Beesource Beekeeping Forums banner

1 - 20 of 27 Posts

·
Registered
Joined
·
1,316 Posts
Discussion Starter · #1 ·
How good are computers these days at counting bees? I have seen this EyesOnHive product, and it claims to count "bees per second," or an aggregate estimate of overall bee activity.

But if I were to mount another BeeCam and point it directly downward at the landing board:
  1. Could a computer distinguish between bees going out and bees coming in, to compute a "net bee flow" in either direction?
  2. Could a computer distinguish between individual bees, so as to "recognize" a returning bee, match it with its outbound event, and determine how long it's been out of the hive?
    • This would be a function of not only the imaging resolution of the camera, but also of the differences between bees.
    • Are the stripe patterns of individual bees sufficiently distinct, like fingerprints on humans or stripes on zebras? Or do they all pretty much look alike within a given colony?
    • An imaging algorithm would have to be able to deal with swollen abdomens from nectar gathering as well as with pollen pants and wear and tear on wings.
    • I suppose a very far-off fantasy would have a tiny laser mounted next to the camera, to etch a name or number onto each bee as it leaves the hive.
 

·
Registered
Joined
·
38 Posts
You will basically have to train a image-recognition neural network as to what a bee looks like and then feed live video, one frame at a time, through the model to determine how many bees are in the picture. This will give you "Total Bees Outside" count.

That's step 1.

#1 - Determining direction would be a bit more difficult but not impossible. You can associate direction bee is pointing with whether it is landing or taking off. It would be relatively easy to train but it would not be very accurate. You can try to combine that with direction of travel, frame by frame, but that would be more difficult to figure out (orientation flights, robbing, general hover etc) and but would improve accuracy.

Anything beyond that is just headache and marginal returns. And to be honest - do you really need to know bee traffic within 1% margin of error?

You will also need more than just "Net Traffic". It doesn't mean much just by itself. Net Traffic of -10b/min could mean [out 10b/min] + [in 0b/min] or [out 110b/min] + [in 100b/min].

Forget about #2.
 

·
Registered
Joined
·
1,287 Posts
For 1 I think it would be possible. You would have to spend a lot of money (1K+) to get a good camera and processing software if you want it quick. You could do it cheaper if you get a good camera and write code yourself, but this will take a lot of time.

For 2 I do not think so. You could probably identify drones vs workers fairly easily but could not track individual bees unless you mark them. In general if you can not identify something to tell them apart on your own you will not be able to program a computer to tell stuff apart. You may be able to identify some other traits (dark stripes, colors, ect) but not individual bees. With Zebra stripes if you compare 2 pics side by side you can tell the stripes are different. I don't think there is enough difference between 1000's of bees with side by side (or overlayed) pictures to positively identify one.

I have seen bee counters on observation hives in a "zoo" type location. I do not know how they actually do it.

If I was trying to count bees I would be more likely to use a different counting method. What comes to mind is a row of tubes (10+) that are big enough for only one bee at a time to fit thru. I would then place 2 sensors (probably a break beam) on each tube. I could then make a system to count the incoming vs outgoing bees by looking at which sensor tripped first.
 

·
Registered
Joined
·
262 Posts
Absolutely possible to detect and count bees, but unlikely to recognize individuals. Machine learning could do this based on what I have seen. Bees are discreet and consistent, the hardest part will be the various lighting conditions throughout the day. Sunlight with shadows can cause some high dynamic range issues.
 

·
Registered
Joined
·
2,410 Posts
"How good are computers these days at counting bees? "

This sort of thing was possible 30 years ago ...

I used to have dealings with a firm called 'Gunsons-Sortex' who supplied machines to examine (and reject where necessary) grains of rice. Tons of rice were fed into gravity-fed hoppers and duly entered the machines - every single grain of rice was optically examined, with rejected grains being identified and subjected to the slightest puff of compressed air which deflected them onto the 'wrong' side of a knife blade, from where they descended by gravity down a shute and into a waste bin. The price difference between sorted and unsorted rice was obviously enough to warrant the use of such machinery.

So - a similar technique could be used - a battery of narrow entrance channels lined with optical sensors, which could very easily determine the direct of travel and make a running count at the same time. An 8-bit processor, 4-6 Mhz job - nothing fancy required. No need for cameras and complex software.
LJ
 

·
Registered
Joined
·
38 Posts
I suppose the answer is 'Yes, with enough time and effort'.

But WHY ????
That was going to be my follow up question as well. I mean, one could build pretty charts with bee traffic data but all that's going to do is confirm spring build-up and summer dearth.

Maybe swarming/robbing detection? But by the time you detect it - it will be already too late to do much about it.

And even doing it yourself you are already looking at 2-3 hives worth of investment. I'd rather just get more boxes and queens.

Maybe market it to the backyard FlowHive crowd?
 

·
Registered
Joined
·
205 Posts
"backyard FlowHive crowd"

great name

I belong to several electronics boards

There is a term used for those who have to have the newest and shiniest and most-hyped audio accessories

"Audio-phools" (take-off on 'audiophile' )

maybe there needs to be a term for the 'follow the latest trend' beekeeping crowd ?
 

