Abstract
Since people spend most of their time indoors, their indoor activities and related issues in health, security and energy consumption have to be understood. Hence, gathering and representing spatial information of indoor spaces in form of 3D models become very important. Considering the available data gathering techniques with respect to the sensors cost and data processing time, single images proved to be one of the reliable sources. Many of the current single image based indoor space modeling methods are defining the scene as a single box primitive. This domain-specific knowledge is usually not applicable in various cases where multiple corridors are joined at one scene. Here, we addressed this issue by hypothesizing-verifying multiple box primitives which represents the indoor corridor layout. Middle-level perceptual organization is the foundation of the proposed method, which relies on finding corridor layout boundaries using both detected line segments and virtual rays created by orthogonal vanishing points. Due to the presence of objects, shadows and occlusions, a comprehensive interpretation of the edge relations is often concealed. This necessitates the utilization of virtual rays to create a physically valid layout hypothesis. Many of the former methods used Orientation Map or Geometric Context to evaluate their proposed layout hypotheses. Orientation map is a map that reveals the local belief of region orientations computed from line segments, and in a segmented image geometric context uses color, texture, edge, and vanishing point cues to estimate the likelihood of each possible label for all super-pixels. Here, the created layout hypotheses are evaluated by an objective function which considers the fusion of orientation map and geometric context with respect to the horizontal viewing angle at each image pixel. Finally, the best indoor corridor layout hypothesis which gets the highest score from the scoring function will be selected and converted to a 3D model. It should be noted that this method is fully automatic and no human intervention is needed to obtain an approximate 3D reconstruction.
Highlights
How much time are you spending at your apartment,/house, office, or other indoor places every day? According to the U.S Environmental Protection Agency (EPA), approximately 318,943,000 people in United States, spend around 90% of their time at indoor places (U.S EPA, 2015)
The main focus of this paper is on 3D modeling of indoor corridors using a single image. 3D modeling of indoor spaces is not a trivial task, and it involves with major problems
The proposed indoor corridor layout estimation approach is following the Manhattan rule assumption to simplify the structure of the indoor corridor layouts
Summary
According to the U.S Environmental Protection Agency (EPA), approximately 318,943,000 people in United States, spend around 90% of their time at indoor places (U.S EPA, 2015). This record shows how crucial the spatial information of the indoor places could be. Given only a monocular image of a corridor scene, we can provide a 3D model allowing the potential viewer to virtually explore the corridor without having to physically visit the scene This adds another dimension to static GIS at indoor places, and is convenient for buildings where direct search in those places is time consuming
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have