Monday, May 18, 2009

Publishing on Blurb - Part 1

In an earlier post I mentioned that I'm editing a book of KAP images to be published by Blurb. I'm excited about doing this project, though the learning curve has been a little steep. This is my first book, and my first experience editing other people's pictures. It's an ongoing story, but it starts here:

The first time I heard of Blurb was during Worldwide KAP Weekend 2008, when Peter Neville offered to take a collection of images from each of the photographers involved in WWKW and publish their photography as a book. Peter used Blurb, a company that does print-on-demand for books. The result was a very professional book that is fun to look at, regardless of the fact that I have two pages of images in it. So when Worldwide KAP Weekend was extended to Worldwide KAP Week in 2009, and when no one else stepped forward to edit the book, I did.

I had already been planning to publish a book of aerial pictures of Hawai`i, but at this time I really don't have enough pictures to justify putting a book together. My goal is to have at least five hundred print-ready pictures to cull through. I figured if it doesn't hurt when I cut out 2/3 of the pictures, I didn't bring enough good work to the table in the first place. So editing the book for WWKW 2009 makes for a good dry-run on my own book, and gives me the opportunity to work with other photographers.

In reading through the forums on Blurb, what became apparent very quickly is that photo editing can make or break a photo book. There is no one standard to move from a digital camera to a digital darkroom to a digital printer. What you see on the camera often bears very little resemblance to what comes off the printer. There are tricks involved. And that's where the learning curve really took off.

The first stage is to put all the images into sRGB color space. This is the color space used by the Epson printers that Blurb uses for printing its books, so the images will wind up in this color space eventually anyway. It's better to put them there first so that all the subsequent editing is representative of what the printer will see. Many cameras can be told to use sRGB color space, which is even better. For my own work, this is what I plan to do. For the WWKW 2009 book, I have to take what I'm given. Conversion first.

The next stage is to get an idea of what the printer will produce. This is where it starts to get tricky. There are two parts to this problem: The first is to make sure the monitor displays in a consistent standardized way. The way to do this is to get a colorimeter and monitor calibration program, and to use it religiously. I wound up getting a Spyder2 colorimeter with the Spyder2Express software. I spent an afternoon madly calibrating all the monitors in my house, and now when I start Photoshop on any of them and bring up a given image, they all render almost identically in terms of contrast, brightness, and color temperature. So far so good.

The second part to the problem is to get the ICC profile for the printers and paper Blurb uses, and to compare the images against it. Keep in mind that no actual image conversion is done at this stage. This just simulates, on a calibrated monitor, what a given image file will look like once it is printed on that printer using that paper and ink. It's not a substitute for an actual printed sample, but a good ICC profile used on a properly calibrated monitor is supposed to be pretty darned good.

The third part, of course, is to get a printed sample to compare against. So I spent most of my weekend putting together a 36 page test book with a number of images that could cause problems when printing. The book has been ordered, and should be here in a week or two. That's just in time for when submissions for the WWKW 2009 book are closed.

In going through the problem pictures for the test book, I ran into a couple of things I expect to see more of when putting the WWKW book together:

A number of images had patches of color that could not be reproduced on Blurb's Epson printers. When a color is outside the range of what a printer or monitor can reproduce, it's said to be outside of that device's color gamut, and that part of the image is said to be a gamut overrun. A lot of what I did with these images was trying various ways to get the colors back into gamut so that they print accurately. It wasn't possible in every case, and a number of images in the test book have patches that are outside of the printer's gamut. In some cases this was intentional, so I could see how the printer handled it. From what I gather it's like having a shadow area in an enlarger print that is just too dark for the paper to handle. It winds up looking muddy, with little to no detail. We'll see.

A number of other images had strong shadows that wanted to block up. "Blocked up" shadows are shadows with little or no detail. When detail does show up, it's typically a result of irregularities in the print medium or in image compression rather than any real details in the image itself. The result looks muddy and unattractive. In processing these, I did what I could to boost the level of the shadows enough that real details showed up when previewing the images using the Blurb ICC profile.

