WEBVTT 00:00:08.534 --> 00:00:10.977 So now that we ... 00:00:11.658 --> 00:00:13.270 ... I'm a programmer .. 00:00:13.972 --> 00:00:15.581 ... I hope that we're going to get a bigger soon ... 00:00:16.302 --> 00:00:20.161 We specialise in masonry. Masonry bridges in particular. 00:00:21.080 --> 00:00:24.773 We do a few other bits and pieces, but the core is basically bridges. 00:00:25.174 --> 00:00:29.281 We're always looking for new ... 00:00:29.563 --> 00:00:31.349 Sorry! 00:00:31.349 --> 00:00:39.302 ... 00:00:39.830 --> 00:00:42.212 Hopefully that doesn't trouble. 00:00:44.316 --> 00:00:53.036 So I think we're looking for ways to understand damage better, and in masonry that is an interesting task. 00:00:53.535 --> 00:00:56.819 And in the last couple of years we've done a lot of work on developing photogrammetry 00:00:57.183 --> 00:01:02.250 which we believe has the potential to transform inspection and assessment of historic bridges ... 00:01:02.788 --> 00:01:04.458 ... and all masonry bridges ... 00:01:04.458 --> 00:01:07.682 But really all masonry bridges are historic bridges. 00:01:10.872 --> 00:01:16.098 I don't have to tell most people here, I think, about photogrammetry, so I'll cut out all my usual slides. 00:01:17.401 --> 00:01:20.171 The basic process is you take a lot of photos, with lots overlap ... 00:01:20.736 --> 00:01:23.389 ... compress them down into a model and then you profit from the results 00:01:24.466 --> 00:01:28.058 Reality is obviously a bit more detailed. 00:01:30.348 --> 00:01:33.595 I'm not going to talk about processing a basic trick that we have here ... 00:01:33.595 --> 00:01:36.232 ... is big workstation and reality capture. Which is fantastic. 00:01:38.620 --> 00:01:49.486 Taking photos is more of an issue, and how you get information out of the data that we produce is the second part. 00:01:51.100 --> 00:01:52.280 A couple of case studies. 00:01:52.650 --> 00:01:59.182 Pontypridd Old Bridge was built in 1750. 00:01:59.358 --> 00:02:02.150 It was the fourth attempt that Williams Edwards made. 00:02:02.475 --> 00:02:05.500 He was given a 7 year contract to cross the Taff at this point. 00:02:05.739 --> 00:02:13.064 The first attempt was a 3 span washed out by floods. 00:02:13.317 --> 00:02:16.619 He thought, well sod that. if I'm going ... to go in 1 step 00:02:16.951 --> 00:02:22.327 That one, the centering collapsed before the bridge was finished. 00:02:22.590 --> 00:02:26.828 Not being put off, he did it again, did the centering a bit better. 00:02:26.891 --> 00:02:28.262 Successfully structured the centering. 00:02:28.262 --> 00:02:29.760 Six weeks later the bridge collapsed. 00:02:29.760 --> 00:02:31.931 And it collapsed in a particular way 00:02:32.151 --> 00:02:34.240 with the crown bursting upwards and the sides going in. 00:02:34.936 --> 00:02:39.571 So he then gathered a little bit more money together from the gentry of the area 00:02:39.820 --> 00:02:44.323 who must have been quite ... 00:02:44.613 --> 00:02:48.442 ... he must have been quite compelling, or maybe it was the fact that he was a well respected preacher. 00:02:48.715 --> 00:02:51.079 Well he must have thought so anyway. 00:02:51.371 --> 00:02:55.459 He gathered these transverse tubes at the sides of it where it's out ... 00:02:55.718 --> 00:02:58.775 ... and if you look at the top of it you'll see there's a partition ... at the top ... 00:02:58.984 --> 00:03:01.277 ... and it's been there since 1786. 00:03:03.519 --> 00:03:07.783 100 years later they built this bridge along the side, that's been widened since. 00:03:08.073 --> 00:03:10.888 And now it runs to about half a meter from the circle ... 00:03:11.202 --> 00:03:17.213 So since 1850, or it seems, the south elevation ... 00:03:18.683 --> 00:03:23.654 Capturing photographs of this, what we're aiming to do is to get set sub-millimeter detail from the whole structure. 00:03:23.918 --> 00:03:30.141 It's big. It's 43 meter span. It's all rubble masonry, basically lots of roughness, lots of detail. 00:03:30.424 --> 00:03:32.975 The crown is about 10 meters off the water. 00:03:33.512 --> 00:03:40.247 Victoria Bridge is very close, which ... 00:03:40.984 --> 00:03:47.193 And at least there was a ford here, so in the driest summer in recent years it was certainly possible walk backwards and forwards underneath. 