvesting stay