The counterpart to blocked up shadows is blown out highlights. In four color printing that basically means that no ink touches the page, and the underlying paper is the only representation at that point in the picture. No detail, no nothing. The trick here is to try to bring highlights down enough that texture and detail is preserved when previewing the images using the Blurb ICC profile.

One oddity that came up during this was a particular B&W image I used for highlight testing. I used it in two instances, one when doing color correcting, and one for comparing two choices in background. During the color correction comparison, the image was saved as an sRGB JPG for both the test image and the control. The test image had +1 green added to it to counter the magenta shift the Blurb printers seem to have.

But in the case of the background test, the image was saved as a grayscale JPG for one page, and as an sRGB JPG for the other. The images look very different once loaded into the Booksmart software. For the record, Blurb makes quite clear that all images are fed to the printers as sRGB JPGs. But when Booksmart does the conversion from grayscale to sRGB, it apparently shifts the gray scale values by quite a bit when compared to how Photoshop does the conversion. This is important to keep in mind when handling B&W images. The conversion to sRGB color space really does have to happen as the first step, or an uncontrolled conversion will happen later.

Other pages in the book test two page full bleed spreads, two page subframed spreads, tiled images, etc. I couldn't test every situation I'm likely to run into, but I tested quite a few of them. Even if the text in the test book is of no interest to anyone other than me, since it describes the details of each test, I hope the photography in the book is of interest, and that someone other than me will be interested in seeing it.

The test book should be here in one to two weeks. In the meanwhile submissions for the WWKW 2009 book open in three days, and I've got a lot of disk space to clear up before that happens. No matter what, one thing I learned while making the test book is that 90% of the work for the WWKW book will be spent preparing the images in Photoshop. I have a procedure for compensating the images for color shift, but each one of the images in the test book took a lot of manual manipulation to get the most out of them. There are no shortcuts.

But I'm looking forward to it.

Tom

Sunday, May 17, 2009

Workflow

Worldwide KAP Week 2009 was a big success for me. I had all manner of plans for places to fly and things to do, but I only managed to visit a small percentage of the sites on my list. I even caught a cold in the middle of WWKW and lost half the available days to it. Nonetheless I came home with a number of images I'm quite happy with, so I call it an unqualified success.

Last year Peter Neville very kindly offered to edit a book of WWKW images, and so the participants were later able to order a copy and have something to show to their friends, family, and people curious about KAP. No one came forward this year to edit the book, so I did. It's been a whirlwind learning experience, but a good one. And once the images start coming in the learning experience will only ramp up. But so will the fun! So it's all good.

The best part of this is that it's forced me to re-examine my own workflow. I found some flaws in what I've been doing in the past, so here's an opportunity to set the record straight:

The Camera

At the moment I'm still shooting JPG files. I know, I know, RAW is better in oh so many ways. But with the Canon compact cameras, I'm not convinced the benefits are all that huge. The biggest one, for me, is that the images don't have JPG compression artifacts in them. This allows a lot more latitude in image processing, but at a cost: the files are huge. On the ground this isn't a huge problem, but in the air it's difficult to know when your chip is full and to know when to swap chips. The result can be lost images. At some point I might make this change, but not now.

The Computer

As soon as I come home from a session, all images from all chips are dumped to my computer and erased off the chips. At this point I pop out all my batteries and load them into chargers. By the time I make a first pass at the images, my batteries are charged, my chips are all empty, and my bag is re-packed and ready to go again.

Original images are kept pristine. I don't like to make changes I can't un-do. Regardless of whether this is done in software, such as with Adobe Lightroom, or if it's done manually through a system of copying and saving that preserves the original file, this is important.

Once the images are on my computer I sit down with my KAP notebook and start going through the pictures. Clear winners are noted by image number and subject name. Sets of images that might work well as a composite stitch are noted as well, along with the subject name. I don't erase bad shots or non-keepers. These can still be used later in the process.

At this point I'll go back through the composite sets and look for signs of parallax shift. Tall thin things are the toughest since they tolerate almost no camera shfit. But if anything is obviously off, that set is either modified or scrapped. As each one clears this Mark I Eyeball check, I'll load it into my pano stitching software and give it a go. All panos are saved without cropping, full size and checked for anomalies. Sometimes these can be fixed in Photoshop. Sometimes they're too extensive and have to be scrapped. Scrapped composites get a line through them in my notebook.