00:03:49.933 --> 00:03:52.745 And we ended up with 14,000 photos. 00:03:55.556 --> 00:04:02.634 That's taken straight from reality capture. And reality capture can take 14,000, or indeed up to 30,000 quickly photos. 00:04:03.563 --> 00:04:05.652 And align them. 00:04:06.117 --> 00:04:10.464 30,000 took me within a gnat's whisker of 256gb of memory. 00:04:12.327 --> 00:04:15.635 But it just does it. 00:04:20.307 --> 00:04:28.552 And you finish up with a model that allows you to step back and look. 00:04:28.811 --> 00:04:36.313 This is the south elevation. No-one has seen that for 170 years. 00:04:41.462 --> 00:04:43.991 In this view ... 00:04:44.521 --> 00:04:48.305 ... 5 million triangles with a very high detailed texture that ... 00:04:48.897 --> 00:04:54.453 ... mesh means that the texture is just slightly blurred on these rock surfaces. 00:04:55.360 --> 00:05:02.604 But we can zoom in and see cracks, and we can see details that we wouldn't really understand without the photographs. 00:05:03.036 --> 00:05:07.792 ... for bridge inspection is much a detailed photograph site as issue here. 00:05:08.696 --> 00:05:09.894 Forced out. 00:05:13.248 --> 00:05:17.631 And so being able to understand that damage in context is pretty good. 00:05:19.895 --> 00:05:25.582 Being able to read the full photographic elevation of light to measure things. 00:05:26.924 --> 00:05:28.535 And we can zoom in to have a look. 00:05:33.234 --> 00:05:36.095 For some reason, it's not quite as good on the software, which is annoying .. 00:05:36.586 --> 00:05:38.551 .. but this allows you to see ... 00:05:38.811 --> 00:05:41.927 ... it's still pretty good. You can see individual bits ... 00:05:45.168 --> 00:05:47.566 A bit on collection. 00:05:47.711 --> 00:05:49.341 We don't use drones. 00:05:49.341 --> 00:05:55.644 We don't have the scale to justify having the the utilisation that justifies the licensing costs. 00:05:55.644 --> 00:06:00.156 And anyway, I've yet to come to a bridge and find that half of it is only accessible to a drone. 00:06:00.610 --> 00:06:05.357 No trees. There's not a bridge next to it. It's next to traffic ... 00:06:05.794 --> 00:06:07.597 It will be useful for the some bits of it ... 00:06:07.984 --> 00:06:09.583 .. but we don't have to deal with this stuff anyway. 00:06:10.147 --> 00:06:15.096 So there we are ... standing in the river we've got some mud weights attached to the bottom of the pole for balance. 00:06:15.096 --> 00:06:17.801 And we can get it to about 8 meters. 00:06:17.801 --> 00:06:20.889 Different bridge, and we can stick it out over the side. 00:06:20.889 --> 00:06:22.196 ... 00:06:23.927 --> 00:06:26.515 And we have camera rigs. 00:06:26.767 --> 00:06:31.138 So this is set of Sony QX1 cameras all strapped to a 6 meter beam. 00:06:32.179 --> 00:06:36.520 The triggers are powered on a wire so they trigger almost constantly ... 00:06:37.234 --> 00:06:42.042 ... they're carbon batteries, so there's always one that's running out of battery. 00:06:43.345 --> 00:06:46.179 You can't take it in and change the battery if it's tied a beam. 00:06:47.458 --> 00:06:51.094 ... Here you could walk into the swinging ... 00:06:51.662 --> 00:06:57.241 ... rigged up the Sony to the side ... 00:06:57.854 --> 00:06:59.690 And why bother? It's a lot of effort. But ... 00:07:01.247 --> 00:07:05.929 The ability to zoom in and see that detail, and then step back and see where it is. 00:07:05.929 --> 00:07:07.896 I keep doing that in the office ... 00:07:07.896 --> 00:07:10.098 ... where it's warm ... 00:07:10.265 --> 00:07:12.120 It's just unnatural. 00:07:12.894 --> 00:07:14.895 The problem is we get a lot of data. 00:07:14.895 --> 00:07:19.243 We don't need the data. We're looking for understanding. 00:07:19.342 --> 00:07:21.031 Just having the model is a step forward. 00:07:24.455 --> 00:07:28.923 We have a point cloud. You've got a vast amount of data ... 00:07:29.233 --> 00:07:33.203 ... you can convert that into a mesh, and bring the data level down and get more access. 00:07:33.515 --> 00:07:36.779 Photogrammetry gives you a bit of a much better place, and just having the model to look at is good. 00:07:37.268 --> 00:07:44.636 But we need to do more to interrogate them and extract information to learn the stories we have to tell 00:07:45.025 --> 00:07:46.957 and then we also need to share those stories with other people. 