Robust cross-domain attitude estimation method and system based on image mixing mechanism

• Main Authors: WANG HAIBIN, JI WENFENG
• Format: Patent
• Language: Chinese; English
• Published: 12.11.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention provides a robust cross-domain attitude estimation method and system based on an image mixing mechanism, and belongs to the technical field of image processing. The method comprises the following steps that: a to-be-processed image is obtained; and an attitude estimation result is obtained according to the obtained image and an OpenPose attitude estimation model, wherein the OpenPose attitude estimation model is trained by adopting a fusion data set, weighted summation is performed on pixels of a certain frame of image in a source domain training set and pixels of a certain frame of image randomly selected in the target domain data set to obtain a fusion image, and fusion operation is performed on each frame of image in the source domain training set in sequence to obtain the fusion data set. According to the method, the data fusion method is introduced on a model input level to carry out cross-domain attitude estimation, the performance of the model on a target domain is improved by adding nois