Object 6D posture prediction method based on RGB image and coordinate system transformation

Abstract

The invention discloses an object 6D posture prediction method based on RGB image and coordinate system transformation. The method includes decoupling 6D posture parameters of the object, and realizing 6D posture prediction of the object by solving six parameter solving problems. The 3D translation of the object is predicted by positioning the center of the object in the image and estimating the distance between the object and the camera. 3D rotation of a predicted object is converted into a pose of a predicted camera through coordinate system transformation, pose parameters of the camera aredecoupled into an azimuth angle, an elevation angle and a rotation angle around a main optical axis, and the three parameters are predicted, so that 3D rotation prediction of the object is indirectlyrealized. The invention provides a universal frame for object 6D posture prediction, 2D target detection and 6D posture prediction can be carried out in one RGB image at the same time, and the methodhas good robustness for the conditions of complex illumination conditions, disordered placement, mutual shielding between objects and the like.

Description

technical field [0001] The invention belongs to the field of object target detection and pose prediction, and specifically relates to a 6D pose prediction method for objects based on RGB images and coordinate system transformation, emphasizing that only RGB information is used to perform 2D target detection and 6D pose prediction at the same time, and objects are mutually occluded and cluttered It has good robustness in situations such as placement and complex lighting conditions. Background technique [0002] At present, object target detection and its 6D pose (3D translation and 3D rotation) prediction are a research hotspot in the field of computer vision, and have important applications in augmented reality, robot operation, unmanned driving, etc. However, the 6D pose prediction problem is still a challenging problem due to the mutual occlusion, cluttered placement, and complex lighting conditions of objects in actual scenes. [0003] Currently, 6D pose prediction metho...

AI Technical Summary

Problems solved by technology

[0004] In view of the fact that the data set in the field of 6D posture prediction is difficult to label, and there are situations such as occlusion and complex lighting conditions in the actual test, the purpose of the present invention is to generate a data set based on OpenGL rendering 3D model, and use the Faster-R-CNN target detection network to locate the object to be predicted. The object pose is converted into the camera pose for prediction through coordinate system transformation, which solves the problems of the existing 6D pose prediction data set is small and the model is poorly adaptable to complex scenes.

Images