US 11,928,184 B2
Generation of a second object model based on a first object model for use in object matching
Ola Friman, Linköping (SE); Anders Moe, Linköping (SE); and Kristoffer Öfjäll, Stenkullen (SE)
Assigned to SICK IVP AB, Linköping (SE)
Filed by SICK IVP AB, Linköping (SE)
Filed on Jun. 16, 2021, as Appl. No. 17/349,241.
Claims priority of application No. 20183322 (EP), filed on Jun. 30, 2020.
Prior Publication US 2021/0406583 A1, Dec. 30, 2021
Int. Cl. G06V 10/74 (2022.01); G06F 18/21 (2023.01); G06F 18/214 (2023.01); G06F 18/22 (2023.01); G06V 10/774 (2022.01); G06V 20/64 (2022.01)
CPC G06F 18/217 (2023.01) [G06F 18/214 (2023.01); G06F 18/22 (2023.01); G06V 10/74 (2022.01); G06V 10/774 (2022.01); G06V 20/64 (2022.01)] 11 Claims
OG exemplary drawing
 
1. Method, performed by one or more devices, for supporting generation of a second object model, based on a first object model, for object matching according to an object matching algorithm, the first object model comprising object features of an imaged reference object, wherein the method comprises:
obtaining sub-models that comprise different sub-features, respectively, of said object features comprised in the first object model, and
providing contribution indicators for the sub-models, respectively, each contribution indicator of said contribution indicators indicating contribution of the sub-feature of the sub-model to incorrect matches, the contribution indicators being based on matching, according to the object matching algorithm, the first object model and the sub-models with at least one model optimization image, “MOI”, comprising predefined training features that when matched with the first object model result in at least said incorrect matches,
obtaining, per MOI, score maps comprising at least sub-model score maps for the sub-models, respectively, each score map being based on matching a corresponding model at different poses with the MOI, each score map comprising scores at different positions of the score map, each of said positions being associated with one or more of said poses, the score of a position indicating a level of matching between said one or more of said poses of the model associated with the position and the MOI,
wherein said provision of the contribution indicators comprises computing the contribution indicators based on said obtained score maps, wherein said score maps, obtained per MOI, further comprises a first model score map for the first object model, and
wherein the computation of each contribution indicator comprises computing, per MOI, a first sum, wherein the terms of the first sum comprise scores, respectively, from at least positions of the sub-model score map that correspond to said incorrect matches.