What is the ‘overlapping’? And why is it so critical in 3D reconstruction?
Quite often, we see the requirement of overlapping images and views in data capturing, here is a short introduction on what it is and why it is so important!
Maintaining high overlap between two images is crucially important for building a 3D model. But why? Let’s find out the reasons!
When users upload images of a certain object to Altizure’s cloud computing system, it will automatically detect thousands of keypoints on each image. If two keypoints on two different images are detected to be the same, they become correctly-matched keypoints. The computer will then extract the features of the object from a set of matched keypoints and generate a 3D model.
The matched keypoints can only be detected on common areas (or overlapping areas) between different images. The larger the common area is, the more keypoints there are. The more matched keypoints are found, the more accurate the 3D model can be.
But what does it mean by “overlap”? Overlap contains both the front-overlap (frontlap) and side-overlap (sidelap).
Sidelap refers to the percentage of overlap between different flight legs. Just as the below picture shows, the same size of image (which is represented by the rectangle) is captured on each spot. Suppose the blue spot indicates Position A, while the green spot indicates B. The shaded area means the overlapping area between Image A and B. So, if the shaded area is 80% of the rectangle, the sidelap will be 80%.
Frontlap means the percentage of overlap between one image and the next, and both images are taken by the drones when flying in the same direction. If the shaded area in the picture is 80% of the rectangle, it means the frontlap will be 80%.
However, the difference between “frontlap” and “sidelap” may not be so clear-cut in situations other than Zigzag flight. We don’t need to specify “overlap” into precise details and divide them into “sidelap” and “frontlap”, but just take it as a whole, which is the overlapping area between two images.
The following pictures show another kind of “overlap”. Both pictures are taken from right above the house. The left one was taken at a lower altitude, and it is a close-up shot at the roof of the house. As for the picture on the right, it is a shot at a much higher altitude, i.e. the vertical view of the entire area. There are some overlapping areas between both pictures. So Altizure’s computing system detects several keypoints in the two different pictures and pair them up. Those yellow lines show which two keypoints are matched together. In another word, if the pictures have enough overlaps, the system can find out keypoints, match them with one another, and generate 3D points to form a 3D model.
Therefore, in a word, one trick to improve the accuracy of your 3D reconstruction model is to increase overlap of your photos. But how to achieve satisfactory overlap in shootings? This is exactly what we are going to show you in the future. The tutorial is coming soon. Please stay stuned with us.
For more information, please visit altizure.com