00:07:47.553 --> 00:07:53.688 So we've been working on a viewer with a few simple tools to help us take the model and say what's going on here. 00:07:55.853 --> 00:08:01.918 A very quick at an an 1830s canal bridge and then a bit of a bit of a look at a Clifton suspension bridge. 00:08:03.725 --> 00:08:06.140 One slightly sad bridge, and also very faint. 00:08:08.568 --> 00:08:15.214 It's just a standard canal bridge. You can see cracks here. 00:08:16.493 --> 00:08:21.631 About an hour on site and 2000 photos got us this and gave us this level of detail, where we can see into every crevice 00:08:22.051 --> 00:08:26.470 we can see all those lost bricks, and we can see how this bit dropped here. 00:08:31.535 --> 00:08:33.336 And we can look around it. 00:08:35.413 --> 00:08:36.471 So far, so good. 00:08:36.789 --> 00:08:40.350 But we need to understand what's happening. Why is the brickwork here? What's going on? 00:08:41.458 --> 00:08:45.127 And what we can do with the viewer is ... 00:08:45.326 --> 00:08:54.049 We know there's something funny about the crown. Let's put in a horizontal plane through the point. 00:09:01.908 --> 00:09:03.576 And it gives us an imaging contact. 00:09:03.859 --> 00:09:05.385 So this is used for the shading engine. 00:09:06.086 --> 00:09:11.430 We can see the intersection. We can see that it's lower here. That there's a ... 00:09:12.016 --> 00:09:14.679 ... at this point. 00:09:17.913 --> 00:09:21.872 At this end there's also mould, but not nearly to the same extent ... 00:09:23.979 --> 00:09:29.731 ... and all of that lets us start to get an understanding of what's going on interactively. 00:09:30.089 --> 00:09:33.560 So we can contour this. We can say we've got a mesh, we're going to do a set of contours. 00:09:34.121 --> 00:09:36.112 We've got to decide what contours and what spacing. 00:09:36.898 --> 00:09:39.325 With this ... 00:09:39.524 --> 00:09:43.500 ... and play and move it around and see what it is and then go and start producing your contours. 00:09:45.626 --> 00:09:48.447 ... 00:09:48.753 --> 00:09:54.487 This is the Clifton suspension Bridge. You may be familiar with it. 00:09:54.939 --> 00:09:57.191 Clifton's over on this side, and Leigh Wood's over here. 00:09:57.662 --> 00:10:02.138 This buttress built up about 10 years ago. 00:10:02.138 --> 00:10:05.978 Somebody doing some maintenance on the deck lifted a stone up and realised oh there's a big hole here. 00:10:07.335 --> 00:10:12.468 There's a shaft that goes down and they looked up and -- from there -- they could see that the shaft goes right down to the depths of the buttress. 00:10:14.144 --> 00:10:21.574 And obviously they explored further and discovered that around the shaft there's this whole network of vaults. 00:10:22.471 --> 00:10:28.908 So the road comes up over the river, runs along the central vault, and there were these two at the back and off the line. 00:10:29.487 --> 00:10:31.972 And nobody had thought, that's a massive abutment ... 00:10:35.133 --> 00:10:38.727 But once you know they're there you have to expect them. You have to know you're safe. 00:10:39.180 --> 00:10:41.479 Out of sight, out of mind. 00:10:42.283 --> 00:10:43.968 Just out of interest. 00:10:45.022 --> 00:10:50.720 We've got a laser scanner there. This was early in our photogrammetry exploration. We didn't know if it was going to work in pitch dark. 00:10:51.137 --> 00:10:58.634 So we needed the geometric information as well, and we got the laser scan, we got 45GBs 00:10:58.878 --> 00:11:00.136 converted that into a textured image 00:11:00.468 --> 00:11:02.242 200MB that we can load onto a normal PC. 00:11:02.577 --> 00:11:06.859 These models are about 200MB as well, per vault. 00:11:09.189 --> 00:11:16.891 This is ... we come in through this tunnel, past this shaft, having gone up another tunnel ... 00:11:17.125 --> 00:11:23.398 ... coming in from the other level through a ladder in the shaft ... 00:11:23.838 --> 00:11:26.805 ...at my height you have to remember to put your hand up before you get into it because you can't move your elbow. 00:11:33.885 --> 00:11:36.621 We find if we look at the bottom bit and turn it around ... 00:11:36.621 --> 00:11:39.678 ... we've got these things that throw cast iron ties into the walls ... 00:11:39.678 --> 00:11:45.863 ... that have been there since this was finished and closed. And they're is packed with timber. 