US 12,294,813 B2
Distributed image analysis method and storage medium for performing image analysis based on image analysis results from servers
Ting Lei, Hangzhou (CN); and Chao Zou, Shenzhen (CN)
Assigned to HUAWEI TECHNOLOGIES CO., LTD., Shenzhen (CN)
Filed by HUAWEI TECHNOLOGIES CO., LTD., Guangdong (CN)
Filed on Jun. 17, 2021, as Appl. No. 17/350,205.
Application 17/350,205 is a continuation of application No. PCT/CN2019/124967, filed on Dec. 13, 2019.
Claims priority of application No. 201811554821.4 (CN), filed on Dec. 18, 2018.
Prior Publication US 2021/0312587 A1, Oct. 7, 2021
Prior Publication US 2022/0138893 A9, May 5, 2022
Int. Cl. H04N 7/18 (2006.01); G06T 1/20 (2006.01); G06T 1/60 (2006.01); G06T 5/50 (2006.01); G06T 7/33 (2017.01); G06V 10/40 (2022.01); G06V 10/74 (2022.01); G06V 20/52 (2022.01)
CPC H04N 7/18 (2013.01) [G06T 1/20 (2013.01); G06T 1/60 (2013.01); G06T 5/50 (2013.01); G06T 7/33 (2017.01); G06V 10/40 (2022.01); G06V 10/74 (2022.01); G06V 20/52 (2022.01)] 10 Claims
OG exemplary drawing
 
1. A distributed image analysis method performed by an image analysis system comprising a plurality of first servers and a second server, wherein the method comprises:
obtaining, by each of the plurality of first servers, a plurality of candidate images of a result object, wherein the obtaining of the plurality of candidate images of the result object comprises:
performing feature extraction on images stored on the first server to obtain a feature value of each of the stored images,
comparing the feature values of the stored images to determine, for each of the stored images, a similarity value indicating similarity between the stored image and each other stored image in the stored images, and
determining images, in the stored images, having similarity values greater than or equal to a similarity threshold as the plurality of candidate images of the result object;
separately sending, by each of the plurality of first servers, a result set obtained by the first server to the second server, wherein the result set comprises an index image of the result object, and the index image is an image determined based on the plurality of candidate images of the result object and is associated with an object frequency indicating a quantity of times that the result object is photographed;
performing, by the second server, feature extraction on each of the index images in the result sets received from the plurality of first servers, to extract a feature value of the index image;
comparing, by the second server, the extracted feature values of the index images in the result sets received from the plurality of first servers, to determine, for each of the index images, a similarity value indicating similarity between the index image and each other index image in the index images in the result sets; and
when determining that the similarity values between at least some of the index images in the result sets received from the plurality of first servers are greater than or equal to a preset value, calculating, by the second server, a sum of object frequencies of those index images between which the similarity values are greater than or equal to the preset value, the sum indicating a total number of times that the result object is photographed,
wherein each of the plurality of first servers is a physical device including a memory and a processing circuitry, and the second server is a physical device including a memory and a processing circuitry,
wherein the method further comprises, after the obtaining, by each of the plurality of first servers, a plurality of candidate images of a result object;
obtaining, by the first server, a resource address of the plurality of candidate images, and
obtaining, by the first server, the result set, wherein the result set comprises the resource address; and
after the calculating of the sum of object frequencies corresponding to the index images,
accessing, by the second server, the plurality of candidate images through the resource address.