Pedestrian Motion Prediction Using Posture Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pedestrian motion prediction systems are inaccurate in determining the possibility of a pedestrian crossing the road until the pedestrian begins to walk and have reduced accuracy for individuals with small steps.
Innovation Solution
A pedestrian motion predicting device that compares the detected shape of a pedestrian with a previously prepared shape of a pedestrian likely to cross, using a combination of detected shapes over time, analyzing shape periodicity, and determining speed continuity to predict the likelihood of a pedestrian crossing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If pedestrian behavior is predicted by the ratio of the opening degree W of the leg portion to the height H, then the prediction method is simple, but the determination accuracy is deteriorated and prediction cannot be made until the pedestrian begins to walk
Solution Approach 1:
The patent changes the prediction parameters from simple geometric ratio (W/H) to multiple posture parameters including upper body inclination angle, arm swing angle, and leg opening degree. This parameter transformation enables earlier detection of crossing intention by analyzing posture changes before the pedestrian actually steps onto the road, thereby improving determination accuracy while maintaining reasonable system complexity
Solution Approach 2:
The system performs preliminary analysis of posture parameters to predict crossing intention before the pedestrian begins actual movement. By detecting changes in upper body inclination, arm swing, and leg positioning that precede the crossing action, the system can issue warnings in advance, resolving the contradiction between early prediction capability and determination accuracy
2Device complexity
If pedestrian behavior is predicted by the ratio of the opening degree W of the leg portion to the height H, then the prediction method is simple, but prediction cannot be made until the pedestrian takes the first step of rush out
Solution Approach 1:
The system analyzes multiple posture parameters (upper body inclination angle, arm swing angle, leg opening degree) to detect crossing intention before the pedestrian takes the first step. This preliminary detection capability provides sufficient warning time for drivers to react, resolving the timing loss issue while keeping the prediction method relatively simple
Solution Approach 2:
The system continuously monitors posture parameters and compares them against threshold values to detect crossing intention. This feedback mechanism enables real-time prediction by identifying when posture changes indicate impending crossing behavior, allowing the system to provide timely warnings before the pedestrian actually crosses
3Device complexity
If pedestrian behavior is predicted by the ratio of the opening degree W of the leg portion to the height H, then the prediction method is simple, but determination accuracy is deteriorated in a pedestrian who rushes out with a small step
Solution Approach 1:
The patent transitions from using a single geometric ratio parameter (W/H) to a multi-parameter posture analysis system that includes upper body inclination angle, arm swing angle, and leg opening degree. This parameter expansion enables accurate detection of crossing intention even when the leg opening is minimal, as the system can detect compensating posture changes in other body parts that indicate crossing intent
Solution Approach 2:
The prediction system combines multiple posture parameters into a composite assessment model. By integrating information from upper body inclination, arm swing, and leg positioning, the system creates a more robust prediction capability that works effectively for all pedestrian types including those with small steps, overcoming the limitations of single-parameter methods
Data Source
AI summary
A subject is to provide a pedestrian motion predicting device capable of accurately predicting a possibility of a rush out before a pedestrian actually begins to rush out. According to the embodiments, the pedestrian is detected from input image data, a portion in which the detected pedestrian is imaged is cut out from the image data, a shape of the pedestrian imaged in the cut-out partial image data is classified by collating the shape with a learning-finished identifier group or a pedestrian recognition template group, and the rush out of the pedestrian is predicted based on a result of the acquired classification.