tv Click - Short Edition BBC News June 26, 2022 10:45pm-11:01pm BST
much for talking to us and you very much for talking to us and he were back for the next review later thank you. that's it. dojoint us later. bye—bye. hello, and welcome to this big, green open space. never before have we appreciated outdoor spaces and parks like we have since lockdown. yeah, unfortunately, though, lockdown also saw an increase in something that's threatening to turn some of our green and pleasant lands into brown and unpleasant landfill, and that is the illegal dumping of rubbish.
there've been well over a million fly—tipping incidents in the uk over the past year, and paul carter's been to see how ai is hoping to help us find the fly—tippers. like many countries, england's seen a surge in the illegal dumping of rubbish during the pandemic. some people pull up and dispose of their waste, while others, well, they literally do it on the fly. here on the outskirts of london, a number of councils have turned to tech to combat the issue. cameras, but not as you know them. these use al to catch fly—tippers. since february, over 80 smart cameras have been installed at known dumping hot spots across a number of councils, including kingston—upon—thames and sutton. they're just one of the initiatives of the south london partnership's
innovate programme. it aims to harness the internet of things, or iot, to manage new challenges that have arisen during the pandemic and to pilot solutions to help people live better and healthier lives. so, i decided to put the cameras through their paces. unlike regular cctv, these cameras don't record continuously and are only triggered when they detect movement. that means it not only limits the amount of footage someone has to look at, it also reduces the carbon footprint of the solution. the footage is then transmitted wirelessly to a secure cloud—based platform and an alert sent out to the council's enforcement team. so, i'vejust dumped the rubbish that we've seen behind me, and you've had that come through to your phone as an alert. yep. can i have a look? yes. so, we get a notification on our desktop, laptop, whatever device we've got. so, you'll be able to see pretty much in real time almost
someone dumping something? yes, 100%. oh, there i am! oh, look at that — bang to rights! and it's very clear, crystal clear. so, actually, we can see the person, we actually can see what the items are on that trolley. is that important sometimes to actually be able to identify what it is that they're dumping as well? very important, because our crew really we want to know what they're going for. if it's got asbestos—related, if it's got anything that is also a health and safety concern. given these cameras are operating in busy urban environments, they're often triggered over 100,000 times a day. the captured footage could just show a passing car or pedestrian. so, why, then, aren't the councils getting thousands of alerts a day? how is the camera able to spot a legitimate fly—tipping event amidst all that noise? well, the magic isn't actually happening in the camera itself, but rather behind the scenes in the cloud. there's an aland machine learning element to this process. how does that work and what are the cameras doing
and what are they looking for? detection of people and that sort of thing is reasonably straightforward, and the cameras will do that themselves. what is the tricky bit is — what is rubbish? it's quite an objective thing, isn't it? yeah. rubbish in one environment is not rubbish in another environment. the cameras are installed, they collect a log of data of movement in that environment and then, our data scientists will then look at that and make sure we tag and we actively review the footage and make sure the tagging is correct. so is the right thing being tagged as rubbish and not rubbish? are we getting false positives, false negatives? and trying to minimise those elements so that what the customer sees is really accurate. so, that's really interesting. so, you're actually using a little bit of a human element to teach them context? that's right. which is quite a difficult thing for an al to learn on its own. and i think that's a really good way of putting it because vision technologies can identify what a box is or what a mattress is, but it doesn't know whether that's good or bad. so, we've got some images on the screen here of some shadowy figures dropping off some rubbish. i can see that it's got a green box
marked around there, so what's it identified? so, what we're seeing here is this is part of one of the automatic ai computer vision models that we're using. this is detecting that there's some rubbish that's come into the scene of the camera, and there's a person pushing that rubbish along — you may recognise him! chuckles. and you can see down the bottom here, so, the model has detected at a level of 68% accuracy and its opinion that that is rubbish. i'd say that's pretty accurate, yeah. yep, yep. and if we look at other instances, there's one here which hasn't been matched. so, now, this is the same trolley, so in theory, this should've actually been tagged. now, in this case, it had a person in front of the trolley, which is probably why it thought, "well, maybe that's not actually rubbish." maybe that's someone doing their shopping? maybe it's someone doing the shopping in front of a rubbish bin. laughter. and so here, we canjust draw a box over this and add it. the model will then improve itself based upon all of the additional feedback that we give it. and it seems that feedback loop is working, with significant drops in the number of reported incidents at former hot spots. the real challenge for councils,
like the one here, is preventing fly—tipping from happening in the first place and not just potentially moving the problem into other areas. but it's clear to see that technology like the one being used here does have real potential to change the area and improve it for people that live and work here. paul there. now, can we see that picture of the trolley again? yes, please! i think that was the teddy i gave him for his birthday! spencer cackles. i think it is! and what's that old chair? what on earth is there? that's definitely the shot of the show, definitely. so, we've have seen how a! can track fly—tippers. i know what you're thinking — can this technology also be used to track puffins? what?! trust me, that's what they're thinking. so, you know in the past, we've talked about how putting up wind turbines might affect the local wildlife? mm. well, nick kwek has been out into the north sea to find out how one energy company is trying to prevent problems for a population of puffins.