The next step is to load the images into Photoshop one by one. I like to start with the composites because they're the trickiest to work with. That leaves the single frame images for the end, which is pure pleasure. No stitching anomalies to worry about, just straight darkroom editing.

Most of the time with the composites is spent with the rubber stamp tool, trying to repair anomalies. I'm fairly conservative about this, and will scrap a composite rather than edit out a bunch of anomalies. If I can tell it's been re-worked, so can anyone else. But this is why I don't crop first. If you can rubber stamp areas that are going to be cropped out later, it's a lot less obvious to the viewer later. Waste not.

Once the composites are repaired and cropped, I'll save them without further editing. At this point they're essentially a single frame image and can be treated that way.

Single frame images start by getting geometric corrections. PTLens is great for this. It will take out barrel or pincushion distortion, and can apply the geometric effects of the tilts and shifts available on a view camera. If an image has a strong horizontal or vertical, this can be an essential step in the process. Curved horizons just plain aren't believable to me unless the picture was taken from the stratosphere or from orbit.

Composites with curved horizons or tilted verticals typically go back to the stitching software for repair. Most have the ability to re-tune the projection of the image and take out things like curved horizons. If a composite needs repair of this sort I'll re-stitch it, tweak the projection, and once again save full size without cropping.

Back in Photoshop, the next step is to crop the image for composition. With a camera on a tripod, the photographer has a great deal of say in how the image is composed. With a camera on a KAP rig, even one with video feedback, there's a random element involved. The camera is never pointed quite the way you want, so the resulting images almost always need some form of cropping. With KAP, this is when I start looking at placing subjects on the 1/3 lines, when I start looking at leading lines and golden mean curves, and when I start looking for balance and motion. On the ground all this happens at the viewfinder or on the ground glass. But for KAP the real art starts in the darkroom or on the computer.

When cropping it is important to keep your final print product in mind. If this is for an 8x10, it needs to be cropped to that ratio. If it's for a CD cover, the ratio needs to be square. If it's for a poster, you need to know the dimensions of the poster, or the image area of the poster to be precise. And if it's for some indeterminate purpose like monitor wallpapers, you may not be able to crop at all. In those cases it's better to leave your options open.

Once the images are cropped, I'll save them under their new names, by subject. "Kua Bay Abstract" "Boogie in the Big Blue" "Can I Play Too" are typical ones. This is where the pictures take on their names. Past that point it's how I think of them, so it's important to choose carefully. Once the name is chosen, I'll jot that down in my notebook beside that frame number or that collection of frame numbers.

From here on out things get a little more dicey. This is where the bulk of the changes in my workflow have happened, because this is where the nit pickey details of getting images ready for print really kick in. At this point we have images that have been assembled, if necessary, had optical distortions of the lens removed, if necessary, and have been cropped for composition and the choice of print medium. But the color of the individual pixels is still untouched.

In darkroom printing, this is the point when a paper is chosen, a developer is selected and mixed, and the test strips get made to check for proper choice of contrast filter, exposure time, development time, etc. In short this is when the final print medium is chosen, and when the picture is tuned for that medium.

In digital printing, the same is true. Since the end result in this case is a book printed at Blurb, that's the example I'll use. The first step is to have a calibrated system so you know what you're going to get on the printed page. I wound up getting a Spyder2Express, which is a combination of a Spyder2 colorimeter and the Spyder2Express software for calibrating monitors. This is a very bare-bones calibration system, but it's all I need in order to calibrate my monitor for printing books at Blurb. More advanced systems allow you to calibrate projector systems, or develop ICC profiles for printers. Since I have the ICC profile for the printers, paper, and inks used on the printers Blurb uses, I really don't need anything that advanced.

