Nonrigid registration method based on implicit vector space

Abstract

The invention discloses a nonrigid registration method based on an implicit vector space, which comprises the following steps: extracting a source model and a target model; carrying out whole registration based on the implicit vector space; establishing the optimal equation of the whole registration; solving the optimal solution of the optimal equation, and finishing the whole registration of the source model and the target model to obtain an rigid transformation process; carrying out local registration based on the implicit vector space; adopting incremental free transformation, and establishing the optimal equation of the local registration based on the implicit vector space; solving the closed form solution of the local registration; and establishing one dense correspondence. The nonrigid registration method is capable of processing an open local model and a closed model simultaneously, thus expanding the application scope of the algorithm and fastening the solving speed. While a natural shape recovery and gesture transformation process is implemented, the dense correspondence on the frame vertexes can be established, thus obtaining the nonrigid registration on dimensional dynamic discrete frame models.

Description

technical field [0001] The invention belongs to the technical field of computer graphics and man-machine interface, and in particular relates to a non-rigid registration method based on an implicit vector space. Background technique [0002] With the development of reverse engineering technology and the improvement of chip manufacturing level, acquisition devices such as depth cameras record the shape and attitude changes of dynamic objects (including rigid and non-rigid transformations) through high-frequency sampling, and output real-time A time-varying frame model. This provides a new technical approach for obtaining dynamic objects in the physical world, and users can realistically reconstruct digital 3D dynamic models without tedious manual editing. Therefore, the level and efficiency of reconstructing the three-dimensional dynamic model can be greatly improved, making the reconstruction process more intuitive and the effect more vivid. However, since the recorded fra...

AI Technical Summary

Problems solved by technology

The disadvantage of this is that for the spatial position near the model boundary, it is difficult to determine the sign of the distance value at this position, and once the sign is determined incorrectly, the correct non-rigid registration result cannot be obtained

Therefore, the traditional non-rigid registration is not applicable widely, and it is difficult to deal with open local models. In fact, due to the occlusion problem in the acquisition process, the acquired models often have holes and are open local models.

In addition, the traditional non-rigid registration method calculates the non-rigid transformation by iteratively solving the optimization equation, and the iterative process is complex and time-consuming

Images