Moving object detection apparatus, method and program

Claim 2

A moving object detection apparatus, comprising: motion vector calculation means for calculating motion vectors from an input image; motion vector removal means for removing a motion vector having high randomness from said calculated motion vectors; motion vector accumulation means for temporally accumulating each said motion vector not removed by said motion vector removal means, and for calculating an accumulated number of occurrences and an accumulated value of each motion vector; and moving object detection means for determining, based on said calculated accumulated value and calculated accumulated number of occurrences of each motion vector, whether each said motion vector corresponds to a moving object, wherein said motion vector accumulation means includes: accumulated value storage means for storing said accumulated values and numbers of said previous motion vectors such that said accumulated values and numbers are to be associated with the preceding positions of said motion vectors, accumulated value readout means for reading out said accumulated values and numbers of said previous motion vectors from said accumulated value storage means, and accumulated value update means for calculating new said accumulated values and numbers of said motion vectors by performing an update based on said accumulated values and numbers of said previous motion vectors and a magnitude and position of said current motion vector, said accumulated values and numbers of said previous motion vectors being stored in said accumulated value storage means, said accumulated value readout means selects, from among said accumulated values and numbers of said previous motion vectors that are associated a position which is spatially a distance equivalent to a magnitude of said current motion vector away from a location of said current motion vector in a direction opposite to said current motion vector or that are associated with vicinity of said position, said accumulated value and number of said motion vector.

Claim 3

The moving object detection apparatus according to claim 1 , wherein said motion vector removal means includes: previous vector storage means for storing said previous motion vectors; randomness calculation means for calculating, based on said current motion vectors and said previous motion vectors read from said previous vector storage means, instantaneous and stationary randomness of groups of motion vectors that are spatially and temporally adjacent to each said motion vector; and random vector removal means for removing said motion vectors that are determined, based on said instantaneous randomness and said stationary randomness, not to correspond to said moving object.

Claim 4

The moving object detection apparatus according to claim 2 , wherein said random vector removal means removes said motion vectors when said instantaneous and stationary randomness are greater than or equal to predetermined threshold values.

Claim 5

A moving object detection apparatus, comprising: motion vector calculation means for calculating motion vectors from an input image; motion vector removal means for removing a motion vector having high randomness from said calculated motion vectors; motion vector accumulation means for temporally accumulating each said motion vector not removed by said motion vector removal means, and for calculating an accumulated number of occurrences and an accumulated value of each motion vector; and moving object detection means for determining, based on said calculated accumulated value and calculated accumulated number of occurrences of each motion vector, whether each said motion vector corresponds to a moving object, wherein said motion vector calculation means includes: area disposition means for disposing a plurality of areas on said input image; previous image storage means for storing said previous input image; and correlation calculation means for calculating a correlation between said areas on said previous input image and corresponding areas on said current input image to obtain said motion vector on each said area disposed by said area disposition means, said previous input image being read from said previous image storage means, said area disposition means disposes each said area on the same position at all times, and wherein said motion vector accumulation means includes: accumulated value storage means for storing said accumulated values and numbers of said previous motion vectors such that said accumulated values and numbers are to be associated with the preceding positions of said motion vectors; accumulated value readout means for reading out said accumulated values and numbers of said previous motion vectors from said accumulated value storage means; and accumulated value update means for calculating new said accumulated values and numbers of said motion vectors by performing an update based on said accumulated values and numbers of said previous motion vectors and a magnitude and position of said current motion vector, said accumulated values and numbers of said previous motion vectors being stored in said accumulated value storage means; and said accumulated value readout means selects, from among said accumulated values and numbers of said previous motion vectors that are associated with a position which is spatially a distance equivalent to a magnitude of said current motion vector away from a location of said current motion vector in a direction opposite to said current motion vector or that are associated with a vicinity of said position, said accumulated value and number of said motion vector.