nine miles off the coast of wick at the far north of scotland, it's wavy and it's windy. you see them there on the horizon! i've come to visit the 84 turbines that make up sse renewables�* beatrice site. this is one of the uk's largest offshore wind farms with enough capacity to power almost 500,000 homes. the uk government, though, wants enough wind energy to be generated to power all of british homes by 2030, and it's cut approval times for new offshore farms from four years to just one. of course, it's not as simple as just sticking these things into the seabed. right now, with offshore wind and the really... the scale of development that we're going to see, we just don't yet know how that's going to impact the ocean, so we always need to think about what's the impact on habitats and wildlife and, you know, especially when we're
harnessing nature's resources, then we need to make sure that we're protecting the natural environment, too. they've been conducting a study with microsoft and avanade on the isle of may in fife, home to fauna such as seals, ducklings and the much—loved puffin. this is a sanctuary for puffins with around 80,000 nesting here each year. it's estimated because traditional counting has been done by eye, so researchers have engineered a way to keep better tabs on them. the effects that we see from building offshore wind farms isn't seen immediately within the local ecology. it's obviously an effect that takes time, and what's really important is that we start monitoring the local ecology to our wind farms so that we understand the impacts we have, so that we can implement corrective actions and potentially have a positive impact. they've installed four artificially intelligent camera systems to count
puffins and monitor their flight, each equipped with their own custom—made marine—grade jackets. you can't buy this off the shelf. right, 0k. and we've had these created specifically for these four cameras, which are on this islet. and so, the puffins are under cctv surveillance? they are, 2a hours a day. are they happy with that? laughs. is that a wee windscreen wiper? that is, indeed, yes, and that allows us to perform periodic maintenance, ensuring that that we can clean the lens of any salty deposits which are picked up from the harsh sea air. puffins generally, when they're congregating around their burrows, tend to face out, looking down this hill. so, we have one camera positioned further down the hill, looking straight on to the puffin, giving a good view of the portrait, whereas this gives a good side—on — a good side—on view of the profile of the puffin. you're quite right — i can see one right there, actually... exactly, yeah. ..looking side—on. so, as part of the trial, we really wanted to understand what would give the a! the best opportunity to recognise a puffin. in february, this was just a barren ground. there was no grass, let alone flowers. and you can see here now all these
white flowers that have bloomed, so the white flowers actually merge with the breast of the puffins in terms of the pixels for the alto pick up, so it was actually tricking the ai. 0h, right, 0k! yeah, and actually resulted in slight inaccuracies until we've retrained the model. this is chiefly a data—gathering exercise — the initial entries for a long—term puffin digital database. already, we're building up quite a strong picture of how the puffins behave during certain times of day. understanding, you know, when there's peak puffin activity versus low puffin activity. so, very soon, we should start to be able to understand when there's any anomalies in this behaviour. but at the moment, it is still very early and we are still trying to understand the data and really start pulling it together. we've got a very narrow window of time in order to ensure that we can protect a lot of ecosystems that are in huge danger, and so it's something that's incredibly close to microsoft's heart. is there a bit of responsibility or a bit of maybe even guilt
sometimes when it comes to addressing the climate change issues surrounding technology companies, having such a carbon footprint? well, we know that, you know, the carbon emissions that microsoft have represents less than 1% of the global carbon emissions. but you're right, you know, the data centre footprint is one that we're really focused on, and i think in the world that we live today, the demand for data and technology is one that's growing. and to satisfy that demand, we'll need more electricity. the thing is, if we're to reach government targets and lessen our dependence on burning fossil fuels, then we're going to have to embrace renewable energy. let's just hope the rollout is a harmonious one, for everyone's sake. what an absolutely stunning view! and there was nick kwek in there, too! anyway, that's it for the shortcut of click for this week. the full—length version is waiting
for you right now on iplayer. thanks for watching, and we'll be back next week! hello. let's take a look at what's expected on monday, and not a particularly cold start for the morning commuters. temperatures around six or seven o'clock around 10—13 degrees. bright enough start in the far west and across eastern parts of england, but in between, through eastern wales, western parts of england, cloud, outbreaks of rain which will turn more showery as that works its way eastwards through the morning and into the afternoon. lots of sunshine developing across western areas later, a scattering of showers, particularly across scotland. but much more sunshine in western areas and with lighter winds compared with sunday, should feel a touch warmer.
a little bit cooler, though, compared with the weekend in the east where we still finished with a bit of cloud and one or two showers. end of the day, though, northern ireland more cloud, outbreaks of rain and strengthening winds start to push their way and that rain becomes more persistent through the night and spreads its way up to south west scotland, the isle of man as well and around the irish sea. a chilly start to tuesday by comparison for eastern areas, but as we go through this week, the south and east of the country should stay largely dry with a few showers around, but always some rain at times to the north and the west. see you soon.
welcome to newsday, live from singapore, i'm st. the headlines. world leaders at the g7 summit call for unity to help ukraine defeat russia. we have to stay together. we did is counting on this from the beginning and somehow nato would in the g7 with splinter and we are not going to. in ukraine, residents reel from the latest act of russian aggression, as the capital kyiv is struck by a barrage of missiles. in south africa, forensic experts investigate how twenty—two young people died at a nightclub. diana ross helps bring the first glastonbury festival for three years to a glorious close