US 11,948,368 B2
Real-time target detection and 3d localization method based on single frame image
Xichuan Zhou, Chongqing (CN); Chunqiao Long, Chongqing (CN); and Yicong Peng, Chongqing (CN)
Assigned to Chongqing University, Chongging (CN)
Filed by ChongQing University, Chongqing (CN)
Filed on Jun. 25, 2021, as Appl. No. 17/357,998.
Claims priority of provisional application 63/079,230, filed on Sep. 16, 2020.
Prior Publication US 2022/0083789 A1, Mar. 17, 2022
Int. Cl. G06K 9/00 (2022.01); G06F 18/22 (2023.01); G06K 9/62 (2022.01); G06N 3/04 (2023.01); G06N 3/045 (2023.01); G06N 3/08 (2023.01); G06T 7/50 (2017.01); G06V 20/56 (2022.01)
CPC G06V 20/56 (2022.01) [G06F 18/22 (2023.01); G06N 3/045 (2023.01); G06N 3/08 (2013.01); G06T 7/50 (2017.01); G06T 2207/10024 (2013.01)] 10 Claims
OG exemplary drawing
 
1. A real-time object detection and 3D localization method based on a single frame image, comprising the following steps:
S1: inputting a 2D infrared image;
S2: performing feature extraction on the 2D infrared image, and extracting features of a deep network and a shallow network respectively;
S3: carrying out 2D object detection and applying to subsequent modules;
S4: estimating vertices, instance-level depth information and center point of a 3D-box respectively;
S5: adding a regularization term for maintaining horizontal locality into prediction of the center point of the 3D-box to constrain and optimize the prediction of the center point of the 3D-box; and
S6: outputting a 2D infrared image with a 3D-box tag in combination with predictions of all modules.