Person searching method and apparatus and image processing device
Embodiments of this disclosure provide a person searching method and apparatus and an image processing device. The method includes: performing person detection on a reference image, determining a query person, and acquiring property information on the query person; performing person detection on at least one video surveillance image, determining all candidate persons, and acquiring property information on all the candidate persons; calculating a match score of each candidate person according to the property information on the query person and property information on each candidate person, and selecting a predetermined number of candidate persons in a descending order of the match scores; and calculating similarities between the query person and the selected candidate persons, and taking a candidate person in the selected candidate persons having a highest similarity with the query person as the query person. In the embodiments of this disclosure, the query person is found from the video surveillance image by person detection and two-step retrieval.
1. A person searching apparatus, characterized in that the apparatus comprises:
a first detecting unit configured to perform person detection on a reference image, determine a query person, and acquire property information on the query person;
a second detecting unit configured to perform person detection on at least one video surveillance image, determine all candidate persons, and acquire property information on all the candidate persons;
a selecting unit configured to calculate a match score of each candidate person according to the property information on the query person and property information on each candidate person, and select a predetermined number of candidate persons in a descending order of the match scores; and
a determining unit configured to calculate similarities between the query person and the selected candidate persons, and take a candidate person in the selected candidate persons having a highest similarity with the query person as the query person;
wherein the determining unit acquires high-dimensional features of the query person and all the selected candidate persons, and calculates a similarity between the query person and each selected candidate person according to the high-dimensional features of the query person and each selected candidate person; and
wherein the determining unit calculates the similarity by using a formula as below:
Similarity
i
=
Q
·
C
i
T
Q
C
i
T
,
i
=
1
,
…
,
n
;
where, i is a sequence number of a selected candidate person, Q is the high-dimensional feature of the query person, Ci is the high-dimensional feature of the selected candidate person, and m is a total number of the selected candidate persons.
2. The apparatus according to claim 1 , wherein the reference image is a pre-captured photo or a video surveillance image.
3. The apparatus according to claim 1 , wherein the video surveillance image is a fish-eye video image, and the apparatus further comprises:
a processing unit configured to de-fish the video surveillance image to obtain a de-fished video surveillance image, so that the second detecting unit performs the person detection on the de-fished video surveillance image.
4. The apparatus according to claim 1 , wherein the property information comprises an up-cloth type, a low-cloth type, an up-cloth color, and a low-cloth color.
5. The apparatus according to claim 1 , wherein the selecting unit extracts a probability of the property information on each candidate person according to the property information on the query person, and takes a product of the probability of the property information and a weight to which the property information corresponds as the match score of the candidate person.
6. The apparatus according to claim 5 , wherein the selecting unit calculates the match score of each candidate person by using a formula as below:
Score=Σ p i *w i ,i= 1, . . . , n;
where, p i is a probability of an i-th property of the candidate person, w i is a weight of the i-th property, i is a sequence number of a property in the property information, and n is a total number of the properties.
7. An image processing device, characterized in that the image processing device comprises the person searching apparatus as claimed in claim 1 .
8. A person searching method, characterized in that the method comprises:
performing person detection on a reference image, determining a query person, and acquiring property information on the query person;
performing person detection on at least one video surveillance image, determining all candidate persons, and acquiring property information on all the candidate persons;
calculating a match score of each candidate person according to the property information on the query person and property information on each candidate person, and selecting a predetermined number of candidate persons in a descending order of the match scores; and
calculating similarities between the query person and the selected candidate persons, and taking a candidate person in the selected candidate persons having a highest similarity with the query person as the query person;
wherein calculating similarities between the query person and the selected candidate persons comprises:
acquiring high-dimensional features of the query person and all the selected candidate persons, and calculating a similarity between the query person and each selected candidate person according to the high-dimensional features of the query person and each selected candidate person; and
wherein the similarity is calculated by using a formula as below:
Similarity
i
=
Q
·
C
i
T
Q
C
i
T
,
i
=
1
,
…
,
n
;
where, i is a sequence number of a selected candidate person, Q is the high-dimensional feature of the query person, Ci is the high-dimensional feature of the selected candidate person, and m is a total number of the selected candidate persons.