Crowdsourced Indoor Semantic Map Updating via Key Frame Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing indoor semantic map systems face challenges in efficiently updating semantic information, leading to high costs and inefficiencies in maintaining accurate maps, especially in dynamic indoor environments, as current methods are labor-intensive and not suitable for real-time updates.
Innovation Solution
A method and system for updating indoor semantic maps using crowdsourced short videos, where mobile terminals capture and preprocess videos to extract key frames and text sequences, utilizing a Markov random field to recognize accurate text and position it on the map, thereby updating the map with minimal resource expenditure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire indoor semantic map is periodically regenerated to update semantic information, then the map accuracy is improved, but the time consumption and resource waste increase significantly
Solution Approach 1:
The patent extracts only the changed portions of the indoor semantic map for updating, rather than regenerating the entire map. This is achieved by comparing new semantic information with existing map data, identifying differences, and updating only those specific regions or entities that have changed, thereby significantly reducing time consumption and resource usage while maintaining map accuracy
Solution Approach 2:
The patent segments the indoor semantic map into multiple independent entities or regions, allowing selective updating of individual segments rather than the entire map. This segmentation enables parallel processing and reduces the overall update time by focusing computational resources only on the changed segments
2Measurement precision
If the entire indoor semantic map is periodically regenerated to update semantic information, then the map accuracy is improved, but the resource consumption increases significantly
Solution Approach 1:
The patent extracts and processes only the changed portions of the indoor semantic map, avoiding unnecessary processing of unchanged regions. This extraction approach reduces computational resource consumption, energy usage, and processing time while maintaining the accuracy of updated map data
Solution Approach 2:
The patent applies partial action by performing updates only where necessary (in changed regions) rather than applying full processing to the entire map. This partial updating strategy reduces resource consumption while achieving the required map accuracy through targeted processing of only the necessary portions
3Productivity
If traditional text recognition methods are used to extract semantic information, then the processing speed is maintained, but the text recognition accuracy deteriorates
Solution Approach 1:
The patent merges multiple text recognition technologies and processing methods into a unified system that combines the strengths of different approaches. This integration includes combining traditional OCR with deep learning-based recognition methods, allowing the system to maintain high processing speed while achieving improved text recognition accuracy through the synergistic effect of multiple techniques
Data Source
AI summary
The present invention discloses an indoor semantic map updating method and system based on semantic information extraction. The method includes: issuing a crowdsourcing task to all mobile terminals; waiting for any mobile terminal to execute the crowdsourcing task, and receiving a task result thereof; preprocessing the task result to obtain a common key frame sequence; extracting an accurate text sequence from the common key frame sequence; and updating an indoor semantic map according to the common key frame sequence and the accurate text sequence. The present invention can encourage the mobile terminal to execute the crowdsourcing task, and update the indoor semantic map and the text semantic information at a lower cost.