People counting device, people counting method and people counting program

1. A people counting device comprising:

a person presence region extractor which extracts a region in which a person is present by performing person detection in relation to an input image;

a person presence state estimator which estimates a person presence state from an extraction result of the person presence region extractor, the person presence state estimator estimates a direction in which persons are overlapping based upon the extraction result of the person presence region extractor;

a portion detector selector which selects a type of portion detector to be used for people count measurement based on an estimation result of the person presence state estimator, and outputs portion detector information;

a portion detection processor which performs a portion detection process for detecting a portion of a person based on the portion detector information output from the portion detector selector; and

a people counter which obtains a number of people within the input image from a portion detection process result of the portion detection processor.

2. The people counting device according to claim 1 ,

wherein the portion detector selector selects at least one of an upper body detector, a lower body detector, a right-side body detector and a left-side body detector as the portion detector.

3. The people counting device according to claim 2 ,

wherein the person presence state estimator determines that the persons are overlapping with one another in a vertical direction when rectangle information extracted by the person presence region extractor is long in the vertical direction and not in the horizontal direction, in comparison with a size of a person, and

wherein the portion detector selector selects the upper body detector when the person presence state estimator determines that the persons are overlapping with one another in the vertical direction.

4. The people counting device according to claim 2 ,

wherein the person presence state estimator determines that the persons are overlapping with one another in the horizontal direction when rectangle information extracted by the person presence region extractor is long in the horizontal direction and not in the vertical direction, in comparison with the size of a person, and

wherein the portion detector selector selects the right-side body detector and the left-side body detector when the person presence state estimator determines that the persons are overlapping with one another in the horizontal direction.

5. The people counting device according to claim 2 ,

wherein the person presence state estimation estimator determines that the persons are overlapping one another over a wide range when rectangle information extracted by the person presence region extractor is long in the horizontal direction and in the vertical direction, in comparison with the size of a person, and

wherein the portion detector selector selects the upper body detector, the lower body detector, the right-side body detector and the left-side body detector when the person presence state estimator determines that the persons are overlapping with one another over a wide range.

6. A people counting method comprising:

extracting a region in which a person is present by performing person detection in relation to an input image;

estimating a person presence state from a result in which a region where a person is present is extracted, the estimating estimates a direction in which persons are overlapping based upon the extraction result of the extracting;

selecting a type of portion detector to be used for people count measurement based on a result in which a person presence state is estimated, and outputting portion detector information;

performing, by a processor, a portion detection process for detecting a portion of a person based on the portion detector information of the selected portion detector; and

obtaining a number of people within the input image from a result of the detection process.

7. A non-transitory computer readable storage medium storing a people counting program, the people counting program comprising instructions, which, when executed by a computer, cause the computer to:

extract a region in which a person is present by performing person detection in relation to an input image;

estimate a person presence state from a result in which a region where a person is present is extracted, by estimating a direction in which persons are overlapping based upon the extraction result of the extracting a region;

select a type of portion detector to be used for people count measurement based on a result in which a person presence state is estimated, and outputting portion detector information;

perform a portion detection process for detecting a portion of a person based on the portion detector information of the selected portion detector; and

obtain a number of people within the input image from a result of the detection process.

8. The people counting method according to claim 6 ,

wherein the selecting selects at least one of an upper body detector, a lower body detector, a right-side body detector and a left-side body detector as the portion detector.

9. The people counting method according to claim 8 ,

wherein the estimating a person presence state determines that the persons are overlapping with one another in a vertical direction when rectangle information extracted by the extracting is long in the vertical direction and not in the horizontal direction, in comparison with a size of a person, and

wherein the selecting a type of a portion detector selects the upper body detector when the estimating a person presence state determines that the persons are overlapping with one another in the vertical direction.

10. The people counting method according to claim 8 ,

wherein the estimating a person presence state determines that the persons are overlapping with one another in the horizontal direction when rectangle information extracted by the person extracting is long in the horizontal direction and not in the vertical direction, in comparison with the size of a person, and

wherein the selecting a type of a portion detector selects the right-side body detector and the left-side body detector when the estimating a person presence state determines that the persons are overlapping with one another in the horizontal direction.

11. The people counting method according to claim 8 ,

wherein the estimating a person presence state determines that the persons are overlapping one another over a wide range when rectangle information extracted by the person extracting is long in the horizontal direction and in the vertical direction, in comparison with the size of a person, and

wherein the selecting a type of a portion detector selects the upper body detector, the lower body detector, the right-side body detector and the left-side body detector when the estimating a person presence state determines that the persons are overlapping with one another over a wide range.

12. The non-transitory computer readable storage medium according to claim 7 ,

wherein the selecting of a portion detector type selects at least one of an upper body detector, a lower body detector, a right-side body detector and a left-side body detector as the portion detector.

13. The non-transitory computer readable storage medium according to claim 12 ,

wherein the estimate of a person presence state determines that the persons are overlapping with one another in a vertical direction when rectangle information extracted is long in the vertical direction and not in the horizontal direction, in comparison with a size of a person, and

wherein the selecting a type of a portion detector selects the upper body detector when the person presence state estimate determines that the persons are overlapping with one another in the vertical direction.

14. The non-transitory computer readable storage medium according to claim 12 ,

wherein the estimate of a person presence state determines that the persons are overlapping with one another in the horizontal direction when rectangle information extracted is long in the horizontal direction and not in the vertical direction, in comparison with the size of a person, and

wherein the selecting a type of a portion detector selects the right-side body detector and the left-side body detector when the person presence state estimate determines that the persons are overlapping with one another in the horizontal direction.

15. The non-transitory computer readable storage medium according to claim 12 ,

wherein the estimate of a person presence state determines that the persons are overlapping one another over a wide range when rectangle information extracted is long in the horizontal direction and in the vertical direction, in comparison with the size of a person, and

wherein the selecting a type of a portion detector selects the upper body detector, the lower body detector, the right-side body detector and the left-side body detector when the person presence state estimate determines that the persons are overlapping with one another over a wide range.

16. The people counting device according to claim 1 ,

wherein the portion detector selector selects at least two of an upper body detector, a lower body detector, a right-side body detector and a left-side body detector as the portion detector.

17. The people counting method according to claim 6 ,

wherein the selecting selects at least two of an upper body detector, a lower body detector, a right-side body detector and a left-side body detector as the portion detector.

18. The non-transitory computer readable storage medium according to claim 7 ,

wherein the selecting of a portion detector type selects at least two of an upper body detector, a lower body detector, a right-side body detector and a left-side body detector as the portion detector.