The relationship between GSD and map scale is simple in theory but the practice of getting sharp ortho results such that they function as a photo-map is not so straightforward.
From the excellent Pix4D support pages:
The printing scale indicates the acceptable error that a printed orthomosaic can have. The acceptable error is based on the smallest difference that a human can distinguish at a drawing, which in a scale 1:1 is about 0.25 mm. This error should be multiplied by the printing scale in order to calculate the acceptable error that a printed orthomosaic should have.
Example: For a printing scale of 1:200 the acceptable error is 0.25 * 200 mm = 50 mm = 5 cm.
The acceptable error defines the GSD that should be obtained for the project. The GSD should be at least 3 times smaller than the acceptable error.
Example: For a printing scale of 1:200 and acceptable error 5 cm, the GSD should be 5 / 3 cm = 1.7 cm. There is no problem if the GSD is smaller (1.5 cm, 1 cm, 0.5 cm).
GSD= Ground Sample Distance. In remote sensing, GSD in a digital photo (such as an orthophoto) of the ground from air or space is the distance between pixel centers measured on the ground. For example, in an image with a one-meter GSD, adjacent pixels image locations are 1 meter apart on the ground.
GSD=[Sensor width (mm) x h (m)]/ [(focal length (mm) x Sensor width (px)]
where, for the L1D-20c:
- Sensor width mm : 13.2
- f (focal length)=10.26 mm
- Sensor width px : 5472
[13.2 x 30]/ [10.26 x 5472]
or GSD=7mm/px for a single image captured at 30m.
At 25-30m ASL the GSD for the Mavic Pro2 is around the 8mm mark, more than good enough for 1:200 scale mapping but for this value to be useful imagery needs to be in sharp focus across the entire frame.
The ‘unsharp’ performance of the camera has driven some users to despair: A surveyor responded to the suggestion the MP2 is a good phtogrammetric platform thus:
“I have used a Mavic 2 Pro for photogrammetry… The image colour quality and sharpness is better than the original Mavic but the lens produces soft, blurry images that make models average at best. The blurred photo edges are a problem…I am so disappointed with the photo quality… “
Which makes me think this surveyor needs to start thinking like a photographer! The little Mavic is no U2 spy plane so getting reasonable image resolution takes patience.
The L1D-20c camera has curious performance and is downright unreliable at the wide stops. Its saving grace is the large sensor which allows image scaling from reasonably efficient heights.
Good results come from
- using a long exposure with a low ISO
- a static aircraft to reduce rolling shutter effects (T mode will help smooth the braking)
- f8 or better to increase sharp focus by increasing depth of field
- flying in strong light conditions
- use still shots not video frame splitting
- use RAW format
- keep it low: 25-30m ASL is fine for 1:200 scale mapping
- deep overlap: aim for 60% for both edge and side of swath
- get the best ground view possible, controlling speed and overlap depend on this.
L1D-20c images tend to display poor focus at the frame edges at stops wider than f8 and the tighter aperture setting mitigates this.
As a survey camera I prefer the Ricoh GR but as an all-in-one flying camera the Mavic Pro2 is pretty good for the money. My brief experience with the P4 is enough to convince me of the superior stability of the Mavic platform, particularly in windy conditions.
A 100% screen grab gives an idea of the 8mmGSD from 30m ASL capture:
The ortho-image (top) was generated from 118 frames taken in 3 swaths at 30m and 2 at 60m ASL. The raised swath being to comply with proximity clearance from the public road.
The camera was set up with aperture fixed at f8 with shutter speeds between 240th and 320th, ISO 200, ev-0.3.
Overlap was 50-60% by eye between positions synced at 10s apart with the aircraft static, this takes practice to use the countdown on the video relay as the the next shot trigger.