Photography – April 2021
Selection of cameras and lenses
17 Co-ordinated Points [T1 to T17] positioned to explore the possibilities with PaMe.

The ideal is for the points defining the measurements to be in the proximity of the baseline defined by the two panoramas, with a good ‘angle of cut’, but not along the axis of the baseline.
The targets were position for some to be in the ideal locations and some not so.
T11 and T18 are on the axis of the baseline and if either of these points were selected PaMe returned the message “Accuracy results can be low due to small Parallax” indicating the points should not be used. The same message is displayed if a wrong point is selected (e.g. T2 and T1 in one image and T2 and T3 in the other).
T1 does not have a good ‘angle of cut’ which is made worse but its elevation as the ‘angle of cut decreases as the point moves away from the horizon through the two panoramas.
The line T10 to T12 is considerably longer than the baseline with reactively small ‘angles of cut’ and showed where the Equirectangular images were not as accurate as they could be. 

The accuracy of the Equirectangular images create with a Fisheye Lens + DSLR + Panohead was proven by using them to add real world colour to the Point Clouds generated by the terrestrial LiDAR Scanners introduced towards the end of the 20th century. The ‘standard’ equipment was a Sigma 8mm Fisheye Lens + Canon or Nikon DSLR + Nodal Ninja NN3 and the results for I (Nikon D5300 + Nikon 10.5mm + Nodal Ninja NN3) and II (Nikon D5300 + Samyang 8mm + Nodal Ninja NN3).

For this exercise I used equipment I had to hand so it is by no means definitive, but I hope provides idea of the possibilities. Although the results from the duel lens cameras are ‘not so good’ these cameras were not developed for measurement and the results show inaccuracies in the equirectangular images which are insignificant visually. These results do not rule out the use of single shot 360° Panoramic cameras for measurement as it is possible that other dual lens camera, such as the 51Mp Ricoh Theta Z1, may give better results and the NCTech iSTAR has proved its accuracy as it has also been used to add real world colours to LiDAR Scanner Point Clouds. 

If you already have the equipment to capture 360° Panoramas (e.g. a dual lens camera) it is a straight forward process to asses if it is suitable for your application:
Take two 360° Panoramas in the same horizontal plane with measurable objects in the locations you want (e.g. boxes, lengths of wood, marked items such as Ranging Poles);
Measure the distance between the two Panoramas or a known length (preferably similar to the distance between the Panoramas) or both;
Take off measurements with PaMe Light which can be purchased for the nominal fee of €8 for 100 measurements, which is more than enough to evaluate the technique.

Setup Four - Plan

Setup Four – Plan

Setup Four – Elevation parallel to the Panorama ‘Base Line’

Setup Four – Elevation parallel to the Panorama ‘Base Line’

Setup Four – Elevation at right angles to the Panorama ‘Base Line’

Setup Four – Elevation at right angles to the Panorama ‘Base Line’

Differences between measurements taken off with PaMe and measurements with theodolite and DISTO

Differences Colours

Variety of Cameras and Lenses

  I   Nikon D5300 + Nikon 10.5mm + Nodal Ninja NN3
8 shots round @ -10° + 4 shots round @ +30° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  17060 x 8530  –  good   Min: 0.05  Av: 2.11  Max: 7.07
B  Equirectangular =  17062 x 8531  –  very good   Min: 0.05  Av: 2.11  Max: 7.07
Protocol: Known distance between panoramas  =  4.774 m
Good results consistent with adding real world colour to LiDAR Scanner Point Clouds.
  II   Nikon D5300 + Samyang 8mm + Nodal Ninja NN3
8 shots round @ -10° + 4 shots round @ +30° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  13520 x 6760  –  very good   Min: 0.05  Av: 1.17  Max: 5.23
B  Equirectangular =  13526 x 6763  –  very good   Min: 0.05  Av: 1.22  Max: 5.59
Protocol: Known distance between panoramas  =  4.774 m
Good results consistent with adding real world colour to LiDAR Scanner Point Clouds.
  III   Nikon D5300 + Sigma 10-20mm @ 10mm + Nodal Ninja NN3
3 rows of 8 shots round @ +45°, 0° and -30° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  16764 x 8382  –  very good   Min: 0.02  Av: 1.22  Max: 5.91
B  Equirectangular =  16762 x 8381  –  very good   Min: 0.05  Av: 1.28  Max: 5.77
Protocol: Known distance between panoramas  =  4.774 m
Rectilinear lenses have a much smaller field of view (fov) than Fisheye lenses of equivalent focal length so three ‘rows’ were used to create the 360° Panorama. The results were good except for measurements including the target at T1.
  IV   Nikon Z5 + 7.5mm Fisheye + Nodal Ninja R1
4 shots round @ 0° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  11136 x 5568  –  very good   Min: 0.06  Av: 1.11  Max: 3.14
B  Equirectangular =  10766 x 5383  –  very good   Min: 0.09  Av: 0.79  Max: 2.08
Protocol: Known distance between panoramas  =  4.774 m
This lens showed distortions in the equirectangular image for measurements that included targets T1 and T10. This lens is considerable cheaper than the other two fisheye lenses and behaves differently as the NPP moves away from the front of the lens as the angle of incidence increases rather than towards the front of the lens as with most fisheye lenses.
  V   Insta360 ONE X2 + Nodal Ninja R1
4 shots round @ Tripod Bush stitched in PTGui 12
A  Equirectangular =  6126 x 3063  –  good   Min: 0.07  Av: 2.09  Max: 5.94
B  Equirectangular =  6120 x 3060  –  good   Min: 0.09  Av: 2.02  Max: 5.35
Protocol: Known distance between panoramas  =  4.774 m
These results were quite poor so this option is not recommended.
  VI   Insta360 ONE X2 - Parallel to Baseline Insta360 ONE X2
Lenses Parallel to Base Line joined in Studio 2020
A  Equirectangular =  6080 x 3040
B  Equirectangular =  6080 x 3040
Protocol: Known distance between panoramas  =  4.774 m
This arrangement provided a better solution than V, but not sufficiently reliable, and not as good as option VII.