One by one, the pictures are loaded into Photoshop on the calibrated system. With the Blurb ICC profile installed, I can preview what the picture will look like in print. Levels and curves are used at this point to render the picture the way I want it to appear in print. Previewing is an important step because it can show you where shadows are starting to block up, where highlights are getting blasted out with details lost, and where individual patches of color are falling outside the gamut of what the printer can reproduce. This is similar to the exposure strips that are done on an enlarger to check exposure time, development time, etc.

During this process, if shadow areas are starting to block up, but the overall exposure level of the image is still too light, selective dodging can be done to bring up the shadow areas so they don't lose detail. Likewise if highlights are starting to blow out, those areas can be selectively burned in to preserve detail. And if there are certain features in the image that need to be emphasized, selective dodging and burning can be done to emphasize those areas of the image.

When doing this I try to make all the changes using layers in Photoshop. This lets me go back later and tweak any individual effect, similar to how test prints and dodge and burn guides are done on an enlarger. I don't really like the Photoshop dodge and burn tool so I will typically do this by selecting an area with the lasso tool or with the magic wand tool, feather the selection, and apply that selection as a layer mask on an adjustment layer. Rather than sliding the black and white points, I typically make changes by sliding the gamma point around. The changes are small, rarely more than 0.15 in either direction. Too much and the effect is obvious, and gross. Subtle is better in this case.

Constant checks with the print preview and gamut overrun using the ICC profile from Blurb makes sure I'm keeping everything in the range of what the printer can handle, and that I'm preserving detail in shadows and highlights.

This is also when color correction is done. "Blue snow" is a common case when a camera's auto white balance can be fooled into making a strange color in an image. Large patches of snow in a wintery scene can result in the snow taking on a blue cast from reflected sky. The Photoshop Color Variations tool can be used like the color correction wheels on a color enlarger to dial in the colors and remove effects like blue snow.

Eventually a usable image is produced. The image is saved as a multi-layer Photoshop file, and then as a flattened JPG file. Blurb, as with most printers, expect a JPG or TIF file in the sRGB color space. They won't take 16-bit files, or multilayer Photoshop files, or RAW, or any of a whole host of formats we take for granted during processing. Keep it simple. But save your working file in case things don't work out the way you expect!

The last step is to send things off to the printer and see how well you did. Even the best monitor calibration and ICC profile is no match for seeing the final image in print. That's when you get to find out just how well you did.

Submissions for the WWKW 2009 book open in three days. This new workflow is just starting to become routine for me. By the time I've gone through the hundreds of images submitted for WWKW 2009, I expect it will be second-nature.

Tom

Friday, April 17, 2009

WWKW 2009 - And Counting

So there are six days 'till WWKW 2009, and I just noticed my shutter servo tab has cracked, and my rig is inoperable.

AAAAAARGH!

I'm glad I saw it, though. Nothing worse than to have gear fail in the air and not know it. The way it happened was this:

One of my leg brackets wound up getting bent about a week ago, resulting in a crack almost completely through the bracket. Rather than risk failure, I ordered a new pair of leg brackets from Brooxes.com along with a folding Picavet.

The folding Picavet really is a thing of beauty. It comes in two pieces, and replaces the stock single-piece Picavet. By pulling a quick pin, you can rotate the top half of the Picavet, which can then be rotated parallel to the rig for some really compact packing. Even better, in the packed position there's a second hole to insert the quick pin so that it locks the Picavet in that position, too.

After installing the new Picavet, re-lacing my Picavet lines because I couldn't keep the pulleys straight, removing the broken leg bracket and my old antenna bracket, and installing the new leg bracket, I finally saw that my shutter servo was sitting at some oddball angle. Oh dear...

I toyed around with the idea of building a new bracket, and even bent up some of my 1/64" brass strip stock while testing out a couple of ideas. In the end I went back to trying to get the GentLED-CHDK2 working with my camera, a Canon Powershot A650IS. When I first got it I had no end of problems trying to get the A650 to recognize the timing coming out of the CHDK2. Things hadn't improved, and more testing only indicated the same behavior.

As a last ditch effort, I asked to see if I could exchange the CHDK2 for a GentLED-CHDK, which does work remarkably well with the A650IS. Hoping for a positive response.

