Object detecting apparatus and learning apparatus for the same
View Patent ↗An object detecting apparatus capable of suppressing an increase of processing loads with high accuracy and a learning apparatus for the same are provided. An object detecting apparatus includes an image window extracting portion ( 210 ) for extracting an image window as a partial area of an image in plural from an input image, and a network identifier ( 590 ) for detecting a presence of an object from extracted image windows respectively by using a node network in which nodes each having an identifier for identifying the object stored in a storing portion ( 502 ) are connected as a network.
1. An object detecting apparatus, comprising:
a setting portion which sets a rectangular template defined by a plurality of rectangular blocks with respect to an image;
a first calculation portion which calculates a reference brightness value based on brightness values of the rectangular blocks contained in the rectangular template, and each individual brightness value of the respective rectangular blocks;
a second calculation portion which calculates spatial distribution information on the basis of magnitude relation between the reference brightness value and the each individual brightness value of the respective rectangular blocks; and
a determination portion which determines whether the image inside the rectangular template contains an object with reference to the spatial distribution information.
2. The object detecting apparatus according to claim 1 , wherein the plurality of rectangular blocks differ in size from one another.
3. The object detecting apparatus according to claim 1 , comprising a labeling portion which labels, when the individual brightness value of the respective rectangular blocks is equal to or larger than the reference brightness value, the corresponding rectangular block as a first label, and labels, when the individual brightness value of the respective rectangular blocks is smaller than the reference brightness value, the corresponding rectangular block as a second label, wherein
the spatial distribution information is constituted by a group of the first label and the second label which the respective rectangular blocks are labeled.
4. The object detecting apparatus according to claim 3 , wherein a strength is calculated on the basis of a difference between an average brightness value of the individual brightness values of all rectangular blocks which are labeled as the first label and an average brightness value of the individual brightness values of all rectangular blocks which are labeled as the second label, and
the determination portion determines whether the image to which the rectangular template is set contains an object with reference to the spatial distribution information and the strength.
5. The object detecting apparatus according to claim 1 , wherein the plurality of rectangular blocks are separate away from one another.
6. The object detecting apparatus according to claim 1 , wherein the plurality of rectangular blocks overlap one another.
7. An object detecting method, comprising:
setting a rectangular template defined by a plurality of rectangular blocks with respect to an image;
calculating a reference brightness value based on brightness values of the rectangular blocks contained in the rectangular template;
calculating each individual brightness value of the respective rectangular blocks;
calculating spatial distribution information on the basis of magnitude relation between the reference brightness value and the each individual brightness value of the respective rectangular blocks; and
determining whether the image inside the rectangular template contains an object with reference to the spatial distribution information.