Note: I am not asking for code just the name of some algorithms I can research
Problem
I'm trying to do a 3D reconstruction from multiple images where the subject is rotating.
I don't have a lot of experience in the CV field and my mathematics is not particularly good so I am reliant on either simply worded papers or existing libraries/tools.
Where I'm At
I have been researching and testing using Structure from Motion (with VisualSFM) to generate a 3d point cloud and Multi View Stereo (with CMPMVS) to reconstruct the scene.
Obviously these algorithms are designed to process images where the camera (or cameras) moves and the scene is stationary so the reconstructions fail for a number of reasons.
Example: I've been working with short videos of a person rotating around a marker on the floor, see below. I've tried removing the background but this seems to make it harder for SfM to build pairs, probably because of the lack of intformation.
Question
Does anyone know the name of an algorithm/pipeline that will be able to reconstruct a series of images with a static background and a rotating object?
Or is it not possible to do a 3d reconstruction in this way?
Related
I need to make an app that detects images and their position, and displays AR content on them. These images will change during the lifetime of the app, and there can be many of them. I'm wondering how to design this kind of app. ARKit can provide this functionality - detect image and it's orientation, and display AR content on it. But the problem is that ARKit can detect only a limited number of images at a time. If I have for example 300 images, then there can be problem. Maybe I could prepare some ML dataset to pre-detect image, and then assign it as an ARKit trackable on the fly? Is this the right approach? What else could I do to make such an app with dynamic and large set of images to detect?
Regarding a ML approach, you can use just about any state-of-the-art object detection network to pull the approximate coordinates of your desired target and extract that section of the frame, passing positives to ARKit or similar. The downside is that training will probably be resource-intensive. It could work, but I can't speak to its efficiency relative to other approaches.
In looking to extend this explanation, I see the ARKit 2.0 handles (what seems to be) what you're trying to do; is this insufficient?
To answer your question in the comments, CoreML seems to offer models for object recognition but not localization, so I suspect it'd be necessary to use their converter after training a model such as these. The input to this network would be frames from camera, and output would be detected classes with probabilities of detection, and approximate coordinates; if your targets are present, and roughly where they are.
Again, though, if you're looking for 2D images rather than 3D+ objects, and especially if it's an ARKit app anyway, it really looks like ARKit's built-in tracking will be much more effective at substantially lower development cost.
At WWDC '19 ARKit 3 was touted to support up to 100 images for image detection. Image tracking supports a lower number of images, which I believe is still under 10. You have to recognize images yourself if you want more than that, currently.
As an idea, you can identify rectangles in the camera feed and then apply a CIPerspectiveCorrection filter to extract a fully 2D image based on the detected rectangle. See Tracking and Altering Images sample code which does something similar.
You then compare the rectangle's image data against your set of 300 source images. ARKit stopped at 100 likely due to performance concerns, but it's possible you can surmount those numbers with a performance metric that's acceptable to your own criteria.
I have a next task: get a room 3d projection from multiple images (possible video stream, doesn't matter). There will be spherical camera (in fact multiple cameras on sphere-like construction), so the case is the right one on the image.
I decided to code it on iOS platform as I'm iOS developer and model cameras with iPhone cam rotating it as shown on the pic above. As I can decompose this task, first I need to get real distance to the objects (walls in most cases, I think). Is it possible? Which algoritms/methods should I use to achieve this? I don't ask you to make the task for me obviously, but give me the direction, because I have no idea, maybe some equations/tutorials/algorithms with explanation to my case. Thank you!
The task of building a 3D model from multiple 2D images is called "scene reconstruction." It's still an active area of research, but solutions involve recognizing the same keypoint (e.g. a distinctive part of an object) in two images. Once you have that, you can use the known camera geometry to solve for the 3D position of that keypoint in the world.
Here's a reference:
http://docs.opencv.org/3.1.0/d4/d18/tutorial_sfm_scene_reconstruction.html#gsc.tab=0
You can google "scene reconstruction" to find lots more, and papers that go into more detail.
This is the setup: A fairly large room with 4 fish-eye cameras mounted on the ceiling. There are no blind spots. Each camera coverage overlaps a little with the other.
The idea is to track people across these cameras. As of now a blob extracting algorithm is in place, which detects people as blobs. It's a fairly decently working algorithm which detects individual people pretty good. Am using the OpenCV API for all of this.
What I mean by track people is that - Say, camera 1 identifies two people, say Person A and Person B. Now, as these two people move from the coverage of camera 1 into the overlapping area of coverage of cam1 and cam2 and into the area where only cam2 covers, cam2 should be able to identify them as the same people A and B cam1 identified them as.
This is what I thought I'd do -
1) The camera renders the image at 15fps and I think the dimensions of the frames are of 1920x1920.
2) Identify blobs individually in each camera and give each blob an unique label.
3) Now as for the overlaps - Compute an affine transformation matrix which maps pixels on one camera's frame onto another camera's frame - this needn't be done for every frame - this can be done before the whole process starts, as a pre-processing step. So in real time, whenever I detect a blob which is in the overlapping area, all I have to do is apply the transformation matrix to the pixels in cam1 and see if there is a corresponding blob in cam2 and give them the same label.