If not, I've got enough Atmel ATtiny hardware lying around to make a fair simulacrum of a GentLED-CHDK, and can get that built before WWKW 2009. And failing that I'll sacrifice another shutter servo to my existing bracket, knowing full-well the mounting tab will likely crack off within the year.

I'm not giving up on my chances of doing massive KAP during WWKW 2009, but I hate having set-backs this late in the game. It makes me paranoid.

Tom

Wednesday, April 8, 2009

Plans for WWKW 2009

World Wide KAP Week 2009 is two days away, slated to start Friday, 24 April, 2009, and extend through 5 May, 2009. My own planning started months ago when the dates were first announced, but as things are getting closer, plans that were once vague maybes are firming up into certainties.

Gear -

I've made a number of changes to my gear, and added some new things to the bag:

4x5 - Lots of changes since the last WWKW, but nothing new in months. All my film holders are loaded, so it's a matter of bringing it everywhere I go, and hoping for dead smooth wind. Maybe also bring some counterweights to try to damp down the remaining vibrations.

Archaeo - I've got a new rig this year that's all ready to go: my ortho-only archaeology rig. This is the rig I also plan to use for garden photography, so that may be where I use it during WWKW 2009.

RC Rig - I've changed out my radio to a 2.4GHz Turborix radio, and added a prototype for an AutoKAP controller that's now under development. Both will get field time this year. I'm also adding a folding Picavet to make packing my rig into my bag even easier.

Anemometer - I bought an anemometer from one of my print sales, so I can keep better track of the conditions in the field, and go back later to compare with the wind models.

Winder - My new winder is now a thoroughly tested and battle-scarred denizen of my KAP bag, and will be my primary winder for the duration.

Locations -

Kohala:
  • Pu`ukohola Heiau
  • Spencer Beach Park (New)
  • Kawaihae Canoe Club (New)
  • Archaeological Sites along the Kohala coast
  • Lapakahi Historical Monument (New)
  • Mahukona + Lighthouse (New)
  • Decay north of Mahukona Park (New)
  • Mo`okini Heiau
  • King Kamehameha's Birthplace
  • Wind Farm (New)
  • Lighthouse
  • Pololu Valley
  • Next valley over (New)
  • Kohala Architecture & Sugar Camps (New)
Kona:
  • Hapuna (New)
  • Beach 69
  • Petroglyphs at Puako (New from kite)
  • Anaeho`omalu Bay
  • King's Trail Petroglyphs at Anaeo`omalu Bay (New)
  • Scoria Abrader Quarries at Waikoloa
  • Waikoloa Beach Resorts (New)
  • Beach Without a Name
  • Wainanali`i Pond and Kiholo Bay
  • Kua Bay
  • Anchialine Ponds along Kona Coast (New)
  • Puako Tide Pools (New)
  • Pu`u Waawaa
Hamakua:
  • Waipi'o Valley
  • Akaka Falls (New)
  • Old trestles on the Hamakua Coast (New)
  • Gulch bridges on the Hamakua Coast (New)
  • Laupahoehoe Point (New)
  • Kalopa (New) (MAYBE)
  • Boiling Pots / Rainbow Falls (New)
  • One of the beach parks, but this is tough (New)
Waimea / Central:
  • Various sites in Waimea
  • Mana Road
  • Whistling Canyon (New)
  • Waimea Resevoirs (New)
  • Head of Waipi`o Valley (New)
Mauna Kea:
  • Mauna Kea Summit (New - Heiau has fallen over)
  • Lake Waiau (New)
  • Adze Quarry (New)
  • Cinder Cone Edge (New)
  • Domes & Dishes
Saddle Road:
  • Pu'u O'o Kipuka Trail (New)
  • Pu`u Huluhulu
  • Hunter's Cabin (New)
  • Silverswords on R14 (New)
  • Pu`u near Kilohana Check-In (New)
South Point:
  • South Point
  • Green Sand Beach (New)
  • Whittington Park
  • Across the bay from Place of Refuge (New)
  • Kealakekua Bay at Captain Cook's Monument (New)
  • HOVE - Road to the Sea, turn north (New)
  • Na`alehu (New)
Puna:
  • Kapoho Tide Pools (New)
  • Kalapana
  • Hike to Pu`u O`o ? (New)
Volcano NP - Contingent on permission from NP Director:
  • Ka`u Desert Trail; Footprints in Lava *
  • Chain of Craters Road Trail *
  • Great Rift? * (This may be hard to get to)

Goals -

I want to fly in new locations, most specifically on the south side of the island. There really is architecture down there, along with the Kapoho tide pools, sites inside and outside of Volcano, Green Sand Beach, and a couple of archaeological sites.

