Different ways to measure/calculate frontal area (A)
 

This has come up in a few threads, decided to start a dedicated one.

The old trick was to photograph the car with a telephoto lens, cut the car out of the photo & weigh the cutout vs. the cutout + the rest of the photo. Using a known measurement on the car in the photo, area can be derived using the proportion of the photo's weight.

A more modern method is to take a digital photo, trace the car's outline & use a pixel-counting program to get the image vs. vehicle proportions.

The problem is, you need a very strong telephoto to make this work. 3x optical isn't enough (I already tried, and it was quite off, when comparing the known length of the front plate to the length of a wiper blade by counting pixels.)

I think a more accurate (but harder) way to do frontal area is to back the car up to a wall, then take a strong flashlight far away and point it at the car. Trace the outline of the shadow on the wall. Then photograph it and do the pixel counting.

EDIT: ideally you would use the sun at sunrise/set, but good luck finding such a wall/unobstructed sightline.

Using a laser level has also been suggested as a way to trace the vehicle outline.

2 points on the digital photo/pixel counting method:

1) I've already tried this, and can confirm that there's way too much distortion with a 3x optical zoom.

Solution: stick the lens of the camera against a binocular lens. I tried this, and think I can make it work. (Focus is the difficult part - I have a somewhat fuzzy image from my first attempt.)

2) Photoshop can do the pixel counting part, once you've made it a 2-colour format. In Photoshop 7.0...

> Image > Histogram

Or make a big one of these:

http://blog.pcnews.ro/wp-content/pho...2/30304-09.jpg


:D

Here's a good illustration of the amount of error you'll be dealing with if you use a camera without enough optical zoom to measure frontal area:

Images resized so the width of the hood between the headlights is the same in both pics...

http://ecomodder.com/forum/attachmen...6&d=1196221942

Left: 2x optical zoom

Right: 2x optical zoom shot through one side of a pair of binoculars from about 150 ft further away. (Sorry, don't know the power of the binocs).

If you can see the rear tires in the photo, it's a sign the zoom isn't strong enough. You can make out about 1 inch of the inside of the rear tire in the right pic, which isn't bad, considering the rear track is a total of 1 inch narrower than the fronts anyway.

If I could get a crisply focused version of the right pic, I think it'd be good enough to use the outline-and-pixel-counting method.

HOWEVER: someone told me that you can just move waaaay back, and keep the vehicle in the center of the lens where there is much less distortion.

That's interesting - didn't know that. So I should try the binocular thing again, but don't try to fill the view right up? Go back another 150 feet and keep the car in the middle of the view? I'll have to sort out getting it to focus properly first - or it'll just be a fuzzy blob.


If I were doing the light-at-a-distance-trace-the-shadow-on-a-wall method, I'd do something like: move the light source a measured amount to keep it as perpendicular as possible to the wall where that area of the shadow is being traced:

Here's another method:

1) position the car parallel and close to a wall
2) make a second cardboard "wall", 90 degrees to it, at least half the width of the car
3) roll the car back and forth through the cardboard wall, trimming as you go
4) photograph the results (edit - no telephoto lens required, since you're taking a pic of a flat surface) & do the pixel counting thing

eg:

http://ecomodder.com/forum/attachmen...8&d=1196222341

Hey, I'm not saying it's practical/easy - or more fun than punching the car through a snowbank. Just an idea.

You could also measure the cutout with a planimeter.

I just took a pic with the 2x optical zoom from waaaaaaay back, and sure enough there's far less perspective distortion in the image than in the pic above where I filled the frame with car @ 2x.

There's still more than the binocs + 2x version, but much less than the 2x photo above (you can't see as much of the rear tires).

So... if folks have a digital camera with a lot of resolution, this may be another option: max out the optical zoom, and go waaaaay down the road before taking the pic. Then crop the resulting image.

Mine's only 2MP, so it won't really work. To get the rear tires to disappear in the image, the car is only about 50 pixels wide in the resulting image :o

So I will have to stick with the binoculars + camera method.

