Pedestrian Risk Estimation Using Ground Area Semantic Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional pedestrian detection systems are limited by installation restrictions, require manual area setting, and struggle to accurately determine a pedestrian's location on various road surfaces, leading to inefficiencies and increased costs.
Innovation Solution
A system and method that utilizes semantic segmentation and ground area mapping to automatically detect and estimate pedestrian risk by generating bounding boxes, segmentation maps, and ground area maps, enabling flexible installation and continuous monitoring without manual area setting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If CCTV is installed to make crosswalk occupy large area in image, then pedestrian detection accuracy is improved, but installation location is physically restricted and requires specific photographing angles
Solution Approach 1:
The patent transitions from 2D image-based detection to 3D spatial understanding by generating ground area maps that represent the physical layout of roads, crosswalks, and sidewalks. This dimensional transformation allows the system to determine pedestrian location accuracy without requiring specific camera installation angles or positions, thereby resolving the contradiction between detection accuracy and installation flexibility.
2Measurement precision
If manual setting of interested area is performed for each CCTV, then detection precision is improved, but operational complexity and time consumption increase
Solution Approach 1:
The system performs self-service by automatically generating ground area maps and identifying road, crosswalk, and sidewalk regions without requiring manual configuration. The semantic segmentation algorithm autonomously divides the image into meaningful ground areas, eliminating the need for operators to manually set interested areas for each CCTV camera while maintaining high location determination accuracy.
3Device complexity
If conventional detection algorithms are used, then implementation simplicity is improved, but ability to determine pedestrian location on various road surfaces deteriorates
Solution Approach 1:
The patent applies segmentation by dividing the image into distinct ground areas (road, crosswalk, sidewalk) using semantic segmentation algorithms. This segmentation approach enables the system to accurately classify and determine which specific road surface a pedestrian is located on, significantly improving location determination precision while adding manageable computational complexity.
Data Source
AI summary
A method of estimating a pedestrian risk situation includes receiving an image captured by a CCTV camera that is installed at a predetermined location, generating a bounding box of a pedestrian within the image, generating a segmentation map by performing semantic segmentation on the image, generating a ground area map by identifying a ground area in the segmentation map, and estimating semantic location information of the pedestrian based on the bounding box and the ground area map.