US 11,704,813 B2
Visual search method, visual search device and electrical device
Liuqing Zhang, Beijing (CN); Guohong Li, Beijing (CN); Yongjia Yu, Beijing (CN); and Xin Qiu, Beijing (CN)
Assigned to Baidu Online Network Technology (Beijing) Co., Ltd., Beijing (CN)
Filed by Baidu Online Network Technology (Beijing) Co., Ltd., Beijing (CN)
Filed on May 11, 2020, as Appl. No. 16/871,206.
Claims priority of application No. 201911204967.0 (CN), filed on Nov. 29, 2019.
Prior Publication US 2021/0166401 A1, Jun. 3, 2021
Int. Cl. G06K 9/00 (2022.01); G06T 7/20 (2017.01); G06T 7/00 (2017.01); G06T 7/70 (2017.01); G06F 1/329 (2019.01)
CPC G06T 7/20 (2013.01) [G06F 1/329 (2013.01); G06T 7/70 (2017.01); G06T 7/97 (2017.01); Y02D 10/00 (2018.01)] 20 Claims
OG exemplary drawing
 
1. A visual search method, comprising:
determining a first object in a visual search process, wherein a power consumption of the first object in the visual search process is greater than a power consumption of a second object in the visual search process, and the first object comprises at least one of a program and an algorithm, the second object comprises at least one of a program and an algorithm; and
performing a visual search through an optimized object of the first object;
wherein the first object comprises a tracking algorithm, and the optimized object comprises an optimized tracking algorithm obtained by optimizing a scheduling of the tracking algorithm; and
wherein performing the visual search through the optimized object of the first object comprises:
acquiring an ith frame of image, wherein i is a positive integer,
performing a target detection on the ith frame of image, to acquire a position of a target in the ith frame of image, and
performing a target tracking with the optimized tracking algorithm comprising:
performing, based on the position of the target in the ith frame of image, the target tracking on n frames of images from m frames of images subsequent to the ith frame of image without performing the target tracking and the target detection on the remaining (m−n) frames of images from the m frames of images, wherein m and n are both positive integers, and n is less than m;
wherein the optimized tracking algorithm is obtained by:
determining, based on a pixel displacement between the ith frame of image and each of N frames of image prior to the ith frame of image, a total pixel displacement;
selecting, based on the total pixel displacement, a tracking frequency coefficient from a tracking frequency coefficient set comprising a plurality of frequency coefficients;
determining, based on the tracking frequency coefficient and a present frequency for acquiring frames of image, a first tracking frequency on which the target tracking is performed; and
optimizing the tracking algorithm with the first tracking frequency to obtain the optimized tracking algorithm;
wherein the first tracking frequency is equal to the present frequency multiplied by the tracking frequency coefficient, and the performing the target tracking with the optimized tracking algorithm comprises:
determining the n frames of images from the m frames of images subsequent to the ith frame of image based on the first tracking frequency, such that a ratio of n to m matches with the first tracking frequency.