2x optical zoom, from close up, medium & faaaar away. Look at the rear tires in each to see diminishing perspective "distortion". (Or the apparent height of the roof line.)

http://ecomodder.com/forum/attachmen...9&d=1196222586

FYI, completed the task for my car: http://ecomodder.com/forum/showthrea...etro-2439.html

i think it's always worth t looking for some blueprints of the car, a site i recently discovered

The-Blueprints.com - 26351 blueprints online has a lot of usefull stuff they often include a frontal view
sometimes even with measures but if not these can be found and applied the the drawing

if you're working with pictures and you know the with and height of the car, you could set it to the correct width in photoshop and that stretch/squash the height untill this matches the scale of the width... this still leaves room for errors but i imagine if you match a crappy 3 view drawing with a picture and real measures and than do a pixel count your arrive pretty close to the real value

here's a few of the best ones for the guys


Suzuki Swift Mk2 1600GLX 4WD Saloon

Suzuki Swift GS 3-Door

You've discussed the roll of perspective projection distortion, but no one seems to have considered optical (lens) distortion.
In the context of the above quoted material, it can be taken it as a given that your digital camera's integrated (not accessory) zoom lens would be considered low-end telephoto lenses. With virtual certainty -

• Non-zoomed, you have some amount of barrel distortion.
• Somewhere in the zoom range you have some mustache distortion.
• At the far end of zoom you have some amount of pincushion distortion.

Recent versions of PaintShop Pro ( and I believe Photoshop, Photoshop Elements as well) have manual tools to correct for pincushion and barrel distortion. There are commercial plug-ins available for the same programs that provide more automatic correction based on camera/lens model (but probably don't include you consumer grade digital in their lists.)

It's still winter here and I'm waiting for warmer weather. When that happens, I'll find a wall to back the car up to such that there is a long, level, unobstructed expanse in front of the car to position my camera shot. Prop a 4x8 sheet of plywood against the wall - vertical but in landscape position orientation. Position my tripod mounted camera* such that the plywood sheet fills about 90% of a picture taken with camera set to maximum Safety Digital Zoom (no in camera pixel interpolation - only cropping) at 1600x1200 resolution. This picture will be used as a reference to determine the best numeric setting to use in the pincushion correction filter.

* Tripod mounted camera: To minimize vertical perspective distortion, height of camera lens center to be 1/4 (height of car + 48")

Moving on to the picture of the car. As controls

• A small object (soup can?) will be positioned on the roof of the car at the point of it's maximum height.
• Measured lengths of fluorescent flagging tap, will be taped to key points on the car's body.
• The measured distance between outer tips of the side mirrors will also be used as a control.

These controls will be used to decide any application of perspective correction after application of lens correction.

When I actually do this, I'll post a thread with pictures.

Or, the low-tech approach (Mechanical skills required: the ability to bang nails, or apply duct tape). Build a simple rectangular frame high enough and wide enough to fit around the widest/tallest points of your vehicle. Calculate the area of the enclosed rectangle. Take a series of measurements of the gap between your vehicle and the frame (the more measurements you take, the more accurate your result). From there it's relatively simple to calculate the difference by various means, either by 'filling in' the space with progressively smaller cubes, or by creating an analog in a drawing programme.

Sounds like a variation of the "cardboard cutout" idea in post #7.

TestDrive: I'd be most interested in the amount of error in area calculations between corrected and uncorrected photos. Sounds like an interesting test.

One drawback of the "cardboard cutout" method is that it doesn't allow you to measure the area under the car, which is a big ol' chunk of space.

MetroMPG, no doubt well south of 5%. Think I'll refrain from any other predictions.
davidgrey50, due to ground effect and all the usual clutter underneath, the conventional wisdom is to include it when measuring frontal area.

Hi,

I would use DataCAD to trace a polyline on the telephoto picture, and then measure the area with the handy Measures menu...report it in square meters or square feet; no problem. Here's two examples using the Aptera; the one on the left is a perspective distorted rendering, the one on the right is a pretty long shot of an actual vehicle:

http://img.photobucket.com/albums/v7...alAreaRev2.png

Using the area from the long photo, the Aptera 2e has a CdA of 0.277 meters. (The hatch grid is 12" squares, btw.)

