Pedestrian Behavior Prediction via Segmented Feature Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pedestrian behavior prediction systems face challenges in efficiently fusing multiple input features and have limitations in processing speed and the number of features they can handle, which hinders their application in safety-critical areas like school zones and smart safety road systems.
Innovation Solution
A system and method that utilize an artificial intelligence model to efficiently fuse multiple visual and non-visual input features by grouping and processing them through separated modules, enabling the prediction of pedestrian behavior and identifying risk-vulnerable situations within a road environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If LSTM-based models are used for pedestrian behavior prediction, then prediction capability is achieved, but processing speed becomes slow
Solution Approach 1:
The patent divides the feature processing into separate modules: visual feature extraction module, non-visual feature extraction module, and fusion module. This segmentation allows each module to process specific features independently, improving overall processing speed while maintaining prediction capability through coordinated operation of specialized modules.
2Measurement precision
If more input features are fused to improve prediction accuracy, then prediction accuracy increases, but device complexity increases
Solution Approach 1:
The patent segments feature processing into dedicated visual and non-visual extraction modules, allowing systematic integration of multiple features (visual features like appearance and contextual features like spatial relationships) without overwhelming complexity. Each module handles specific feature types, making the overall system manageable while achieving high accuracy through comprehensive feature fusion.
3Adaptability or versatility
If existing researches are applied to school zones or smart safety road systems, then pedestrian safety application is enabled, but additional consideration for safety requirements is insufficient
Solution Approach 1:
The patent applies different processing strategies to different feature types: visual features are processed through convolutional neural networks for detailed pattern recognition, while non-visual features are processed through recurrent neural networks for temporal pattern analysis. This local quality approach allows the system to address specific safety requirements of different environments (school zones, crosswalks) with appropriate processing depth and focus for each feature type.
Data Source
AI summary
A method of generating pedestrian behavior prediction information includes setting image information from present to a certain time in the past as the target observation image, extract multiple visual input feature information and non-visual input feature information from the target observation image, grouping the multiple visual input feature information and non-visual input feature information, and inputting the grouped information to separated processing modules and generating pedestrian behavior prediction information by concatenating output results of the processing modules.