Search

CMP SfM Web Service is a remote procedure call service operated at the Center of Machine Perception (CMP) of the Czech Technical University in Prague. The majority of available procedures are implementations of Computer Vision methods developed at CMP. The service can be accessed through web page and command line scripting interfaces. This paper presents the service through users' perspective, providing a brief description of the CMP Structure-from-Motion methods and the respective job types. We also explain the technology behind the web service and the most common use cases of the browser interface.

3D building models, being the main part of a digital city scene, are essential to all applications related to human activities in urban environments. The development of range sensors and Multi-View Stereo (MVS) technology facilitates our ability to automatically reconstruct level of details 2 (LoD2) models of buildings. However, because of the high complexity of building structures, no fully automatic system is currently available for producing building models. In order to simplify the problem, a lot of research focuses only on particular buildings shapes, and relatively simple ones. In this paper, we analyze the property of topology graphs of object surfaces, and find that roof topology graphs have three basic elements: loose nodes, loose edges, and minimum cycles. These elements have interesting physical meanings: a loose node is a building with one roof face; a loose edge is a ridge line between two roof faces whose end points are not defined by a third roof face; and a minimum cycle represents a roof corner of a building. Building primitives, which introduce building shape knowledge, are defined according to these three basic elements. Then all buildings can be represented by combining such building primitives. The building parts are searched according to the predefined building primitives, reconstructed independently, and grouped into a complete building model in a CSG-style. The shape knowledge is inferred via the building primitives and used as constraints to improve the building models, in which all roof parameters are simultaneously adjusted. Experiments show the flexibility of building primitives in both lidar point cloud and stereo point cloud.

We present a novel method for 3D surface reconstruction from an input cloud of 3D points augmented with visibility information. We observe that it is possible to reconstruct surfaces that do not contain input points. Instead of modeling the surface from input points, we model free space from visibility information of the input points. The complement of the modeled free space is considered full space. The surface occurs at interface between the free and the full space. We show that under certain conditions a part of the full space surrounded by the free space must contain a real object also when the real object does not contain any input points; that is, an occluder reveals itself through occlusion. Our key contribution is the proposal of a new interface classifier that can also detect the occluder interface just from the visibility of input points. We use the interface classifier to modify the state-of-the-art surface reconstruction method so that it gains the ability to reconstruct weakly supported surfaces. We evaluate proposed method on datasets augmented with different levels of noise, undersampling, and amount of outliers. We show that the proposed method outperforms other methods in accuracy and ability to reconstruct weakly supported surfaces.

We present a multi-view stereo method that avoids producing hallucinated surfaces which do not correspond to real surfaces. Our approach to 3D reconstruction is based on the minimal s-t cut of the graph derived from the Delaunay tetrahedralization of a dense 3D point cloud, which produces water-tight meshes. This is often a desirable property but it hallucinates surfaces in complicated scenes with multiple objects and free open space. For example, a sequence of images obtained from a moving vehicle often produces meshes where the sky is hallucinated because there are no images looking from the above to the ground plane. We present a method for detecting and removing such surfaces. The method is based on removing perturbation sensitive parts of the reconstruction using multiple reconstructions of perturbed input data. We demonstrate our method on several standard datasets often used to benchmark multi-view stereo and show that it outperforms the state-of-the-art techniques .