1
Camera Calibration / Most efficient way to take photos: GSD and number of Photos
« on: June 29, 2021, 04:33:48 PM »
I have a question abut the most efficient way to take photos get the best quality model (texture).
If I have a 20MP camera (4: 3 -> 5164px x 3873px), what is the distance at which to photograph a cuboid with a side of 2.5 meters for example?
In order to have a model with a GSD (Ground Sample Distance) comparable to the resolution of the human eye 0.2 mm / px (150DPI) I estimated a distance of 50 cm. In this condition, the FOV allows you to photograph a rectangle of approximately 1 meter x 70 centimeters. Taking into account an overlap of the photos of 50% as recommended by the manual, it is necessary to move horizontally by 50 centimeters and vertically by 35 centimeters. Consequently, to photograph a cube with a side of 2.5 meters, 40 photos per face are required and therefore 240 shots. Then, considering the 12 edges (8 horizontal and 4 vertical), another 8x5 + 4x8 = 72 photos are required. For a total of 312 photos.
What I am wondering is if this account is wrong or if Metashape is capable of generally a high resolution texture result even with a smaller number of shots (staying at a greater distance for example). Also I do not understand if in this reasoning that I make a mistake in the way of photographing. In the advice in the manual I understand that it is always convenient to take a portion of the floor and / or ceiling. By doing the indicated procedure, however, do not take the sky or the floor in certain shots.
Finally I would like to ask another question. Taking for example the usual 2.5-meter cuboid, if I didn't have the ability to photograph the object from above with a drone but using a selfie stick, I can still get reliable results by photographing at two different heights and 2 different angles (45°-30°) without photographing parallel to the face up?
Thanks
If I have a 20MP camera (4: 3 -> 5164px x 3873px), what is the distance at which to photograph a cuboid with a side of 2.5 meters for example?
In order to have a model with a GSD (Ground Sample Distance) comparable to the resolution of the human eye 0.2 mm / px (150DPI) I estimated a distance of 50 cm. In this condition, the FOV allows you to photograph a rectangle of approximately 1 meter x 70 centimeters. Taking into account an overlap of the photos of 50% as recommended by the manual, it is necessary to move horizontally by 50 centimeters and vertically by 35 centimeters. Consequently, to photograph a cube with a side of 2.5 meters, 40 photos per face are required and therefore 240 shots. Then, considering the 12 edges (8 horizontal and 4 vertical), another 8x5 + 4x8 = 72 photos are required. For a total of 312 photos.
What I am wondering is if this account is wrong or if Metashape is capable of generally a high resolution texture result even with a smaller number of shots (staying at a greater distance for example). Also I do not understand if in this reasoning that I make a mistake in the way of photographing. In the advice in the manual I understand that it is always convenient to take a portion of the floor and / or ceiling. By doing the indicated procedure, however, do not take the sky or the floor in certain shots.
Finally I would like to ask another question. Taking for example the usual 2.5-meter cuboid, if I didn't have the ability to photograph the object from above with a drone but using a selfie stick, I can still get reliable results by photographing at two different heights and 2 different angles (45°-30°) without photographing parallel to the face up?
Thanks