I'd like to fly some of the less-likely locations for Hawaii. I've done tons of beaches. I want to show off cattle country, desert country, volcano country, etc. There's more to Hawaii than beaches. Besides, this is a vitally important component of the book I'd like to do later in the year of low altitude aerials of Hawaii.

I want to take some 4x5 photos from the air. I'd really like at least one rock solid image I can print at wall mural size. If I can get a couple of solid ones, I'll be happy as can be.

Mostly I want to have fun, knowing that all over the world people will be flying cameras on kites anywhere and everywhere they are, and sharing what they see. I was blown away by the output for WWKW 2008. I can't wait to see what happens for WWKW 2009!

-- Tom

Thursday, March 5, 2009

I Bought an Anemometer

Not the most original title for a journal entry, but it's the truth. I did buy an anemometer: a Kestrel 2000. I can't say it was an impulse buy since I've been looking at anemometers a little over two years now. But the decision to get it right now? I suppose it was.

I've been holding off on getting an anemometer, probably for relatively silly reasons. I can gauge wind pretty well, and the wind models I use are generally accurate. But I've had a bad run of KAP outings recently when I would either find the wind too weak to lift a rig, or too strong to safely fly a kite. I've come home frustrated, more often than not, with little to show for it. An anemometer can't make the wind blow, but it can send a very clear message if the conditions just aren't right.

Another reason I decided to buy one now is that World Wide KAP Week 2009 is coming up soon, and given my finances, there's no way I'd be able to get the kite I've been lusting after (a Dan Leigh Trooper RS, a high-wind delta that can handle the conditions at Upolu Point and Southpoint.) So I took what finances I had and put them into the anemometer. It will be a useful addition to my KAP bag, and if it gives me better awareness of the flying conditions at a new site, it could make the difference between a successful outing and one that ends in damaged or destroyed gear. A sound investment.

But in a larger sense, it was just plain time. A few years ago, if asked which hobby was actually paying for itself the best, I would have to say machining and writing. I wrote a couple of articles for a machining magazine, and when they were published it was the first time I'd ever actually had a hobby start paying for itself. The funny thing is, I used the money from the articles to pay for KAP gear rather than tools. To be fair, I did use some of it to buy two years worth of subscriptions to my machining magazines, so it did go back into the hobby to some extent. But it also helped bootstrap the next.

Today a more truthful answer is that it's KAP that's bringing in the most real revenue. I've sold a number of prints, and have had requests for even larger prints though so far none of those requests have led to a sale. Nonetheless, my KAP hobby has done more to pay for itself than any other hobby I've got. By and large, about half my KAP gear was paid for by print sales. This is better than my ratio for my machine shop and those magazine articles, but it's still not a hundred percent.

I'm at a bit of a leaping-off point. Selling prints is great, but I don't really think that's where I'm going to stop in terms of letting KAP pay for itself. I'm looking at other options, most of which involve doing paying gigs. But for paying gigs to work out, there has to be some guarantee that the flight can happen, and that the right tools are used for the conditions. And for new kite purchases to make sense, I have to be able to demonstrate to myself that I've been asked to fly in conditions that would warrant them.

Hence the anemometer. WWKW 2009 is an excellent opportunity to characterize each of the kites I use, to find out the wind range for each one where I can reliably lift my KAP rig, and the line angle I'm likely to get. This information can be used to plan shots in ways I've never been able to do, and to provide a better guarantee that I can pull off a particular shot.

My goal, by the end of WWKW 2009, is to have a portfolio of images to show to prospective clients, and the ability to know if my gear can pull off a shot that a customer asks for. The anemometer is the first step in this direction.

