Panoramic redundant viewpoint filtering method based on man-machine isomorphism

Through the human-computer homogeneous panoramic redundant viewpoint filtering method, combined with progressive self-step learning and multimodal information, the problem of insufficient accuracy and generalization of redundant viewpoint filtering in panoramic images is solved, and more efficient Viewpoint classification enhances the user experience of virtual reality technology.

CN119068340AActive Publication Date: 2024-12-03CHINA UNIV OF PETROLEUM (EAST CHINA)

Patent Information

Application Number
CN202411186645.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-28
Publication Date
2024-12-03
Estimated Expiration
2044-08-28

Smart Images

  • Figure CN119068340A_ABST
    Figure CN119068340A_ABST
Patent Text Reader

Abstract

The invention discloses a panoramic redundant viewpoint filtering method based on man-machine isomorphism, and the method comprises the steps: building an initial credible and low credible sample set through a full-automatic multi-threshold division strategy, building a deep neural network model, considering the time sequence characteristics of different viewpoint track sequences, and carrying out the full-automatic multi-threshold division strategy through learning the dependency relationship and mode between the sequences. And the viewpoints of the panoramic image are accurately and reliably classified. Through a self-paced learning strategy, more credible samples are gradually mined from easy to difficult to add into a training sample set, and in each progressive learning stage, the updated training sample set is utilized to guide the retraining of the model. Along with the deepening of training times, the model continuously learns and captures viewpoint features in more complex scenes, and finally realizes accurate classification of panoramic viewpoint data. The method is wide in application scene, and redundant viewpoint filtering plays an important role in the field closely related to panoramic images. The viewpoint data is of great significance to research and understand how the user explores the panoramic scene.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Method and apparatus for determining and analyzing a location of visual interest

    CN101466305A

  • Multi-pedestrian detection and attitude estimation method under view angle of ultra-low-altitude unmanned aerial vehicle

    CN116152682A

  • Gaze point acquisition method and apparatus, electronic device and readable storage medium

    WO2023051215A1

Cited By

  • Two-person interactive watching behavior recognition and mode conversion positioning method

    CN121708647A