Photogrammetry, so what is it? What if I told you that its origins date back to the grand old master himself, Leonardo Da Vinci and also Albrecht Duerer, in the 1400 and 1500’s. Perhaps not in the same sense as we use the term today, but they were the first to use mathematical science to determine perspective within their drawings. From this developed the use of mathematical principles to determine the co ordinates of a point in 3d space from two or more images. Imagine that but using modern digital cameras and the power of computers to match every pixel from two different photographs (pairs) and then using hundreds of photos over hectares of terrain.
So how does it work, initially I just thought it was voodoo magic, but apparently not, and when I researched it, its not so difficult to understand. Taking a photograph is the process of capturing the 3d world and applying it to a 2d canvas. Photogrammetry , in its simplest form, reverses this process and gives us the 3rd dimension back. Take us, humans, actually take most living things on this planet that have eyes, but just the ones with two for now!, If we only possessed one, depth perception would be neigh on impossible, but two, side by side, gives us the ability to judge depth. This is not so dissimilar to the way a camera, with one lens can be used while mounted on an aerial system. While moving forward it take a series of photographs, the second image overlapping the first, the third overlapping the second, etc, using pairs of these images, two eyes, and knowing the focal distance and sensor size, imaginary (or computed) rays can be formed back to the camera to produce a line of sight, with orientation and position, to a specific point, in this case a line for every pixel, triangulation is then used on both images, where the rays for matching pixels intersect, that point gives us depth.
Now, the act of converting the 3d world into a 2d photograph and then converting it back to 3 dimensions is not full proof, the very act of this process is going to cause some data loss, even the very best laser scanners will only capture what is in view and at a resolution that is determined by the instruments used in either process. All that said and done, compared with off the shelf terrain data, ie 5m Ordnance Survey terrain data for example, photogrammetry can produce cm level resolution and accuracies and precisions that are also in the cm range, and the beauty of it is it is very scalable. Photos taken 5m above the ground will produce sub mm data, photos taken at 100m above the ground will give you sub decimetre data.
So, what next? The whole process usually follows a similar procedure, regardless of software. Initially processing software will reduces the quality of the images by a factor of 2 – 4. Then, from this, an initial low density point cloud is produced, this also aligns the images. then a dense point cloud is produced from the full resolution images. Both point clouds are assigned a colour value from the pixels they were derived from. Meshing the points is the next step, if desired, followed by a full ortho-mosaic, all photos aligned to form a vertical panorama, any colour correction can be done to create a seamless image of the area photographed. So, meshed data in any industry standard format, we personally prefer .DXF or .OBJ, point clouds in XYZ and ortho-mosaics as tiffs.
Surveying? Generally, when people talk about a topographic survey, they usually expect a CAD file of some description that includes point heights, lines that indicate man made structures and quite often contour data. In an urban area, the data will be mainly made up off man made features that interact with the terrain they are on. All this data will very often have a co-ordinate system associated with it, which places it within a real world mapping context. They can also be part of a clients local grid. All our photogrammetric data is geo reference using our ground survey of the same area. We use a networked RTK positioning system that gives us sub cm accuracy of targets that are captured within our aerial images. This places the finished model into the Ordnance Survey Grid System. Our final data output gives an accurate model with everything included. We can then post process this to provide a multitude of data sets specific to a clients needs. All done in house.
Please view the example below.
This post has been seen 4353 times.