Insta360 ONE X2 - Parallel to Baseline

  VII  

Insta360 ONE X2- Right Angles to Baseline

Insta360 ONE X2
Lenses @ Right Angles to Base Line joined in Studio 2020
A  Equirectangular =  6080 x 3040
B  Equirectangular =  6080 x 3040
Protocol: Known distance between panoramas  =  4.774 m
This was the best result using this camera. Orientating the camera at right angles to the baseline means that the weakest part of the equirectangular image where the two hemispheres are joined is aligned with the baseline where measurements are to be avoided anyway.

Insta360 ONE X2- Right Angles to Baseline

  VIII   Samsung Gear 360 + Nodal Ninja R1
4 shots round @ NPP stitched in PTGui 12
A  Equirectangular =  8266 x 4133  –  very good   Min: 0.09  Av: 1.76  Max: 5.13
B  Equirectangular =  8270 x 4135  –  very good   Min: 0.08  Av: 1.36  Max: 5.69
Protocol: Known distance between panoramas  =  4.774 m
This is the best solution with this camera, but leaves doubts about reliability where there are poor ‘angles of cut’.
  IX  

Insta360 ONE X2 - Right Angles to Baseline

Samsung Gear 360
Lenses Parallel to Base Line joined in Action Director
A  Equirectangular =  7776 x 3888
B  Equirectangular =  7776 x 3888
Protocol: Known distance between panoramas  =  4.774 m
Using this camera with the principal rays of the lenses aligned with the baseline produced the worst set of results in this setup and should definitely be avoided as a measurement tool.

Insta360 ONE X2 - Right Angles to Baseline

  X  

 Samsung Gear 360 - 90°

Samsung Gear 360
Lenses @ Right Angles to Base Line joined in Action Director
A  Equirectangular =  7776 x 3888
B  Equirectangular =  7776 x 3888
Protocol: Known distance between panoramas  =  4.774 m
Orientating the he camera at right angles to the baseline provided a little improvement to IX so should still be avoided as a measurement tool. These results are consistent with my investigations of the angular accuracy of this camera.

Samsung Gear 360 - 90°

First Set - Lines

First Set - A

First Set - B

Back to the Top of the Page

Different Arrangements

  I a   Nikon D5300 + Nikon 10.5mm + Nodal Ninja NN3
8 shots round @ -10° + 4 shots round @ +30° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  17060 x 8530  –  good   Min: 0.05  Av: 2.11  Max: 7.07
B  Equirectangular =  17062 x 8531  –  very good   Min: 0.05  Av: 2.11  Max: 7.07
Protocol: Known distance between panoramas  =  4.774 m
Good results consistent with adding real world colour to LiDAR Scanner Point Clouds.
  I b   Nikon D5300 + Nikon 10.5mm + Nodal Ninja NN3
8 shots round @ -10° + 4 shots round @ +30° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  17060 x 8530  –  good   Min: 0.05  Av: 2.11  Max: 7.07
B  Equirectangular =  17062 x 8531  –  very good   Min: 0.05  Av: 2.11  Max: 7.07
Protocol: Known length  T14 to T19 =  5.204 m
Good results consistent with adding real world colour to LiDAR Scanner Point Clouds.
  II a   Nikon D5300 + Samyang 8mm + Nodal Ninja NN3
8 shots round @ -10° + 4 shots round @ +30° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  13520 x 6760  –  very good   Min: 0.05  Av: 1.17  Max: 5.23
B  Equirectangular =  13526 x 6763  –  very good   Min: 0.05  Av: 1.22  Max: 5.59
Protocol: Known distance between panoramas  =  4.774 m
Good results consistent with adding real world colour to LiDAR Scanner Point Clouds.
  II b   Nikon D5300 + Samyang 8mm + Nodal Ninja NN3
8 shots round @ -10° + 4 shots round @ +30° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  13520 x 6760  –  very good   Min: 0.05  Av: 1.17  Max: 5.23
B  Equirectangular =  13526 x 6763  –  very good   Min: 0.05  Av: 1.22  Max: 5.59
Protocol: Known length  T14 to T19 =  5.204 m
Good results consistent with adding real world colour to LiDAR Scanner Point Clouds.
  III a   Nikon D5300 + Sigma 10-20mm @ 10mm + Nodal Ninja NN3
3 rows of 8 shots round @ +45°, 0° and -30° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  16764 x 8382  –  very good   Min: 0.02  Av: 1.22  Max: 5.91
B  Equirectangular =  16762 x 8381  –  very good   Min: 0.05  Av: 1.28  Max: 5.77
Protocol: Known distance between panoramas  =  4.774 m
The results were good except for measurements including the target at T1.
  III b   Nikon D5300 + Sigma 10-20mm @ 10mm + Nodal Ninja NN3
3 rows of 8 shots round @ +45 and -30° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  16766 x 8383  –  very good   Min: 0.01  Av: 1.02  Max: 5.36
B  Equirectangular =  16762 x 8381  –  very good   Min: 0.01  Av: 1.19  Max: 5.31
Protocol: Known distance between panoramas  =  4.774 m
The upper and lower rows overlapped by 10% so it was interesting to see if it was possible to ‘get away’ with stitching just these two rows, but the results were not as good as stitching all three rows to create the equirectangular image.
   III c   Nikon D5300 + Sigma 10-20mm @ 10mm + Nodal Ninja NN3
Sigma 10mm to 20mm Rectilinear Lens @ 10mm3 rows of 8 shots round @ (Horizontal row) in Portrait orientation stitched in PTGui 12
A  Equirectangular =  16766 x 8383  –  very good   Min: 0.01  Av: 1.02  Max: 5.36
B  Equirectangular =  16762 x 8381  –  very good   Min: 0.01  Av: 1.19  Max: 5.31
Protocol: Known distance between panoramas  =  4.774 m
Stitching the single horizontal row does leave large ‘holes’ at the zenith and nadir, but if all the relevant detail is visible in the single row it could be argued that it is not necessary to stitch all three rows. It is imperative that the equirectangular proportions (width = 2 x height exactly) is maintained and this did provide a better solution than stitching only the upper and lower rows, but still not as good as stitching all three rows. Target T1 was outside the image from Panorama A so no measurements were possible to it. To capture and process all three rows does not take much longer than a single row so it is worth capturing and processing all rows when using rectilinear lenses.
  III d   Nikon D5300 + Sigma 10-20mm @ 10mm + Nodal Ninja NN3
3 rows of 8 shots round @ +45°, 0° and -30° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  16764 x 8382  –  very good   Min: 0.02  Av: 1.22  Max: 5.91
B  Equirectangular =  16762 x 8381  –  very good   Min: 0.05  Av: 1.28  Max: 5.77
Protocol: Known length  T14 to T19 =  5.204 m
The results were good except for measurements including the target at T1.
  IV a   Nikon Z5 + 7.5mm Fisheye + Nodal Ninja R!
4 shots round @ in Portrait orientation stitched in PTGui 12
A  Equirectangular =  11136 x 5568  –  very good   Min: 0.06  Av: 1.11  Max: 3.14
B  Equirectangular =  10766 x 5383  –  very good   Min: 0.09  Av: 0.79  Max: 2.08
Protocol: Known distance between panoramas  =  4.774 m
This lens is different from other Fisheye Lenses in behaviour and cost, as mentioned in IV, but is suitable for situations where the points defining the measurements are in the proximity of the baseline defined by the two panoramas, with a good ‘angle of cut’, but not along the axis of the baseline.
  IV b   Nikon Z5 + 7.5mm Fisheye + Nodal Ninja R!
4 shots round @ +7.5° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  10984 x 5492  –  very good   Min: 0.06  Av: 1.14  Max: 3.62
B  Equirectangular =  11070 x 5535  –  very good   Min: 0.08  Av: 1.21  Max: 3.80
Protocol: Known distance between panoramas  =  4.774 m
Tilting the lens up by 7.5° ensures the zenith is ‘closed’ and has produced better results in this case.
  IV c   Nikon Z5 + 7.5mm Fisheye + Nodal Ninja R!
4 shots round @ 0° in Portrait orientation stitched in PTGui 12
A  Equirectangular =  11136 x 5568  –  very good   Min: 0.06  Av: 1.11  Max: 3.14
B  Equirectangular =  10766 x 5383  –  very good   Min: 0.09  Av: 0.79  Max: 2.08
Protocol: Known length  T14 to T19 =  5.204 m
These results using this lens illustrate that it is important to give cognisance to the geometry of the set up to get the optimum outcome.

Second Set - LinesSecond Set - A

Second Set - B

Back to the Top of the Page