Abstract: Techniques for computing a volume of interest for photogrammetric 3D reconstruction are provided. According to one set of embodiments, a computer system can receive a plurality of observations of a scene captured by a camera operator via one or more cameras and, for each observation in the plurality of observations, determine a likelihood distribution function that outputs, for each point in 3D real coordinate space, a likelihood that the point was of interest to the camera operator in view of the observation. The computer system can further combine the likelihood distribution functions for the plurality of observations into an aggregate likelihood distribution function and can compute a volume of interest using the aggregate likelihood distribution function. The computer system can then perform a photogrammetric 3D reconstruction of the scene, where the photogrammetric 3D reconstruction is constrained to objects within the volume of interest.

Abstract: Techniques for computing a volume of interest for photogrammetric 3D reconstruction are provided. According to one set of embodiments, a computer system can receive a plurality of observations of a scene captured by a camera operator via one or more cameras and, for each observation in the plurality of observations, determine a likelihood distribution function that outputs, for each point in 3D real coordinate space, a likelihood that the point was of interest to the camera operator in view of the observation. The computer system can further combine the likelihood distribution functions for the plurality of observations into an aggregate likelihood distribution function and can compute a volume of interest using the aggregate likelihood distribution function. The computer system can then perform a photogrammetric 3D reconstruction of the scene, where the photogrammetric 3D reconstruction is constrained to objects within the volume of interest.