Image Processing and Modeling of Virtual Scenes Based on Nonlinear ICA

The intelligent role in virtual environment can be located when using vision to get the target image, but there is a source separation problem in case of unknown complex natural signals. In order to resolve the source separation problem, this paper proposed the target three-dimensional modeling based on the visual image processing of virtual robot; applied the method based on the nonlinear Principle component analysis indicator and style of learning rules of ICA to separate the signal image, and put forward the mathematical model and rule, then introduced the feature extraction of the signal image and took the natural Litchi as an example to descript the effectiveness. Finally, the applications of ICA for Virtual Scene Modeling were indicated.