So, Questions :
1) Would this system give me a badly-working system which tracks people decently ?
2) So, for the affine transform, do I have to convert the fish-eye to rectilinear image ? (My answer is yes, but am not too sure)
Please feel free to point out possible errors and why certain things might not work in the process I've described. Also alternate suggestions are welcome! TIA
1- blob extraction is not enough to track a specific object, for people case I suggest HoG - or at least background subtraction before blob extraction, since all of the cameras have still scenes.
2- opencv <=2.4.9 uses pinhole model for stereo vision. so, before any calibration with opencv methods your fisheye images must be converted to rectilinear images first. You might try calibrating yourself using other approaches too
release 3.0.0 will have support for fisheye model. It is on alpha stage, you can still download and give it a try.
I want to do something like this but in reverse-- so that the cameras are outside and pointing inward. Let's start with the abstract and get specific:
1) Are there any TOOLS that will do this for me? How close can I get using existing software?
2) Say the nearest tool is a graphics library like OpenCV. I've taken linear algebra and have an undergraduate degree in CS but without any special training in graphics. Where should I go from there?
3) If I really am undergoing a decade-long spiritual quest of a self-teaching+programming exercise to make this happen, are there any papers or other resources that you aware of that might aid me?
I think the demo you linked uses a 360° camera (see the black circle on the bottom) and does not involve stitching in any way.
About your question, are you aware of this work? They don't do stitching either, just blending between different views.
If you use inward views, then the objects you will observe will probably be quite close to the cameras, while standard stitching assumes that objects are far away. Close 3D objects mean high distortion when you change the viewpoint (i.e. parallax & occlusions), which makes it difficult to interpolate between two views. Hence, if you want stitching, then your main problem is to correctly handle parallax effects & occlusions between the views.
In my opinion, the most promising approach would be to do live stereo matching (i.e. dense 3D reconstruction) between the two camera images closest to your current viewpoint, and then interpolate the estimated disparities to generate an expected image. However, it's not likely to run in real-time, as demonstrated in the demo you linked, and the result could be quite ugly...
EDIT
You can also have a look at this paper, which uses a different but interesting approach, however maybe not directly useful in your case since it requires the new viewpoint to be visible in the available images.
If I take a picture with a camera, so I know the distance from the camera to the object, such as a scale model of a house, I would like to turn this into a 3D model that I can maneuver around so I can comment on different parts of the house.
If I sit down and think about taking more than one picture, labeling direction, and distance, I should be able to figure out how to do this, but, I thought I would ask if someone has some paper that may help explain more.
What language you explain in doesn't matter, as I am looking for the best approach.
Right now I am considering showing the house, then the user can put in some assistance for height, such as distance from the camera to the top of that part of the model, and given enough of this it would be possible to start calculating heights for the rest, especially if there is a top-down image, then pictures from angles on the four sides, to calculate relative heights.
Then I expect that parts will also need to differ in color to help separate out the various parts of the model.
As mentioned, the problem is very hard and is often also referred to as multi-view object reconstruction. It is usually approached by solving the stereo-view reconstruction problem for each pair of consecutive images.
Performing stereo reconstruction requires that pairs of images are taken that have a good amount of visible overlap of physical points. You need to find corresponding points such that you can then use triangulation to find the 3D co-ordinates of the points.
Epipolar geometry
Stereo reconstruction is usually done by first calibrating your camera setup so you can rectify your images using the theory of epipolar geometry. This simplifies finding corresponding points as well as the final triangulation calculations.
If you have:
the intrinsic camera parameters (requiring camera calibration),
the camera's position and rotation (it's extrinsic parameters), and
8 or more physical points with matching known positions in two photos (when using the eight-point algorithm)
you can calculate the fundamental and essential matrices using only matrix theory and use these to rectify your images. This requires some theory about co-ordinate projections with homogeneous co-ordinates and also knowledge of the pinhole camera model and camera matrix.
If you want a method that doesn't need the camera parameters and works for unknown camera set-ups you should probably look into methods for uncalibrated stereo reconstruction.
Correspondence problem
Finding corresponding points is the tricky part that requires you to look for points of the same brightness or colour, or to use texture patterns or some other features to identify the same points in pairs of images. Techniques for this either work locally by looking for a best match in a small region around each point, or globally by considering the image as a whole.
If you already have the fundamental matrix, it will allow you to rectify the images such that corresponding points in two images will be constrained to a line (in theory). This helps you to use faster local techniques.
There is currently still no ideal technique to solve the correspondence problem, but possible approaches could fall in these categories:
Manual selection: have a person hand-select matching points.
Custom markers: place markers or use specific patterns/colours that you can easily identify.
Sum of squared differences: take a region around a point and find the closest whole matching region in the other image.
Graph cuts: a global optimisation technique based on optimisation using graph theory.
