Method and device for determining duplicate video

1. A method of processing data, comprising:

acquiring multiple types of video features of a query video;

performing sequence comparisons for a plurality of existing videos respectively to obtain sequence comparison results according to the multiple types of video features of the query video;

performing a first ranking on the plurality of existing videos according to the sequence comparison results, and taking first n existing videos as first candidate videos according to a result of the first ranking, where n is a positive integer; and

determining a duplication level of the query video according to the sequence comparison results of the first candidate videos;

wherein the acquiring multiple types of video features of a query video further comprises:

extracting frames of the query video to obtain a plurality of frame images of the query video,

extracting multiple types of image features of each of the plurality of frame images as first image features, wherein the extracting multiple types of image features of each of the plurality of frame images comprises:

for each of the plurality of frame images, acquiring one or more shift vectors, determining an end point pointed by each shift vector by using each of the one or more shift vectors and taking any pixel in each of the plurality of frame images as a starting point, and

determining the multiple types of image features of each of the plurality of frame images according to an overall condition of differences between each starting point and each corresponding end point.

2. The method of claim 1 , wherein the acquiring multiple types of video features of a query video further comprises:

determining each video feature of the query video as a first video feature to obtain multiple types of first video features according to each of the first image features with an identical type of the plurality of frame images of the query video.

3. The method of claim 2 , wherein the extracting multiple types of image features of each of the plurality of frame images comprises:

for each of the plurality of frame images, performing multiple types of pooling processes step by step to obtain the multiple types of image features of the plurality of frame images as a pooling feature, wherein the multiple types of pooling processes comprise a maximum pooling process, a minimum pooling process and an average pooling process.

4. The method of claim 2 , wherein the performing sequence comparisons for a plurality of existing videos respectively to obtain sequence comparison results according to the multiple types of video features of the query video comprises:

acquiring multiple types of video features of one of the plurality of existing videos as second video features, wherein each of the second video features comprises a plurality of second image features;

determining a unit similarity between each of the second image features and each of the first image features of the same type respectively to obtain multiple types of unit similarities;

determining a minimum value or an average value of the multiple types of unit similarities, and determining a similarity matrix of the plurality of existing videos according to the minimum value or the average value of the multiple types of unit similarities; and

determining a sequence comparison score according to the similarity matrix, wherein the sequence comparison score is configured for indicating a similarity of the plurality of existing videos and the query video.

5. The method of claim 4 , wherein the determining a sequence comparison score according to the similarity matrix comprises:

determining the sequence comparison score according to a straight line in the similarity matrix.

6. The method of claim 4 , wherein the performing sequence comparisons for a plurality of existing videos respectively to obtain sequence comparison results according to the multiple types of video features of the query video further comprises:

determining a duplicate video segment of the plurality of existing videos and the query video according to the similarity matrix.

7. The method of claim 2 , wherein the performing sequence comparisons for a plurality of existing videos respectively to obtain sequence comparison results according to the multiple types of video features of the query video comprises:

performing a second ranking on the plurality of existing videos according to each of the first image features in at least one of the multiple types of first video features, according to a result of the second ranking, taking first k of the plurality of existing videos as second candidate videos, where k is a positive integer; and

performing a sequence comparison for each of the second candidate videos respectively to obtain the sequence comparison result.

8. The method of claim 7 , wherein the performing a second ranking on the plurality of existing videos according to each of the first image features in at least one of the multiple types of first video features comprises:

performing a term frequency-inverse document frequency ranking on the plurality of existing videos using each of the first image features in at least one of the multiple types of first video features as an index request.

9. The method of claim 7 , wherein the according to each of the first image features with an identical type of the plurality of frame images of the query video, determining each video feature of the query video as a first video feature comprises:

binarizing the first image features; and

determining the first video feature according to binarized first image features of the plurality of frame images.

