Pedestrian Detection Ground Contact Correction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing human detection systems inaccurately identify the ground contact position of pedestrians when a covering object is present, leading to incorrect distance calculations from the vehicle.

Innovation Solution

A human detection device and method that utilizes a front-view camera, image processing units, and dictionaries to detect humans, identify the ground contact position, and correct it based on the head-to-height ratio and feature extraction, ensuring accurate identification even with covering objects.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a covering object is detected as part of the pedestrian candidate, then the detection coverage is improved, but the ground contact position identification accuracy deteriorates

Engineering Contradiction:
Improvedetection coverageVSAvoidground contact position identification accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The detection system segments the pedestrian candidate into multiple parts: head portion, body portion, and potential covering objects. By extracting the head portion specifically and using it as a reference, the system can distinguish between the actual pedestrian and covering objects, thereby maintaining detection coverage while improving ground contact position identification accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The head portion extraction serves as an intermediary reference point. Since the head portion is less likely to be confused with covering objects and has a more stable position, it acts as a mediator to correctly identify the pedestrian's ground contact position even when covering objects are present in the detection area.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Device complexity

If the lower end portion of the detected human is used to identify ground contact position, then the identification process is simplified, but the distance calculation accuracy deteriorates when covering objects are present

Engineering Contradiction:
Improveidentification process complexityVSAvoiddistance calculation accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The system extracts the head portion from the detected human candidate and uses it as a reference point for identification. This extraction approach maintains relative simplicity in the identification process while significantly improving distance calculation accuracy, as the head portion is less likely to be misidentified as a covering object compared to the lower end portion.

Inventive Principle:
Principle #2Taking out (Extraction)

3Quantity of substance

If the detection system includes all objects in the detection area, then the detection completeness is improved, but the false positive rate increases when covering objects are present

Engineering Contradiction:
Improvedetection completenessVSAvoidfalse positive rate
Core Design Contradiction:
Quantity of substanceVSReliability

Solution Approach 1:

The system applies different quality criteria to different parts of the detection area. By focusing on the head portion characteristics (such as shape, size, and position relative to the body) and using it as a reference, the system can distinguish true pedestrians from covering objects, thereby maintaining detection completeness while reducing false positives.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3410420B1Person detection device, person detection method, and person detection program
Publication Date: 2024.03.06 JVC KENWOOD CORP
  • EP3410420B1 patent drawingFigure 1
  • EP3410420B1 patent drawingFigure 2
  • EP3410420B1 patent drawingFigure 3

AI summary

A human detection device (20) comprises an image acquiring unit (21) configured to acquire an image captured by an imaging device (10), a human detecting unit (22) configured to detect a human from the acquired image, a ground contact position identifying unit (23) configured to identify a ground contact position of the human on the basis of a lower end portion of the detected human, a feature portion extracting unit (24) configured to extract a feature portion of the detected human, a ratio calculating unit (25) configured to calculate a ratio of the feature portion on the basis of a vertical size of the detected human and a vertical size of the extracted feature portion and a correction determining unit (26) configured to determine whether the identified ground contact position is to be corrected on the basis of the identified ground contact position or the calculated ratio, in which the ground contact position identifying unit (23) corrects the identified ground contact position on the basis of the vertical size and the ratio of the feature portion.