For specific implementations you can use Google Scholar to search through the current literature. Here is one highly cited paper comparing various techniques:
A Taxonomy and Evaluation of Dense Two-Frame Stereo Correspondence Algorithms.
Multi-view reconstruction
Once you have the corresponding points, you can then use epipolar geometry theory for the triangulation calculations to find the 3D co-ordinates of the points.
This whole stereo reconstruction would then be repeated for each pair of consecutive images (implying that you need an order to the images or at least knowledge of which images have many overlapping points). For each pair you would calculate a different fundamental matrix.
Of course, due to noise or inaccuracies at each of these steps you might want to consider how to solve the problem in a more global manner. For instance, if you have a series of images that are taken around an object and form a loop, this provides extra constraints that can be used to improve the accuracy of earlier steps using something like bundle adjustment.
As you can see, both stereo and multi-view reconstruction are far from solved problems and are still actively researched. The less you want to do in an automated manner the more well-defined the problem becomes, but even in these cases quite a bit of theory is required to get started.
Alternatives
If it's within the constraints of what you want to do, I would recommend considering dedicated hardware sensors (such as the XBox's Kinect) instead of only using normal cameras. These sensors use structured light, time-of-flight or some other range imaging technique to generate a depth image which they can also combine with colour data from their own cameras. They practically solve the single-view reconstruction problem for you and often include libraries and tools for stitching/combining multiple views.
Epipolar geometry references
My knowledge is actually quite thin on most of the theory, so the best I can do is to further provide you with some references that are hopefully useful (in order of relevance):
I found a PDF chapter on Multiple View Geometry that contains most of the critical theory. In fact the textbook Multiple View Geometry in Computer Vision should also be quite useful (sample chapters available here).
Here's a page describing a project on uncalibrated stereo reconstruction that seems to include some source code that could be useful. They find matching points in an automated manner using one of many feature detection techniques. If you want this part of the process to be automated as well, then SIFT feature detection is commonly considered to be an excellent non-real-time technique (since it's quite slow).
A paper about Scene Reconstruction from Multiple Uncalibrated Views.
A slideshow on Methods for 3D Reconstruction from Multiple Images (it has some more references below it's slides towards the end).
A paper comparing different multi-view stereo reconstruction algorithms can be found here. It limits itself to algorithms that "reconstruct dense object models from calibrated views".
Here's a paper that goes into lots of detail for the case that you have stereo cameras that take multiple images: Towards robust metric reconstruction
via a dynamic uncalibrated stereo head. They then find methods to self-calibrate the cameras.
I'm not sure how helpful all of this is, but hopefully it includes enough useful terminology and references to find further resources.
Research has made significant progress and these days it is possible to obtain pretty good-looking 3D shapes from 2D images. For instance, in our recent research work titled "Synthesizing 3D Shapes via Modeling Multi-View Depth Maps and Silhouettes With Deep Generative Networks" took a big step in solving the problem of obtaining 3D shapes from 2D images. In our work, we show that you can not only go from 2D to 3D directly and get a good, approximate 3D reconstruction but you can also learn a distribution of 3D shapes in an efficient manner and generate/synthesize 3D shapes. Below is an image of our work showing that we are able to do 3D reconstruction even from a single silhouette or depth map (on the left). The ground-truth 3D shapes are shown on the right.
The approach we took has some contributions related to cognitive science or the way the brain works: the model we built shares parameters for all shape categories instead of being specific to only one category. Also, it obtains consistent representations and takes the uncertainty of the input view into account when producing a 3D shape as output. Therefore, it is able to naturally give meaningful results even for very ambiguous inputs. If you look at the citation to our paper you can see even more progress just in terms of going from 2D images to 3D shapes.
This problem is known as Photogrammetry.
Google will supply you with endless references, just be aware that if you want to roll your own, it's a very hard problem.
Check out The Deadalus Project, althought that website does not contain a gallery with illustrative information about the solution, it post several papers and info about the working method.
I watched a lecture from one of the main researchers of the project (Roger Hubbold), and the image results are quite amazing! Althought is a complex and long problem. It has a lot of tricky details to take into account to get an approximation of the 3d data, take for example the 3d information from wall surfaces, for which the heuristic to work is as follows: Take a photo with normal illumination of the scene, and then retake the picture in same position with full flash active, then substract both images and divide the result by a pre-taken flash calibration image, apply a box filter to this new result and then post-process to estimate depth values, the whole process is explained in detail in this paper (which is also posted/referenced in the project website)
Google Sketchup (free) has a photo matching tool that allows you to take a photograph and match its perspective for easy modeling.
EDIT: It appears that you're interested in developing your own solution. I thought you were trying to obtain a 3D model of an image in a single instance. If this answer isn't helpful, I apologize.
Hope this helps if you are trying to construct 3d volume from 2d stack of images !! You can use open source tool such as ImageJ Fiji which comes with 3d viewer plugin..
https://quppler.com/creating-a-classifier-using-image-j-fiji-for-3d-volume-data-preparation-from-stack-of-images/