·
Registered
Joined
·
1,316 Posts
Discussion Starter · #11 ·
I suppose the answer is 'Yes, with enough time and effort'.

But WHY ????
For science! For a deeper understanding of how bees live, work, and die. For the same reason Tom Seeley and his grad students (probably mostly his grad students!!!) sit on a wooden box to catch and individual mark _hundreds_ of bees during his experiments. So that we all benefit from their collected data, wisdom, and analysis.
 

·
Registered
Joined
·
1,316 Posts
Discussion Starter · #12 ·

·
Registered
Joined
·
3,530 Posts
I saw a study years ago where they put barcodes on the bees (much lighter) and a camera at the hive and then cameras at various feeders they had setup.
They were studying the bees preference for the different feeders
 

·
Registered
Joined
·
1,287 Posts
But WHY ????
I do stuff like this because it is interesting, I am curious, and/or it is a good learning experience. I have never done anything with computer vision but I have built robots, added lights and a horn to a cozy coupe, and put a datalogger on my hive because they all seemed interesting.
 

·
Registered
Joined
·
1,287 Posts
Mark my words :D; laser-marking of individual bees will be feasible and affordable (for scientific purposes) in less than 10 years. Maybe 5.
I am not sure if laser marking or ink jet spraying would work better. By ink jet spraying I am talking about the expiration dates that you see printed on containers. Time will tell.
 

·
Registered
Joined
·
5,486 Posts
I don't know if you program it to recognize a bee, but you could program general 'particle' parameters in which helps the software to identify single bee units coming and going. If you look how seed counters work, that's what you do. Some of the digital ones though process images so you could distinguish incoming and out going possibly, but that's more of a software issue and how you want to approach it. Do you use multiple pictures to determine direction of movement, or do you tell the software to try to determine which side is the head of the bee?
 

·
Registered
Joined
·
1,316 Posts
Discussion Starter · #18 ·
I am not sure if laser marking or ink jet spraying would work better. By ink jet spraying I am talking about the expiration dates that you see printed on containers. Time will tell.
I'd think that ink would get groomed right off.

And if we used lasers, we could also zap any phoretic mites!
 

·
Registered
Joined
·
1,316 Posts
Discussion Starter · #19 ·
https://www.kickstarter.com/projects/985910122/eyesonhives-the-health-monitor-for-honey-bees was an attempt, but no recent activity.

There was a local guy in Houston that was working on a gate counter system, but I don't have any info on it.

This all leads back to the queston of "why?" I suspect that a scale would provide much more actionable info for a fraction of the cost.
Yup! I linked to EyesOnHive in my original post. They posted some useful spam on a technical thread of mine last year. I'd never seen spam before that was so on-topic!

Yes, I might also put in a scale, as soon as I can find one that suits my needs, and can be read out electronically. That would give me data on the overall health of the hive.

But recognizing individual bees would give us information such as:
  1. For how long are foragers outside of the hive?
  2. Does this change over each bee's lifetime?
  3. How many bees never come back?
  4. How many lifetime trips did they make before they stopped coming back? Is it total # of trips? Or total # of minutes?
  5. How many "foreign" bees attempt to enter the hive? If we mark only on exit, we might get some data on that. Do foreign bees get adopted and start foraging?
  6. etc.
 

·
Registered
Joined
·
10 Posts
Yup! I linked to EyesOnHive in my original post. They posted some useful spam[/URL] on a technical thread of mine last year. I'd never seen spam before that was so on-topic!
Hi all! Sorry I'm late to the conversation - Eyesonhives is very much still active, alive and beesy! I should have returned to that thread - your comment about my 'spam' is pretty hilarious. I'm an engineer not a great marketer - I just saw a conversation and figured people might want to know that we'd purpose built something that was off-the-shelf ready to monitor bees! This past 2 years we've been focused on working with researchers (UC Davis is a big one!), but we still love citizen science and the spirit behind it:

"For science! For a deeper understanding of how bees live, work, and die... So that we all benefit from their collected data, wisdom, and analysis."
^ :thumbsup:

Eyesonhives is an easy-to-use tool to visually monitor and receive a data summary of the activity of a beehive. You're exactly right that it measures 'bees per second' - with an edge-network image processing algorithm on board the device. That lets it send up just the videos of interest, and keeps the costs of customer bandwidth and server storage data lower. We're also working on a refresh of the website at https://info.eyesonhives.com.

We have started working with a researcher who is interested in the 'direction' of bee travel - this is certainly possible, it's not offered as part of the standard Eyesonhives for now. Eyesonhives cannot distinguish individual bees at this point (although I swear there's a couple of drones that are always hanging around one of my hives for the last few days, that I think they're the same fellows).

You have some great ideas and questions. Re the foreign bees - through Eyesonhives videos we do see guard bees often challenging 'drifters', sometimes completely rejecting entrance to the hive. When robbing is happening in the apiary, this happens on an almost saddening scale.

Anyhwo, great conversation, we'd be delighted if you chose to get started on your project with Eyesonhives, if not, no worries! We just started using Calendly, so if you want to connect for a 15 minute chat, please feel free to block out some time https://calendly.com/eyesonhives/15min

Cheers,

Kelton & the Eyesonhives team
 
1 - 20 of 27 Posts
Top