There's an old adage that as soon as you start to do a hobby for money, it turns into a job. The point, of course, is to keep hobbies as hobbies and to keep work as work. But I'd be lying if I said I wasn't thrilled when I saw my first article printed in that magazine. It only made me want to write more. And when my first KAP print sold, I was ecstatic. I hope this next leg of the venture works out as well.

-- Tom

Thursday, February 26, 2009

KAP Aerial Archaeology and Photosynth

I've been frustrated in my attempts to turn KAP aerial archaeology shots into maps, or even into stitched panoramas that bear any real resemblance to reality. For the most part this is because I really don't have the right tools to get the job done. Photogrammetry, and georectification are not simple to do, and most consumer-level tools aren't suited to the job. Typically this involves software like Photomodeler, Leica Photogrammetry Suite, and GIS software. It also typically requires a grid of GPS-referenced ground control points. I'm not positioned to do any of these things.

But I did run across something that produced oddly pleasing results. It's a Microsoft R&D product called Photosynth. It's not so much stitching software (though I use Microsoft's Image Composite Editor for most of my stitching these days). Instead, Photosynth assumes the shots are not taken from the same vantage point, and finds all the points at which the shots overlap. This in turn is used to create a point cloud of intersection points between shots, so a viewer can place themselves at any of the vantage points from which the photos were taken, and see the scene from that point.

All of which is hard to say, but a lot easier to see. I tried this out using one of the flights from last weekend:



It worked remarkably well. Even more fun is to view it with the images turned off so you only see the point cloud of the intersection points between all the shots. Photosynth does a really good job of picking up the three-dimensional structure of the scene.

I tried it with some of my other KAP flights, just on a whim. The archaeology flights were almost all orthogonal to the ground, but most of the time when I'm out doing KAP I'm busy taking oblique shots rather than orthos. I was amazed at how well Photosynth dealt with this:




I think there's a lot of potential for what Photosynth is doing. There are all sorts of other features I'd ask for (and have asked for) beyond the existing Photosynth interface. Among others, it's ideally suited for picking out proper stereo pairs from a whole collection of images. It would be wonderful to have them all gathered, arranged, cropped, and ready to go.

Even better would be 3D model extraction from a random collection of photos of a scene. One KAP flighti n particular resulted in a really good point cloud for a large truck: a deuce-and-a-half. I'd love to be able to export that truck as a 3D model.

I don't know how much time I'll devote to this, but right now it's pretty fun. And for now walking around an archaeological site using my aerial still shots is as close as I can come to what i want: real 3D scene reconstruction.

-- Tom

Monday, February 23, 2009

More Archaeological KAP

I got to do some more archaeological KAP last weekend. The archaeologist I worked with on Ahu A`Umi was on island a few weeks ago, and tried to fly his camera over the lava flows south of the resorts at Anaeho`omalu Bay. It used to be an active quarry for scoria, which in turn was made into abrader tools for working wood, bone, shell, etc. He wanted to see if the quarry and abrader workshops were visible from the air. Foul weather made for a disastrous KAP trip, so I offered to give it a try the next weekend. We traded some email planning out a strategy (wind direction, time of day, angle of light, etc.) and Saturday morning my son and I headed out.

The wind was certainly better than what he'd faced the previous weekend, but it was still turbulent, tossy, and a little too strong. With some reluctance I put up a Flowform 16 and 15' of fuzzy tail, and got down to business. The nature of lava makes it difficult to walk a consistent grid pattern, and I was reluctant to plant my big feet in anything of archaeological value that might be fragile. Nonetheless I spent about an hour and a half pacing around and chanting my mantra: "C'mon baby, steady wind steady wind. C'mon, don't no NOnono... ok, better, c'mon baby..." (I wonder if I'm the only KAPer who talks to his kites.)

At the end of that first session I took my son to a coffee shop at the nearby resort. He had milk and an apple turnover while I sipped coffee and reviewed my shots. As it turns out I'd used the wrong camera settings, and this combined with the turbulent wind resulted in over half the shots being too blurry to use. But there were enough sharp pictures to convince me to give it another try the next day.