10. A device for determining duplicate video, comprising:

at least one processor; and

at least one memory communicatively coupled to the at least one processor and storing instructions that upon execution by the at least one processor cause the device to:

acquire multiple types of video features of a query video;

perform sequence comparisons for a plurality of existing videos respectively to obtain sequence comparison results according to the multiple types of video features of the query video;

perform a first ranking on the plurality of existing videos according to the sequence comparison results, and take first n existing videos as first candidate videos according to the result of the first ranking, wherein n is a positive integer; and

determine the duplication level of the query video according to the sequence comparison results of the first candidate videos;

wherein instructions that upon execution by the at least one processor cause the device to acquire multiple types of video features of a query video further comprise instructions that upon execution by the at least one processor cause the device to:

extract frames of the query video to obtain a plurality of frame images of the query video,

extract multiple types of image features of each of the plurality of frame images as first image features, wherein instructions that upon execution by the at least one processor cause the device to extract multiple types of image features of each of the plurality of frame images further comprise instructions that upon execution by the at least one processor cause the device to:

for each of the plurality of frame images, acquire one or more shift vectors, determine an end point pointed by the shift vector by using each of the one or more shift vectors, and take any pixel in each of the plurality of frame images as a starting point, and

determine the multiple types of image features of each of the plurality of frame images according to an overall condition of differences between each starting point and each corresponding end point.

11. The device of claim 10 , wherein the at least one memory further stores instructions that upon execution by the at least one processor cause the device to:

determine each video feature of the query video as a first video feature to obtain multiple types of first video features according to each of the first image features with an identical type of the plurality of frame images of the query video.

12. A non-transitory computer-readable storage medium, storing non-transitory computer-readable instructions to perform operations when the non-transitory computer-readable instructions are executed by a computing device, the operations comprising:

acquiring multiple types of video features of a query video;

performing sequence comparisons for a plurality of existing videos respectively to obtain sequence comparison results according to the multiple types of video features of the query video;

performing a first ranking on the plurality of existing videos according to the sequence comparison results, and taking first n existing videos as first candidate videos according to a result of the first ranking, where n is a positive integer; and

determining a duplication level of the query video according to the sequence comparison results of the first candidate videos;

wherein the acquiring multiple types of video features of a query video further comprises:

extracting frames of the query video to obtain a plurality of frame images of the query video,

extracting multiple types of image features of each of the plurality of frame images as first image features, wherein the extracting multiple types of image features of each of the plurality of frame images comprises:

for each of the plurality of frame images, acquiring one or more shift vectors, determining an end point pointed by the shift vector by using each of the one or more shift vectors and taking any pixel in each of the plurality of frame images as a starting point, and

determining the multiple types of image features of each of the plurality of frame images according to an overall condition of differences between each starting point and each corresponding end point.

13. The device of claim 11 , wherein the at least one memory further stores instructions that upon execution by the at least one processor cause the device to:

for each of the plurality of frame images, perform multiple types of pooling processes step by step to obtain the multiple types of image features of the plurality of frame images as a pooling feature, wherein the multiple types of pooling processes comprise a maximum pooling process, a minimum pooling process and an average pooling process.

14. The device of claim 11 , wherein the at least one memory further stores instructions that upon execution by the at least one processor cause the device to:

acquire multiple types of video features of one of the plurality of existing videos as second video features, wherein each of the second video features comprises a plurality of second image features;

determine a unit similarity between each of the second image features and each of the first image features of the same type respectively to obtain multiple types of unit similarities;

determine a minimum value or an average value of the multiple types of unit similarities, and determine a similarity matrix of the plurality of existing videos according to the minimum value or the average value of the multiple types of unit similarities; and

determine a sequence comparison score according to the similarity matrix, wherein the sequence comparison score is configured for indicating a similarity of the plurality of existing videos and the query video.

15. The device of claim 14 , wherein the at least one memory further stores instructions that upon execution by the at least one processor cause the device to:

determine the sequence comparison score according to a straight line in the similarity matrix.

16. The device of claim 14 , wherein the at least one memory further stores instructions that upon execution by the at least one processor cause the device to:

determine a duplicate video segment of the plurality of existing videos and the query video according to the similarity matrix.

17. The device of claim 11 , wherein the at least one memory further stores instructions that upon execution by the at least one processor cause the device to:

perform a second ranking on the plurality of existing videos according to each of the first image features in at least one of the multiple types of first video features, according to a result of the second ranking, take first k of the plurality of existing videos as second candidate videos, where k is a positive integer; and

perform a sequence comparison for each of the second candidate videos respectively to obtain the sequence comparison result.

18. The device of claim 17 , wherein the at least one memory further stores instructions that upon execution by the at least one processor cause the device to:

perform a term frequency-inverse document frequency ranking on the plurality of existing videos using each of the first image features in at least one of the multiple types of first video features as an index request.