Pedestrian Detection via Position-Appearance Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle external environment recognition systems face challenges in quickly and accurately identifying pedestrians within a detection area, particularly due to the sudden appearance of pedestrians from various directions and the need to avoid erroneous recognition, which can lead to malfunction and delay in collision avoidance.
Innovation Solution
A vehicle external environment recognition device that utilizes a combination of three-dimensional position derivation, object identification, positional point derivation, appearance point derivation, and pedestrian identification modules to rapidly and accurately identify pedestrians by integrating positional and appearance information, including contour and color tone analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If contour geometric model matching is used to identify pedestrians, then identification accuracy is improved, but identification time increases
Solution Approach 1:
The system performs preliminary actions by deriving three-dimensional positions of subject parts and grouping them into objects before conducting detailed pedestrian identification. This preliminary object identification and positioning prepares data structures and reduces the search space for subsequent pedestrian detection, enabling faster processing without sacrificing accuracy
Solution Approach 2:
The pedestrian identification process is segmented into multiple independent modules: three-dimensional position derivation, object identification based on grouped subject parts, and pedestrian identification based on positional information. This segmentation allows each module to process specific aspects independently and in parallel, reducing overall identification time while maintaining comprehensive accuracy
2Reliability
If collision avoidance control is implemented, then safety is improved, but risk of erroneous recognition increases
Solution Approach 1:
The system employs feedback mechanisms where the three-dimensional position information and object identification results are continuously refined and verified before final pedestrian identification. The positional data from multiple subject parts is aggregated and cross-checked, providing feedback loops that validate detections and reduce false positives while maintaining high sensitivity for actual pedestrians
3Area of stationary object
If multiple imaging devices are used to cover all directions, then detection coverage is improved, but system complexity increases
Solution Approach 1:
The imaging devices are designed with multi-functionality to perform multiple tasks simultaneously: capturing images for pedestrian detection, deriving three-dimensional positions of subject parts, and providing data for object identification. This universal usage of the same hardware for multiple processing stages reduces the need for additional dedicated devices while maintaining comprehensive detection coverage
Data Source
AI summary
A vehicle exterior environment recognition device includes a three-dimensional position deriving module that derives three-dimensional positions of subject parts in a real space based on an image obtained by imaging a detection area, an object identifying module that groups the subject parts of which differences in the three-dimensional positions are within a predetermined range to identify an object, a positional point deriving module that derives a positional point based on positional information of the object, an appearance point deriving module that derives an appearance point based on appearance information of the object, and a pedestrian identifying module that identifies the object to be a pedestrian when a pedestrian point that is obtained by at least adding the positional point to the appearance point equals to or greater than a predetermined threshold.