Claim 6

A moving object detection apparatus, comprising: motion vector calculation means for calculating motion vectors from an input image; motion vector removal means for removing a motion vector having high randomness from said calculated motion vectors; motion vector accumulation means for temporally accumulating each said motion vector not removed by said motion vector removal means, and for calculating an accumulated number of occurrences and an accumulated value of each motion vector; and moving object detection means for determining, based on said calculated accumulated value and calculated accumulated number of occurrences of each motion vector, whether each said motion vector corresponds to a moving object, wherein said motion vector calculation means includes: area disposition means for disposing a plurality of areas on said input image; previous image storage means for storing said previous input image; and correlation calculation means for calculating a correlation between said areas on said previous input image and corresponding areas on said current input image to obtain said motion vector on each said area disposed by said area disposition means, said previous input image being read from said previous image storage means, said area disposition means disposes each said area on the same position at all times, and wherein said motion vector accumulation means includes: accumulated value storage means for storing said accumulated values and numbers of said previous motion vectors such that said accumulated values and numbers are to be associated with the preceding positions of said motion vectors; accumulated value readout means for reading out said accumulated values and numbers of said previous motion vectors from said accumulated value storage means; and accumulated value update means for calculating new said accumulated values and numbers of said motion vectors by performing an update based on said accumulated values and numbers of said previous motion vectors and a magnitude and position of said current motion vector, said accumulated values and numbers of said previous motion vectors being stored in said accumulated value storage means; and said accumulated value readout means selects, from among said accumulated values and numbers of said previous motion vectors that are associated with a position which is spatially a distance equivalent to a magnitude of said current motion vector away from a location of said current motion vector in a direction opposite to said current motion vector or that are associated with a vicinity of said position, said accumulated value and number of said motion vector.

Claim 7

The moving object detection apparatus according to claim 4 , wherein said accumulated value readout means selects said accumulated value and number of said previous motion vector whose difference with respect to said current motion vector is smallest.

Claim 8

The moving object detection apparatus according to claim 4 , wherein said accumulated value readout means selects said accumulated value and number of said previous motion vector whose accumulated value produces a maximum dot product with said current motion vector.

Claim 9

The moving object detection apparatus according to claim 4 , wherein: said accumulated value update means includes previous vector storage means for storing said previous motion vectors; and said accumulated value update means resets said accumulated value and number of said previous motion vector associated with said position that is determined, based on said previous motion vectors stored in said previous vector storage means and said current motion vector calculated by said motion vector calculation means, not to correspond to said moving object, said accumulated value and number of said previous motion vector being stored in said accumulated value storage means.

Claim 10

A moving object detection method, comprising: calculating motion vectors from an input image; removing a motion vector having high randomness from the calculated motion vectors; temporally accumulating each motion vector not removed from the calculated motion vectors, and calculating an accumulated number of occurrences and an accumulated value of each motion vector; and determining by use of a moving object detection device, based on the calculated accumulated value and the calculated accumulated number of occurrences of each motion vector, whether each motion vector corresponds to a moving object, wherein the temporally accumulating and calculating includes: storing said accumulated values and numbers of said previous motion vectors such that said accumulated values and numbers are to be associated with the preceding positions of said motion vectors, reading out the stored accumulated values and numbers of said previous motion vectors, and calculating new said accumulated values and numbers of said motion vectors by performing an update based on the stored accumulated values and numbers of said previous motion vectors and a magnitude and position of said current motion vector, said reading out involves selecting, from among said accumulated values and numbers of said previous motion vectors that are associated with a position which is spatially a distance equivalent to a magnitude of said current motion vector away from a location of said current motion vector in a direction opposite to said current motion vector or that are associated with a vicinity of said position, said accumulated value and number of said motion vector.

Claim 11

A moving object detection method, comprising: calculating motion vectors from an input image; removing a motion vector having high randomness from the calculated motion vectors; temporally accumulating each motion vector not removed from the calculated motion vectors, and calculating an accumulated number of occurrences and an accumulated value of each motion vector; and determining by use of a moving object detection device, based on the calculated accumulated value and the calculated accumulated number of occurrences of each motion vector, whether each motion vector corresponds to a moving object, wherein the temporally accumulating and calculating includes: storing said accumulated values and numbers of said previous motion vectors such that said accumulated values and numbers are to be associated with the preceding positions of said motion vectors, reading out the stored accumulated values and numbers of said previous motion vectors, and calculating new said accumulated values and numbers of said motion vectors by performing an update based on the stored accumulated values and numbers of said previous motion vectors and a magnitude and position of said current motion vector, said reading out involves selecting, from among said accumulated values and numbers of said previous motion vectors that are associated with a position which is spatially a distance equivalent to a magnitude of said current motion vector away from a location of said current motion vector in a direction opposite to said current motion vector or that are associated with a vicinity of said position, said accumulated value and number of said motion vector.

Claim 12

A computer-readable storage medium having stored thereon a program for causing an apparatus to perform a moving object detection method, the method comprising: calculating motion vectors from an input image; removing a motion vector having high randomness from the calculated motion vectors; temporally accumulating each motion vector not removed from the calculated motion vectors, and calculating an accumulated number of occurrences and an accumulated value of each motion vector; and determining, based on the calculated accumulated value and the calculated accumulated number of occurrences of each motion vector, whether each motion vector corresponds to a moving object, wherein the temporally accumulating and calculating includes: storing said accumulated values and numbers of said previous motion vectors such that said accumulated values and numbers are to be associated with the preceding positions of said motion vectors, reading out the stored accumulated values and numbers of said previous motion vectors, and calculating new said accumulated values and numbers of said motion vectors by performing an update based on the stored accumulated values and numbers of said previous motion vectors and a magnitude and position of said current motion vector, said reading out involves selecting, from among said accumulated values and numbers of said previous motion vectors that are associated with a position which is spatially a distance equivalent to a magnitude of said current motion vector away from a location of said current motion vector in a direction opposite to said current motion vector or that are associated with a vicinity of said position, said accumulated value and number of said motion vector.

Claim 13

A computer-readable storage medium having stored thereon a program for causing an apparatus to perform a moving object detection method, the method comprising: calculating motion vectors from an input image; removing a motion vector having high randomness from the calculated motion vectors; temporally accumulating each motion vector not removed from the calculated motion vectors, and calculating an accumulated number of occurrences and an accumulated value of each motion vector; and determining, based on the calculated accumulated value and the calculated accumulated number of occurrences of each motion vector, whether each motion vector corresponds to a moving object, wherein the temporally accumulating and calculating includes: storing said accumulated values and numbers of said previous motion vectors such that said accumulated values and numbers are to be associated with the preceding positions of said motion vectors, reading out the stored accumulated values and numbers of said previous motion vectors, and calculating new said accumulated values and numbers of said motion vectors by performing an update based on the stored accumulated values and numbers of said previous motion vectors and a magnitude and position of said current motion vector, said reading out involves selecting, from among said accumulated values and numbers of said previous motion vectors that are associated with a position which is spatially a distance equivalent to a magnitude of said current motion vector away from a location of said current motion vector in a direction opposite to said current motion vector or that are associated with a vicinity of said position, said accumulated value and number of said motion vector.

Claim 14

A moving object detection apparatus, comprising: a motion vector calculation section calculating motion vectors from an input image; a motion vector removal section removing a motion vector having high randomness from said calculated motion vectors; a motion vector accumulation section temporally accumulating each said motion vector not removed by said motion vector removal section, and calculating an accumulated number of occurrences and an accumulated value of each motion vector; and a moving object detection section determining, based on said calculated accumulated value and calculated accumulated number of occurrences of each motion vector, whether each said motion vector corresponds to a moving object, wherein said motion vector accumulation section includes: an accumulated value storage section storing said accumulated values and numbers of said previous motion vectors such that said accumulated values and numbers are to be associated with the preceding positions of said motion vectors, an accumulated value readout section reading out said accumulated values and numbers of said previous motion vectors from said accumulated value storage section, and an accumulated value update section calculating new said accumulated values and numbers of said motion vectors by performing an update based on said accumulated values and numbers of said previous motion vectors and a magnitude and position of said current motion vector, said accumulated values and numbers of said previous motion vectors being stored in said accumulated value storage section, said accumulated value readout section selects, from among said accumulated values and numbers of said previous motion vectors that are associated with a position which is spatially a distance equivalent to a magnitude of said current motion vector away from a location of said current motion vector in a direction opposite to said current motion vector or that are associated with a vicinity of said position, said accumulated value and number of said motion vector.

Claim 15

A moving object detection apparatus, comprising: a motion vector calculation section calculating motion vectors from an input image; a motion vector removal section removing a motion vector having high randomness from said calculated motion vectors; a motion vector accumulation section temporally accumulating each said motion vector not removed by said motion vector removal section, and calculating an accumulated number of occurrences and an accumulated value of each motion vector; and a moving object detection section determining, based on said calculated accumulated value and calculated accumulated number of occurrences of each motion vector, whether each said motion vector corresponds to a moving object, wherein said motion vector accumulation section includes: an accumulated value storage section storing said accumulated values and numbers of said previous motion vectors such that said accumulated values and numbers are to be associated with the preceding positions of said motion vectors, an accumulated value readout section reading out said accumulated values and numbers of said previous motion vectors from said accumulated value storage section, and an accumulated value update section calculating new said accumulated values and numbers of said motion vectors by performing an update based on said accumulated values and numbers of said previous motion vectors and a magnitude and position of said current motion vector, said accumulated values and numbers of said previous motion vectors being stored in said accumulated value storage section, said accumulated value readout section selects, from among said accumulated values and numbers of said previous motion vectors that are associated with a position which is spatially a distance equivalent to a magnitude of said current motion vector away from a location of said current motion vector in a direction opposite to said current motion vector or that are associated with a vicinity of said position, said accumulated value and number of said motion vector.