The strangest part of the whole experience was that the entire time I was reviewing the pictures I'd taken, the sound system in the coffee shop was playing, Tex Ritter's "I Got Spurs That Jingle Jangle Jingle". I kid you not.

The next day the wind was much more laminar and of more reasonable speed. My son wasn't interested in going this time, so I took my daughter, who brought a book. This isn't unusual for her. What was unusual was that she only brought one. Normally a quick trip somewhere involves at least a small stack of books. "I never know what I want to read when I get there!" pretty much sums up her reason for this. I've quit asking.

The second day went much better than the first. I launched at a second site the archaeologist had marked out, and moved from there back to the original site I'd flown over the previous day. I found a better spot to fly from, and got a better grid the second go 'round. Toward the end of the session the wind was dying down, so I walked the rig back to my launch spot, where my daughter was still reading. I think she looked up maybe once the whole time. So much for getting out to fly kites with your kids. At least she guessed right on her book. It was what she wanted to read.

My daughter had heard of the coffee shop outing with my son the previous day, and was eager to get the same treat. So back we went, my daughter toting her book and me toting a camera with what I hoped were some better pictures than the day before. This time we both wound up getting chocolate croissants, which we munched while she read and I reviewed my pictures. What I saw put a smile on my face. Not only did the steady wind and the correct camera settings work out, the pictures very clearly showed quarry sites, abrader workshops, and even some petroglyphs I hadn't seen from the ground. Euphoria!

But as I continued to flip through the pictures it slowly dawned on me that the coffee shop was still playing, "I Got Spurs That Jingle Jangle Jingle". They didn't miss a beat.

Those two days generated about 3.5GB of data, probably 2GB of which is actually any good. I sent the disks off this morning in the hope they'll be there some time tomorrow. Meanwhile I've got "My Spurs Go Jingle Jangle Jingle" so firmly embedded in my head that even two hours of Linkin Park couldn't clear it out. I've given up. I fear that the very mention of scoria abrader quarries will forever summon up the jaunty notes of "I Got Spurs That Jingle Jangle Jingle":

I got spurs that jingle, jangle, jingle
As I go ridin' merrily along
And they sing, 'Away, too glad, you're single'
And that song ain't so very far from wrong

Oh Tex...

-- Tom

Sunday, January 4, 2009

Root Cause Analysis

In a previous career, I worked as a systems administrator for twelve years. When working with computer problems, one of the first things you need to learn to do is to see past the symptoms, and treat the underlying condition. It's something called "root cause analysis", and it's the backbone of computer troubleshooting.

In my present career, I do instrumentation at an international astronomical observatory. When diagnosing and maintaining scientific instrumentation, one of the first things you need to learn to do is see past the symptoms and treat the underlying condition. Root cause analysis again.

Recently, my KAP rig has suffered from an oddball problem: Each time I try to rotate the pan axis, there's a lag, and then it starts moving. The first time this happened in the field, I swapped batteries and went back into the air. When the symptoms didn't go away, I assumed both sets of batteries were low.

In a later flight I realized the small pinion in my pan axis gear reduction was loose. DOH! Without thinking about it, I tightened the screw that holds the pinion in place, and put it back in the air. Problem solved, right?

As it turns out, wrong.

About the third time I tightened that screw, I realized something else was wrong. I hadn't done proper root cause analysis; I'd just treated the symptoms of a loose screw. A little more poking around revealed the culprit: The pan axis itself was binding. This put additional torque on the pinion gear, which put extra force on the screw, which pulled it loose, which made the pan axis have backlash, which made for the lag when I tried to rotate it.

ARGH!

The cause of the tight pan axis was a bushing that was too tight in its hole. I made a new one with 0.005" clearance, and everything ran like a well-oiled clock. I also used thread lock compound on the pinion gear screw, so it shouldn't get loose as easily. For good measure I also used a synthetic grease on the two bushings on the pan axis, so hopefully this will be the last time I deal with this problem.

Which is the real benefit of root cause analysis: never having to see a problem come back.

Time to go fly a kite!

Tom