Road User Detection With Pose-Based Movement Profiling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for recognizing and classifying road users, particularly persons in road traffic, are not sufficiently precise and robust, especially in capturing varied modes of behavior and transport, posing challenges for reliable trajectory planning.
Innovation Solution
A method involving optical sensors, such as mono or stereo cameras, and neural networks to detect and classify persons, determine their pose and presence of transport means, and assign movement profiles based on detected features, using skeleton-like representations and sensor data fusion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detailed classification of persons and their transport means is implemented, then measurement precision and reliability of road user recognition is improved, but device complexity and data processing requirements increase
Solution Approach 1:
The classification process is segmented into multiple stages: first detecting the person, then determining pose, then identifying transport means, and finally assigning movement profiles. This staged approach breaks down the complex classification task into manageable segments that can be processed sequentially, improving precision without overwhelming system complexity at any single stage.
Solution Approach 2:
The system transitions from 2D image data to 3D pose information by detecting key points and generating skeleton-like representations. This dimensional transformation enables more accurate classification of persons and their transport means by adding spatial depth information to the classification process.
2Measurement precision
If comprehensive data acquisition and labeling for detailed classification is performed, then classification accuracy is improved, but loss of time and data processing effort increase
Solution Approach 1:
Movement profiles and their associated parameters are predetermined and stored in advance. Once a person and their transport means are identified, the system can immediately retrieve and assign the appropriate pre-defined movement profile, avoiding time-consuming real-time analysis of movement patterns and significantly reducing processing time.
Solution Approach 2:
The system uses skeleton-like representations as simplified copies of actual person poses. These abstracted skeletal models capture essential pose information without requiring full detailed analysis of every image pixel, enabling efficient classification while maintaining accuracy.
3Measurement precision
If skeleton-like representations and pose detection are used, then measurement precision of person characterization is improved, but device complexity and computational requirements increase
Solution Approach 1:
The system extracts only the essential key points needed for pose determination from the full image data, rather than processing every pixel. By taking out only the critical skeletal key points, the system achieves accurate pose detection while significantly reducing computational complexity compared to full image analysis.
4Adaptability or versatility
If multiple classification categories for persons and transport means are implemented, then adaptability to various road user behaviors is improved, but difficulty of detecting and measuring increases
Solution Approach 1:
The system dynamically adapts its classification approach by first detecting the person, then determining their pose, and based on that pose, identifying the appropriate transport means category. This dynamic, adaptive process allows the system to handle diverse road user behaviors efficiently without requiring all classification algorithms to run simultaneously, reducing overall detection difficulty.
Data Source
AI summary
The present disclosure relates to a method and system for recognizing road users, in particular persons in road traffic The method includes the following steps:recording at least one ambient image or a sequence of ambient images an optical sensor of a vehicle,detecting a person in the ambient image,determining a pose of the detected person;determining a presence of a means of transport of the person; andassigning a movement profile to the person based on the determined means of transport of the person.