Many people, including some in the industry, do not know the three-match (air triangulation in the connection point matching, that is, computer vision in the SFM algorithm of the same name point matching, such as the SIFT algorithm) and dense matching (also known as stereo matching, stereo correspondence or dense matching) what is the difference, so often ask questions such as the following:
1. Short three matches when the dense point is a dense match it?
2. empty three-match after the orientation you can get the ortho image, right?
3. Can you do a dense match with sift?
As everyone knows, the answers to the above three questions are: no!
Let's show you what the difference is between an empty three-match and a dense match.
1. Definitionnull three-match: in the aerial triangulation (SFM), a point matching with the same name is used to determine the connection point between some images as the adjustment condition. dense matching: In the production of Dsm/dem, in order to calculate the area of each object square point three-dimensional coordinates, so as to reconstruct the entire area of the terrain and the same name point matching. 2. Differences
| Project |
Empty three matches |
Dense match |
| Downstream technology |
Aerial triangulation measurements |
Space Ahead Rendezvous |
| Purpose of calculation |
Image foreign position Element |
Topography of the surveying area |
| Method kernel |
Feature Matching |
Grayscale Matching |
| Distribution Features |
Sparse |
Intensive |
| Typical algorithms |
Sift |
SGM |
3. ContactThe two also have a connection, the first is to find different images of the same name feature points, so the essence is consistent, second, the empty three matching is the basis of dense matching, there is no complete empty three dense match meaningless.
The above is from the definition of the difference between the two and the relationship, in fact, can also look at the issue in a different perspective. 4. Change one angle to see the differencefirst, let's review the two basic theorems of surveying and mapping:Rear intersection: The point coordinates are determined by observing two known points from an unknown point. forward intersection: The unknown point coordinates are determined by observing the same unknown point from two known points respectively. as shown, the red dots represent unknown points, the green dots represent known points, the red lines represent the known points from unknown points, and the Green Line represents the observation of unknown points from known points.
Although the two completed the same work, that is, two known points to calculate an unknown point, but the actual application of the difference is very large, summed up as follows:rear intersection: "Active minority ", ask the way mode, that is, from a few unknown points of active observation to solve these unknown points. forward intersection:"passive majority", missionary mode, that is, most unknown points are solved by the observation of a few known points.
An example of an image is as follows:Xiao Ming was sent to Wuhan missionary, asked to bring back the personal information of every religious, so Xiao Ming intends to walk two steps:First, find a few people to ask the way to determine whether they are in Wuhan. then, everyone began to preach, and to record the information of the believers.
go back and forth to see the process of photogrammetry:Now there are a number of aerial images of a certain area, requesting the dsm/dem of the area, so our work is divided into two steps:first, the aerial triangulation is performed to determine the coordinate posture of each image (the element of the foreign position). then, a dense match is carried out on each stereo, and the three-dimensional coordinates of each object square point are calculated by the forward intersection, thus obtaining the dsm/dem of the region.
so, comparing the above example, the difference between an empty three (SFM) match and a dense match is obvious.
The difference and relation between the three-match and dense match of space
