Rail train perception method based on occupancy grid world model

Through the rail train perception method based on the raster world model, combined with the information of current and historical scene image frames, three-dimensional scene features are generated and predictions are made, the problem of low accuracy and reliability of perception results in the prior art is solved, and the safety of rail trains is significantly improved.

CN119559614BActive Publication Date: 2025-05-09BEIJING AINIBABY HEALTH MANAGEMENT CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202510115987.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-01-24
Publication Date
2025-05-09
Estimated Expiration
2045-01-24

AI Technical Summary

Technical Problem

The existing train perception methods only make judgments based on the current frame, lack prediction of future scenarios, resulting in low accuracy and reliability of perception results, affecting the safety of rail trains.

Method used

The track train perception method based on the occupancy raster world model is adopted. By obtaining the current and historical scene image frames, voxel features and category features are extracted, stitching and occupancy raster scene encoding are performed, three-dimensional scene features are generated, and the occupancy raster world model is predicted based on this to determine the scene perception results.

Benefits of technology

The accuracy and reliability of the perception results of rail trains are improved, and potential dangers are identified in advance by predicting future scenarios, which improves the safety of trains.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119559614B_ABST
    Figure CN119559614B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of visual perception technology, and provides a rail train perception method based on an occupation grid world model, including: obtaining a scene image frame to be perceived, including a current scene image frame and a historical scene image frame; extracting voxel features of the scene image frame to be perceived; splicing the category features and voxel features of the scene image frame to be perceived to obtain a splicing feature, performing occupation grid scene coding on the splicing feature to obtain a three-dimensional scene feature; performing occupation grid world model prediction based on the three-dimensional scene feature to determine the scene perception result. The method provided by the present invention extracts voxel features from the scene image frame to be perceived including the historical scene image frame; splices the category features and voxel features, performs occupation grid scene coding on the splicing feature to obtain a compact three-dimensional scene feature; performs occupation grid world model prediction based on the three-dimensional scene feature, thereby achieving accurate and reliable scene perception, and thus improving the safety of rail trains.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Three-dimensional panoramic segmentation method and device based on multi-view camera

    CN116912488A