Probability distribution constructing method, probability distribution constructing apparatus, storage medium of probability distribution constructing program, subject detecting method, subject detecting apparatus, and storage medium of subject detecting program

A probability distribution constructing apparatus that constructs, based on plural images obtained by repeatedly photographing a field where a plurality of persons exist using a fixedly disposed camera, appearance probability distribution distributions of head portions of the plurality of persons repeatedly appearing in various sizes in the plural images, the probability distribution constructing apparatus comprising:

• an image acquiring section that acquires the plural images from the camera;
• an extracting section that extracts, from the plural images acquired by the image acquiring section, similar head portion shapes each of which is similar a predetermined typical shape of head portion of a person, together with sizes of the similar head portion shapes and positional information of the camera on a view angle;
• a filtering section that determines whether the similar head portion shapes extracted by the extracting section are head portions of persons or not, and that extracts head portions determined in the determination;
• and
• a probability distribution calculating section that calculates, (a) for each of plural sizes of the extracted head portions, an appearance location probability distribution whose probability variable is each location in the view angle and which represents appearance probabilities of head portions classified as having the same size as the each of the plural sizes, of the extracted head portions or (b) for each location in the view angle, an appearance size probability distribution whose probability variable is each of plural sizes of the extracted head portions and which represents appearance probabilities of head portions appeared in the each location in the view angle,
• wherein the filtering step is a step of (1) counting, for each of the extracted similar head portion shapes in the extracting step, the number of other similar head portion shapes except for the each of the extracted similar head portion shapes which exist in a search area having a predetermined size and being spread around the each of the extracted similar head portion shapes in the view angle and which are classified as having the same size as size of the each of the extracted similar head portion shapes, and (2) determining the each of the extracted similar head portion shapes (2a) as being one of the head portions when the number of the other similar head portion shapes obtained by the counting is equal to or larger than a predetermined threshold value being one or more and (2b) as not being one of the head portions when the number of the other similar head portion shapes obtained by the counting is less than the threshold value.

A non-transitory computer-readable medium that stores a probability distribution constructing program that is executed in a computer and allows the computer to operate as a probability distribution constructing apparatus that constructs, based on plural images obtained by repeatedly photographing a field where a plurality of persons exist using a fixedly disposed camera, appearance probability distributions of head portions of the plurality of persons repeatedly appearing in various sizes in the plural images, the probability distribution constructing program causes the computer to implement:

• an image acquiring section that acquires the plural images from the camera;
• an extracting section that extracts, from the plural images acquired by the image acquiring section, similar head portion shapes each of which is similar to a predetermined typical shape of a head portion of a person, together with sizes of the similar head portion shapes and positional information of the camera on a view angle;
• a filtering section that determines whether the similar head portion shapes extracted by the extracting section are head portions of persons or not, and that extracts head portions determined in the determination;
• and
• a probability distribution calculating section that calculates, (a) for each of plural sizes of the extracted head portions, an appearance location probability distribution whose probability variable is each location in the view angle and which represents appearance probabilities of head portions classified as having the same size as the each of the plural sizes, of the extracted head portions or (b) for each location in the view angle, an appearance size probability distribution whose probability variable is each of plural sizes of the extracted head portions and which represents appearance probabilities of head portions appeared in the each location in the view angle,
• wherein the filtering step is a step of (1) counting, for each of the extracted similar head portion shapes in the extracting step, the number of other similar head portion shapes except for the each of the extracted similar head portion shapes which exist in a search area having a predetermined size and being spread around the each of the extracted similar head portion shapes in the view angle and which are classified as having the same size as size of the each of the extracted similar head portion shapes, and (2) determining the each of the extracted similar head portion shapes (2a) as being one of the head portions when the number of the other similar head portion shapes obtained by the counting is equal to or larger than a predetermined threshold value being one or more and (2b) as not being one of the head portions when the number of the other similar head portion shapes obtained by the counting is less than the threshold value.