No announcement yet.

photographic equipment requirements

  • Filter
  • Time
  • Show
Clear All
new posts

  • photographic equipment requirements

    I'm in the process of upgrading my photographic equipement (1 FF camera+1 sharp fixed focal wide angle), that I currently use for small projects with Zephyr. Typical project is photogrammetry of ancien stones (granitic) looking for time-erased engravings. So dense point cloud and mesh quality are of importance in my projects. "quality" being refereed to as the accuracy and precision of the position of points in the dense point cloud.

    I've already read the tutorial on that cameras and have few questions.

    One is linked to the size of the pixels, ie. it is advised they are larger than 2 ?m.
    is this advice a way to look for a signal in the image that has the lowest noise level possible, ie. a high S/N ratio ?
    is this advice linked to the diffraction ? Or are other considerations linked to this adivice ?

    another question: In the cameras we find today, we see a great dynamic range, up to something like 14/15 bits. How the dynamic range might influence the quality of the dense point cloud quality ?

    another question: a sharp lens is also advised. My primary understanding is that the lens shall exhibit high FTM curves. Then, how the micro-contrast is of importance in the quality of the dense point cloud ?

    Last question: are they other consierations that play an important role when looking for high precision/accuracy position of point cloud, at leat from taking the picture point of vue ?

    Best regards,

  • #2
    Hi AroBass - Welcome to the forum. To answer your questions:

    - Regarding accuracy, we have a theoretical calculator which can be found here: - There are a lot of variables at play which can contribute to the accuracy of your 3D digitization, but for subjects which are feature-rich like ancient carved stones, your accuracy will typically be much higher than when shooting non-ideal surfaces such as homogenous or reflective media.

    - I have a working knowledge of cameras and squeeze performance out of the ISO and exposure values, keeping ISO down low reduces the level of sensor noise in your images, but exposure is critical to the final result. I use an Xrite ColorChecker passport (v2) and ensure the 18% grey panel is smack bang in the middle of my camera's histogram to make the most of the A7Riii's dynamic range.

    - Having a wide dynamic range is useful if you're processing your images from RAW to JPG prior to input into Zephyr as it enables you to bring down the highlights and boost the shadows prior to conversion to 8bit JPG. There is minimal gain from using 16bit images when it comes to processing, and considerable storage issues can arise when taking a large number of images (processing times are longer too).

    - MTF charts are a bit out of my comfort zone, but through real-world testing, I have found that using high F-Stops with diffraction is a relative non-issue when it comes to photogrammetric processing. It's far more important to have your subject in deep focus than the minor loss in sharpness across the frame as a result of diffraction.

    - Micro contrast in your images will come through with the mesh filters, but won't really make much difference to the dense cloud.

    - Roberto or Andrea will be able to expand on the technical 2micrometre query and another other deeply technical accuracy questions you may have.
    Last edited by cam3d; 2020-10-07, 04:02 AM.


    • #3
      Thanks Cam3d for those elements and for sharing your experience.

      - I new about the calculator but found it painfull to use, since it is trial and error. I would appreciate to have the formulas to choose my best startegy with regards to accuracy by plotting directly the resutls.

      - I have to admit I did miss using the colorchecker (I use the classic one mostly for white balance). good point here.

      - My use of Zephyr has been using raw format and yes noticed lengthy processing time. I expected better quality in the ouptut compared to jpeg. I'm not sure to understand why there is a minimal gain from using 16 bits images. Anyway, I take into consideration that for my next shots and with regards to my camera I will not focus on the 16bits side.

      - well noted for the mesh filters. I'm still learning on the setailed usage of the parameters in the photoconsistency mesh optimisation.


      • #4
        Hi AroBass ,

        I also would like to share some of my personal experiences:
        - if you are planning to buy a super compact cam, try it out, rent one first, they could overheat and overheating sensor could generate incredible annoying issues, as bad alignment.
        - avoid cameras that can produce "shutter socks" it can be a marginal issue due to classic photography, but could destroy your overall PG results, no matter what sharp lens, great sensor is struggling under that shutter shock.
        - Take a very close look at luminance colour noise, luminance channel is one of the cardinal data when depth map is calculated.

        I was lucky enough to experience all these issues by buying my Canon M6 mark II, that seemed a perfect PG camera on paper (except the pixel size, in that filed it lacking even on paper haha)
        By capturing the same object (did the test in warm environment) with a a7 mark II vs my Canon M6 Mark II under the same light conditions, same focal length same angles, same image number the result was not a little bit better on a7 side, but it was day-and-night!

        So sharp lens, modern sensor, and excellent reviews due to classic photography usage could still end up in terrible results due to PG.

        After you made your cam candidates, I suggest to: rent, borrow and try, compare.


        • #5
          Hi, thanks for sharing your experiences.
          - renting trying and comparing before bying is indeed a good idea. I will check for this option.

          - you mentions that «luminance channel is one of the cardinal data when depth map is calculated». Can you elaborate on this point ? That would mean I need to check different camera, taks some shots alls being equals in the shotting conditions, then convert raw to black and white and look for luminance noise, as one of decision criteria ?

          - In the same vein, has anyone ( cam3d ?) tested 4x pixel shift with their camera and compare results with and without ? (I do not mention 16x pixel shift). Does it improve the accuracy of point cloud and mesh ?


          • #6
            you should know well your camera(s)
            keep in mind that if you shoot in a low light environment the noise given from hi iso will be high, but if there is plenty of light you can go high with iso and have very low noise!

            as long as we feed zephir with jpgs, just use a camera that ouputs good jpgs (pentax, fuji...)
            I use 2 pentax cameras (apsc+ff) and use their jpgs straight out of the camera...I know that jpgs from my old apsc are "good" until iso1600-3200, but for pg I prefer to stay under 640, on the ff jpgs are "good" until 12800-25600 but prefer to stay under 2000.
            a neutral background in a controlled light environment, custom wb, manual exposure, tripod + remote, accurate focusing, good depth of field and many many photos are the keys for the success at first attempt
            I use jpg setting very flat, maybe contrast-1 and sharpness+1...that's it

            if I was shooting outdoor I would probably use masquerade