Vehicle Occupant Counting Using Depth Image Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional vehicle occupant counting methods lack accuracy in determining the number of occupants in a vehicle, particularly when using depth information and combining face detection with depth data for accurate counting.

Innovation Solution

A vehicle occupant count monitoring system that employs a three-dimensional time of flight camera to capture depth images, combining facial recognition and upper half-body recognition using learning dictionaries and deep learning techniques, and determines the number of occupants by associating three-dimensional coordinates to improve counting accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional face recognition methods are used for occupant counting, then the system is simple to implement, but the counting accuracy deteriorates when occupants are positioned closely or in complex scenarios

Engineering Contradiction:
Improveoccupant counting accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent transitions from two-dimensional face recognition to three-dimensional depth image analysis. By capturing depth information and converting it to three-dimensional coordinates, the system can distinguish occupants in complex scenarios and prevent repetitive counting, thereby improving counting accuracy without excessively increasing system complexity

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent segments the depth image into multiple recognition blocks and processes each block independently to identify faces and upper bodies. This segmentation approach allows the system to handle complex scenarios with multiple occupants by analyzing each region separately and combining results, improving accuracy while maintaining manageable system complexity

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If only face detection is used for occupant counting, then the processing is simple, but the accuracy deteriorates when faces are not visible or occupants are positioned closely

Engineering Contradiction:
Improveoccupant detection accuracyVSAvoiddetection complexity
Core Design Contradiction:
Measurement precisionVSDifficulty of detecting and measuring

Solution Approach 1:

The patent combines face detection results with upper body detection results from depth images. By merging these two detection approaches and associating them through three-dimensional coordinate matching, the system achieves more robust occupant counting that works even when faces are not visible, improving detection accuracy while managing the complexity through coordinated processing

Inventive Principle:
Principle #5Merging (Combining)

3Measurement precision

If depth images are captured and processed for occupant counting, then the counting accuracy improves, but the processing time and computational load increase

Engineering Contradiction:
Improveoccupant counting accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary processing by converting depth images to three-dimensional coordinates and segmenting the image into recognition blocks before detailed analysis. This preliminary structuring of data facilitates faster subsequent processing and matching operations, reducing overall processing time while maintaining the accuracy benefits of depth image analysis

Inventive Principle:
Principle #10Preliminary action

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

The system enhances the accuracy of occupant counting by effectively combining facial and upper half-body recognition, preventing repetitive counting and improving the detection of occupants even in complex scenarios, such as those with multiple occupants positioned closely.

Implementation Method 1

a three-dimensional time of flight camera captures a depth image

Methodology Applied
Scientific EffectTime of flight: Time of Flight

Data Source

PatentEP3584746B1Vehicle occupant count monitoring system
Publication Date: 2023.06.28 YAZAKI CORP
  • EP3584746B1 patent drawingFigure 1
  • EP3584746B1 patent drawingFigure 2
  • EP3584746B1 patent drawingFigure 3

AI summary

A vehicle occupant count monitoring system (1) includes: an imaging unit (3) that captures a depth image (G) including a distance to an occupant (OB) in a cabin of a vehicle; a facial recognizing unit (83) that recognizes a face of the occupant (OB) from the depth image (G) captured by the imaging unit (3); an upper half body recognizing unit (84) that recognizes an upper half body of the occupant (OB) from the depth image (G) captured by the imaging unit (3); and a counting unit (85) that counts the number of occupants (OB) by combining a recognition result of the face using the facial recognizing unit (83) and a recognition result of the upper half body using the upper half body recognizing unit (84).