00:11:46.748 --> 00:11:50.359 You can see there that they were vertical timbers, with just a bit missing in the middle. 00:11:50.679 --> 00:11:54.528 This one was filled with water before the others were discovered. 00:11:54.881 --> 00:11:58.145 And it was obviously a very stable level. It's moved up and down by about a foot. 00:11:58.601 --> 00:12:02.707 And the timber has rotted where that's happened and there's now coral down there ... 00:12:04.232 --> 00:12:09.528 You can look up the top and view along these things which are 10 meters up in pitch dark 00:12:09.528 --> 00:12:12.058 and you can see that tie is shining quite brightly. 00:12:12.977 --> 00:12:14.923 There's no attention to it. 00:12:18.751 --> 00:12:20.871 If we look to ... 00:12:21.157 --> 00:12:23.478 ... we can see a crack. 00:12:24.609 --> 00:12:26.606 We couldn't see that until we looked at the model. 00:12:27.105 --> 00:12:28.608 There's no way ... 00:12:28.747 --> 00:12:32.636 ... that by the time you've wriggled into that space, shining a torch ... 00:12:32.965 --> 00:12:34.698 ... that you're got to catch features like that ... 00:12:34.806 --> 00:12:41.305 ... but once you got this you can look at and take your time with about, you can see things that you wouldn't see otherwise. 00:12:45.848 --> 00:12:50.378 We can move the plain across here as well. 00:12:51.427 --> 00:12:53.230 And ... the plain. 00:12:57.309 --> 00:13:01.064 I've set that to just roundabout about springing where the whole starts turning in. 00:13:01.845 --> 00:13:07.350 You can turn it over clipping plain and change the clipping direction. 00:13:09.950 --> 00:13:13.463 You can go to photographic ... view from below. 00:13:16.747 --> 00:13:18.359 And then next to him ... 00:13:27.796 --> 00:13:28.494 ... colouring. 00:13:28.798 --> 00:13:38.190 So if we take a normal vector for every triangle and cover the mesh with that, we can very quickly see changes in matter. 00:13:38.754 --> 00:13:47.093 So you can see that although the shape of the bottom planes have curved sides and corners that is carried right onto the bulk of the top. 00:13:47.458 --> 00:13:50.380 But we can also see very quickly that there are orange stripes all the way around here 00:13:50.380 --> 00:13:55.469 ... 00:13:55.469 --> 00:14:02.687 So we can start to learn that the walls are built quite easily once we got to the top and built freehand as far as we could. 00:14:04.026 --> 00:14:06.367 Because working up there, if we built a center... 00:14:06.678 --> 00:14:08.519 ... you'd be fighting trying to work out ... 00:14:08.764 --> 00:14:12.724 ... and you stand at the platform ... 00:14:12.917 --> 00:14:14.709 and only at the very top do you put the center in and finish it. 00:14:15.056 --> 00:14:19.364 So you can see that it's smooth at the sides. Smooth at the top where the wood would be worked on ... 00:14:20.239 --> 00:14:22.780 ... there's always roughness in ... masonry. 00:14:35.854 --> 00:14:37.329 And there are interesting things like this. 00:14:37.936 --> 00:14:42.133 These stones were embedded in the stones. The stones actually slipped forwards as they were building it. 00:14:46.786 --> 00:14:48.921 We're going back there next week to do the top level. 00:14:49.414 --> 00:14:52.691 So those models are of the bottom, and we're going to back there and do the top one. 00:14:53.113 --> 00:14:54.431 I'm a bit nervous about that ... 00:14:54.795 --> 00:14:58.890 ... 00:15:00.849 --> 00:15:06.468 I've been developing a bit of extra kit to try and speed up the process and not just go like ... 00:15:09.904 --> 00:15:13.831 So developing an understanding of defects is really hard. 00:15:15.170 --> 00:15:20.026 Textured meshes are a lot more useful than point cloud in this respect, and that's not just if you're in the dark. 00:15:25.034 --> 00:15:28.890 It's possible to get sub millimeter details for very big structures ... 00:15:30.847 --> 00:15:37.920 And being able to take snapshots in time at very high detail gives you an ability to rethink 00:15:38.413 --> 00:15:42.472 ... whether it's just because you've thought of something and you need to go back and have a look 00:15:42.868 --> 00:15:47.155 or because you're looking at it again for a second time. 00:15:48.276 --> 00:15:50.082 Thank you. 00:15:51.206 --> 00:16:03.035 [APPLAUSE]