Method and device for three-dimensional object recognition segmentation and pose measurement based on cad model

Abstract

The invention discloses a CAD model-based three-dimensional target recognition and partitioning and position-pose measurement method and device. The method comprises the steps that two-dimensional target recognition is conducted to extract a target rectangular frame; a CNN network structure of a Faster RCNN network is reused according to the target rectangular frame; component constraint is conducted on feature points of the target rectangular frame, the constrained feature points are input to the reused CNN network structure, and the feature points with the robustness are extracted; a matching model suitable for a target is acquired; a projection matrix of a target relative space model is recovered, and the translation and rotation states are acquired; and a model obtained after the matching model is correspondingly translated and rotated is projected into a target area according to the projection matrix of the target relative space model and the translation and rotation states to complete target partitioning. According to the method, comprehensive understanding on the target is built, a basis is built for scene understanding, and the advantages of being high in responding speed and precision and good in robustness are achieved.

Description

technical field [0001] The present invention relates to the technical field of image recognition, in particular to a CAD (Computer Aided Drafting, computer-aided design) model-based method and device for three-dimensional target recognition segmentation and pose measurement. Background technique [0002] At present, with the development of computer vision theory and the improvement of hardware equipment level, a large amount of 3D target data has been generated in the field of recognition. Since 3D target data contains complete information of the target, it can make up for the inherent defects of 2D target recognition, so 3D target recognition Technology is getting more and more attention. Its main task is to identify the type of target in the image, and give the relative position and pose of the target in the scene. [0003] In related technologies, three-dimensional target recognition and attitude measurement mainly include two methods: a contact method and a non-contact ...

AI Technical Summary

Problems solved by technology

However, the calculation of continuous variables in tasks such as attitude and segmentation has not yet achieved sufficient accuracy.

In addition, in many cases without ranging equipment, single-camera 3D reconstruction recognition and pose measurement have not yet achieved sufficient accuracy and efficiency

On the other hand, for scene cognition, the 3D semantic segmentation of the target is also a very important task. In the past, the segmentation method using neural network needs to label each pixel during training, which is very labor-intensive and resource-intensive. Other methods of level labeling cannot achieve a good enough segmentation effect

More importantly, the 3D scene understanding of the target only stays on dividing it into different tasks such as recognition, segmentation, pose measurement, etc., and cannot establish an overall understanding of the target

Images