You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
At 8,192x4,096, the map used for determining On Water, has 33,554,432 pixels. So, it would be accurate between 1 and 2 decimal places for a given lat/long:
At 1 decimal place, there are 3,600 longitudes and 1,800 latitudes (3,600x1,800) - so, 6,480,000 points.
At 2 decimal places, there are 36,000 longitudes and 18,000 latitudes (36,000x18,000) - so, 648,000,000 points.
At 1 decimal place, a given latitude is accurate to within 11.112km.
At 2 decimal places, a given latitude is accurate to within 1.1112km. Source
So, at 33.5 million points the accuracy of the source map is somewhere between 1 and 2 decimal places (between 1.1 and 11.1 km).
I believe the accuracy roughly works out to:
$$ {11.1112 km \over \sqrt{33,554,432 \over 6,480,000}} = 4.88 km $$
Of course, this means we would need quite a large number of images to try increasing the accuracy, as we would improve the accuracy by
$$ {\sqrt{number-of-images}} $$
(eg 4 images will double accuracy, 16 images will quadruple accuracy, etc)
So, if we want to shoot for 3 decimal places of accuracy (111 meters), we would need
$$ {8,192 \times 4,096 \times x = 360,000 \times 180,000} $$
Solving for x, the answer is 1931.19. The closest power of 2 would be 2048. So, if it were possible to get 2048 8k tiles of the earth, this program could more accurately capture rivers (that are at least 100-111 meters across) and coastlines. I wouldn't suggest going much more than 2048 images, as already that would take up 4GB of space, and beyond that, the space required for getting higher accuracy would work it's way up exponentially, and become impractical for normal use-cases.
I think, programmatically, this shouldn't be too difficult to solve - the only adaptation would be adjusting the CoordsToPixels to account for the larger number of total pixels, and then use a divisor and modulus to determine which image and which part of which image to use.
What I don't know how to do is source the images. I don't know where you got your current 8k image, or where I might look to find bigger images/b&w map image tiles. Do you have any ideas?
The text was updated successfully, but these errors were encountered:
At 8,192x4,096, the map used for determining On Water, has 33,554,432 pixels. So, it would be accurate between 1 and 2 decimal places for a given lat/long:
At 1 decimal place, there are 3,600 longitudes and 1,800 latitudes (3,600x1,800) - so, 6,480,000 points.
At 2 decimal places, there are 36,000 longitudes and 18,000 latitudes (36,000x18,000) - so, 648,000,000 points.
At 1 decimal place, a given latitude is accurate to within 11.112km.
At 2 decimal places, a given latitude is accurate to within 1.1112km.
Source
So, at 33.5 million points the accuracy of the source map is somewhere between 1 and 2 decimal places (between 1.1 and 11.1 km).
I believe the accuracy roughly works out to:
$$ {11.1112 km \over \sqrt{33,554,432 \over 6,480,000}} = 4.88 km $$
Of course, this means we would need quite a large number of images to try increasing the accuracy, as we would improve the accuracy by
$$ {\sqrt{number-of-images}} $$
(eg 4 images will double accuracy, 16 images will quadruple accuracy, etc)
So, if we want to shoot for 3 decimal places of accuracy (111 meters), we would need
$$ {8,192 \times 4,096 \times x = 360,000 \times 180,000} $$
Solving for x, the answer is 1931.19. The closest power of 2 would be 2048. So, if it were possible to get 2048 8k tiles of the earth, this program could more accurately capture rivers (that are at least 100-111 meters across) and coastlines. I wouldn't suggest going much more than 2048 images, as already that would take up 4GB of space, and beyond that, the space required for getting higher accuracy would work it's way up exponentially, and become impractical for normal use-cases.
I think, programmatically, this shouldn't be too difficult to solve - the only adaptation would be adjusting the
CoordsToPixels
to account for the larger number of total pixels, and then use a divisor and modulus to determine which image and which part of which image to use.What I don't know how to do is source the images. I don't know where you got your current 8k image, or where I might look to find bigger images/b&w map image tiles. Do you have any ideas?
The text was updated successfully, but these errors were encountered: