Pedestrian Detection Using Lower Body Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing exterior environment recognition systems face challenges in accurately detecting and tracking pedestrians at night due to the difficulty in identifying uniform movement and the upper body area, which affects smooth operation of brake and cruise control systems.
Innovation Solution
An exterior environment recognition device that groups image portions based on relative relationships and divides them to determine person candidates using shape, size, symmetry, and luminance, with a focus on the lower body area for improved accuracy and efficiency in pedestrian detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the upper half of the body area is used for pedestrian detection, then the detection can be performed in daylight conditions, but the detection fails at night when headlights only illuminate the lower half of the body
Solution Approach 1:
The patent divides the pedestrian detection task into two segments: detecting the lower half of the body (illuminated by headlights) and inferring the upper half position based on geometric relationships. This segmentation allows reliable detection at night by focusing only on the illuminated portion while using spatial reasoning to identify the complete pedestrian.
2Measurement precision
If the whole body area is specified, then accurate pedestrian recognition can be achieved in daylight, but the specification becomes impossible at night when only the lower half is visible
Solution Approach 1:
Instead of trying to detect the entire body and infer the lower half, the patent inverts the approach by detecting the lower half first (which is visible) and then inferring the position and extent of the upper half based on geometric relationships and typical human proportions. This inversion makes detection feasible under nighttime conditions.
3Speed
If traditional tracking methods are used for pedestrians, then movement can be monitored, but the tracking fails when the upper body area cannot be specified due to lighting conditions
Solution Approach 1:
The patent performs preliminary detection of the lower body area and uses this information to predict and track the complete pedestrian position. By establishing the detection framework based on the visible lower half first, the system can then track movement continuously even when the upper body remains invisible, maintaining both speed and reliability in tracking.
Data Source
AI summary
There are provided an environment recognition device and an environment recognition method. An exterior environment recognition device obtains an image in a detection area, generates a block group by grouping, based on a first relative relationship between blocks, multiple blocks in an area extending from a plane corresponding to a road surface to a predetermined height in the obtained image, divides the block group into two in a horizontal direction of the image, and determines, based on a second relative relationship between two divided block groups, whether the block group is a first person candidate which is a candidate of a person.