What would the numbers be if you had corrected the picture for lens distortion prior to working in DataCAD?

I took the long shot and based on the pole (tree?? Note curve relative to the straight blue line I added) on the left and corner of the building on the right, I applied correction for pincushion distortion. Would the DataCAD numbers come out the same for the corrected shot?

http://i490.photobucket.com/albums/r...Correction.png

Now, note the curve of the road relative to flat... roads are always tapered down from the center, so it should be slanted down on the left of the pic, but it should also be either nearly flat, or curved upward toward the car, not concave.

You're right, Christ.
My guess-timated lens correction is obviously incorrect.

I wouldn't say it's incorrect, I'd just say that it brought focus to another problem... if you correct one thing, does it make the perceived fish-eye effect worse to other parts of the photograph? Should you then correct more for everything at once? Man... setting CRT monitors up is sooo hard sometimes.

I haven't tried it, just read this page and thought about it, but wouldn't it be possible to do a pixle count of an image of a car's silhouette, like say by putting a stretched white sheet in front of the car between it and the camera. If you do it at night and use bright enough flood lights far enough behind the car to get an outline underneath and all the way around, and you keep the camera and the lights both at the centerlines of the car's height and width, it should be the same as the cutout, zero distortion, right?

After much searching I couldn't really find a relevant frontal area number for my 2005 Accord, so I found technical drawings on the-blueprints.com that lunarhighway mentioned above. They're not the exact same trim (I don't have the foglight recesses or the side mirror reflector/indicators), but it looks close enough.

Put it into Photoshop, selected the area of the car, then looked at the histogram (in CS5.1, Windows -> Histogram) for the number of pixels. Oddly, the number of pixels reported in the histogram is significantly less than the actual number of pixels you get when you multiply the dimensions of the image, however after adjusting for that, I got 23.8 square feet as a frontal area, which isn't far out enough to be obviously wrong at least.

Removing the side mirrors apparently lowers this by 2.5%. Neat.

http://i78.photobucket.com/albums/j1...ps68085f1a.jpg

CAR and DRIVER's 'DRAG QUEENS' frontal area
 

In C&D's June,2014 article about the Cd comparisons they did,Don Sherman commented that they photographed the cars from a distance of 150-feet,with a 200mm camera lens,then imported the digital image into Siemens Solid Edge CAD software to ascertain the frontal area.

MetroMPG had to crank the handle on this page for a whole page before it took off...

I like this best. A turn-out on Highway 101 at sunset, and a 4x8 sheet of white coroplast (folded in half for transportation). Use an East-West shadow to judge exactly when the light is horizontal. Trace 1/2 on one side and the other on the obverse.


Then fill it in with a Sharpie, photograph it, put it in the GIMP, crop it to a bounding box, blur it to infinity, sample the resulting gray, convert the RGB to a percentage and take that area of the bounding box. Done. :thumbup:

When I was campaigning my salt flats race car, I wanted to know it's frontal area, so I used the shadow method. At the time, I didn't have a way to drive it (not street legal and/or engine not in it at the time), so I had it on my flatbed car trailer.

Then I took a sheet of particleboard and built a simple brace so I could stand it up at the back of the car trailer behind the car. The board and brace was held up and held together by c-clamps and "quick grip" clamps. Then I went to a place I could park it facing the setting sun and jacked up the hitch until flatbed of the trailer was aiming at the right angle up (the sun was still a few degrees up).

I used the symmetry of shadow on the board to help get the aiming right. Once it was right, I quickly traced the outline of the shadow on the board with a fine tip marker.

Then I drew a grid on the board, spacing the horizontal lines one inch apart. Then I just carefully measured the lengths of each of those lines, wrote them on one side of the shadow outline, and added up the numbers. The result is what I used as the frontal area of the car in square inches. ;)

....To get square feet, just divide by 144. :thumbup: