US 11,682,190 B2
Method, system, and device for detecting an object in a distorted image
Hampus Linse, Lund (SE); Song Yuan, Lund (SE); and Johan Förberg, Lund (SE)
Assigned to Axis AB, Lund (SE)
Filed by Axis AB, Lund (SE)
Filed on Apr. 3, 2020, as Appl. No. 16/839,083.
Claims priority of application No. EP19168368 (EP), filed on Apr. 10, 2019.
Prior Publication US 2020/0327691 A1, Oct. 15, 2020
Int. Cl. G06V 10/50 (2022.01); G06T 7/73 (2017.01); G06T 5/50 (2006.01); G06V 10/75 (2022.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06V 20/00 (2022.01)
CPC G06V 10/50 (2022.01) [G06T 5/50 (2013.01); G06T 7/73 (2017.01); G06V 10/7515 (2022.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06V 20/00 (2022.01); G06T 2207/20021 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/20221 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method for detecting an object in a single image of a scene, the single image being distorted by a predetermined distortion, the method comprising:
receiving the single image of the scene;
receiving an inverse function of a mathematical representation of the predetermined distortion,
wherein the mathematical representation of the predetermined distortion is an image transform which, when applied to a rectilinear image corresponding to the scene, results in the single image that is distorted by the predetermined distortion; and
detecting the object in the single image, wherein the detecting the object in the single image comprises:
sliding a sliding window over a plurality of positions in the single image that is distorted by the predetermined distortion, the sliding window comprising a feature detection pattern; and
for each position of the plurality of positions in the single image, transforming the sliding window based on a position-specific value of the inverse function of the mathematical representation of the predetermined distortion,
wherein the transforming the sliding window at each position of the plurality of positions in the single image comprises transforming the feature detection pattern of the sliding window such that resulting position-specific distortions of the feature detection pattern of the transformed sliding window corresponds to position-specific portions of the predetermined distortion such that the object is detected in the single image based on at least one of the position-specific distortions.