Entrance detection from street-level imagery
Architecture that detects entrances on building facades. In a first stage, scene geometry is exploited and the multi-dimensional problem is reduced down to a one-dimensional (1D) problem. Entrance hypotheses are generated by considering pairs of locations along lines exhibiting strong gradients in the transverse direction. In a second stage, a rich set of discriminative image features for entrances is explored according to constructed designs, specifically focusing on properties such as symmetry and color consistency, for example. Classifiers (e.g., random forest) are utilized to perform automatic feature selection and entrance classification. In another stage, a joint model is formulated in three dimensions (3D) for entrances on a given facade, which enables the exploitation of physical constraints between different entrances on the same facade in a systematic manner to prune false positives, and thereby select an optimum set of entrances on a given facade.
1. A computer-implemented method for detection of entrances in images of structural facades, the method comprising:
identifying a plurality of entrance candidates in an image of a facade of a structure, each candidate a potential entrance to the structure;
generating projections of the plurality of entrance candidates into multi-dimensional space to resolve conflicts; and
selecting one or more of the entrance candidates as true entrances based on the projections.
2. The method of claim 1 , further comprising:
generating a plurality of entrance hypotheses based on the plurality of entrance candidates; and
generating classifications of the plurality of entrance candidates to differentiate facade entrances from non-entrances based on the plurality of entrance hypotheses.
3. The method of claim 1 , further comprising:
determining a number of the plurality of entrance candidates that map to a back projected facade, the back projected facade being the facade back projected into a 3D space; and
determining for each of the back projected entrance candidates a probability that each is a true entrance.
4. The method of claim 3 , wherein the determination of the probability that each candidate is a true entrance is based on a determination of an entrance density value for the corresponding facade, the entrance density value indicating a number of entrances per unit distance for the corresponding facade.
5. The method of claim 3 , wherein the determination of the probability that each candidate is a true entrance is based on a non-overlapping constraint limiting a possible number of positions of entrances in the 3D space.
6. The method of claim 3 , wherein the determination of the probability that each candidate is a true entrance is based on a visibility of the candidate entrance from a second view in the 3D space.
7. The method of claim 3 , wherein the determination of the probability that each candidate is a true entrance is optimized based on a stochastic optimization approach.
8. A system for detection of entrances in images of structural facades, comprising:
at least one processor configured to execute computer-executable instructions in a computer readable hardware storage memory, the computer-executable instructions, when executed by the processor, cause the processor to:
identify a plurality of entrance candidates in an image of a facade of a structure, each candidate a potential entrance to the structure;
generate projections of the plurality of entrance candidates into multi-dimensional space to resolve conflicts; and
select one or more of the entrance candidates as true entrances based on the projections.
9. The system of claim 8 the computer-executable instructions comprising further instructions, that when executed by the processor, cause the processor to:
generate a plurality of entrance hypotheses based on the plurality of entrance candidates; and
generate classifications of the plurality of entrance candidates to differentiate facade entrances from non-entrances based on the plurality of entrance hypotheses.
10. The system of claim 8 , the computer-executable instructions comprising further instructions for the projection of entrance candidates into multi-dimensional space, that when executed by the processor, cause the processor to:
determine a number of the plurality of entrance candidates that map to a back projected facade, the back projected facade being the facade back projected into a 3D space; and
determine for each of the back projected entrance candidates a probability that each is a true entrance.
11. The system of claim 10 , wherein the determination of the probability that each candidate is a true entrance is based on a determination of an entrance density value for the corresponding facade, the entrance density value indicating a number of entrances per unit distance for the corresponding facade.
12. The system of claim 10 , wherein the determination of the probability that each candidate is a true entrance is based on a non-overlapping constraint limiting a possible number of positions of entrances in the 3D space.
13. The system of claim 10 , wherein the determination of the probability that each candidate is a true entrance is based on a visibility of the candidate entrance from a second view in the 3D space.
14. The system of claim 10 , wherein the determination of the probability that each candidate is a true entrance is optimized based on a stochastic optimization approach.
15. A computer-readable hardware storage medium comprising computer-executable instructions for detection of entrances from images of structural facades that when executed by a microprocessor, cause the microprocessor to:
identify a plurality of entrance candidates in an image of a facade of a structure, each candidate a potential entrance to the structure;
generate projections of the plurality of entrance candidates into multi-dimensional space to resolve conflicts; and
select one or more of the entrance candidates as true entrances based on the projections.
16. The computer-readable hardware storage medium of claim 15 , the computer-executable instructions comprising further instructions, that when executed by the processor, cause the processor to:
generate a plurality of entrance hypotheses based on the plurality of entrance candidates; and
generate classifications of the plurality of entrance candidates to differentiate facade entrances from non-entrances based on the plurality of entrance hypotheses.
17. The computer-readable hardware storage medium of claim 15 , the computer-executable instructions comprising further instructions for the projection of entrance candidates into multi-dimensional space, that when executed by the processor, cause the processor to:
determine a number of the plurality of entrance candidates that map to a back projected facade, the back projected facade being the facade back projected into a 3D space; and
determine for each of the back projected entrance candidates a probability that each is a true entrance.
18. The computer-readable hardware storage medium of claim 17 , wherein the determination of the probability that each candidate is a true entrance is based on a determination of an entrance density value for the corresponding facade, the entrance density value indicating a number of entrances per unit distance for the corresponding facade.
19. The computer-readable hardware storage medium of claim 17 , wherein the determination of the probability that each candidate is a true entrance is based on a non-overlapping constraint limiting a possible number of positions of entrances in the 3D space.
20. The computer-readable hardware storage medium of claim 17 , wherein the determination of the probability that each candidate is a true entrance is based on a visibility of the candidate entrance from a second view in the 3D space.