In-Cabin 3D Object Classification Using Reflectivity Maps

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current driver assistance systems lack the capability to accurately and seamlessly transition from autonomous to manual mode, requiring improved classification of objects and recognition of driver gestures within the vehicle interior for enhanced safety and user experience.

Innovation Solution

A method and device utilizing a 3D image capture unit and processing unit to generate a reflectivity map from 2D intensity and 3D depth images, enabling precise classification of objects and recognition of gestures by calculating reflectivity, particularly effective in the near-infrared range, using time-of-flight cameras and optical bandpass filters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If object classification is improved for better driver monitoring, then safety and gesture recognition are enhanced, but system complexity and computational requirements increase

Engineering Contradiction:
Improveobject classification accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the classification process by first identifying skin regions based on color information, then separately analyzing gesture patterns within those regions. This division allows complex gesture recognition to be broken down into manageable steps: skin detection → region extraction → gesture classification, reducing overall system complexity while maintaining high accuracy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system transitions from analyzing only intensity images to incorporating color space information (HSV color model) as an additional dimension. By adding color as a classification dimension, the system achieves more accurate skin detection and gesture recognition without significantly increasing hardware complexity, as color processing can be performed through software algorithms

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

2Measurement precision

If 3D depth information is utilized for better object classification, then classification accuracy is improved, but data processing requirements and computational load increase

Engineering Contradiction:
Improveobject classification accuracyVSAvoidcomputational load
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system applies local quality by focusing computational resources only on skin-detected regions rather than processing the entire image. Once skin regions are identified using color information, the system concentrates depth analysis and gesture recognition computations specifically within those localized areas, significantly reducing overall computational load while maintaining high classification accuracy

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system performs preliminary skin detection using color information before conducting detailed 3D depth analysis. This preliminary filtering step identifies only the relevant regions that require intensive processing, allowing the system to prepare and prioritize computational resources in advance, thereby reducing overall processing requirements

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

This solution provides accurate and reliable classification of objects and gesture recognition, enhancing the transition between autonomous and manual driving modes, improving safety and user experience through precise monitoring of the driver's state and interactions.

Implementation Method 1

Each of the 3D images includes at least one 2D intensity image and at least one 3D depth map of the scene

Methodology Applied
Scientific EffectTime of flight: Time of Flight

Implementation Method 2

The illumination unit preferably operates in the near infrared (NIR) range

Methodology Applied
Scientific EffectNear-infrared illumination: Infrared Radiation

Implementation Method 3

The 3D image capture unit can, for example, comprise an optical bandpass filter that only transmits the wavelength used by the illumination, thereby eliminating a large portion of the interfering background light

Methodology Applied
Scientific EffectOptical filtering: Filter (optical)

Data Source

PatentEP3934929B1Method for classifying objects within a motor vehicle
Publication Date: 2024.09.04 VALEO SCHALTER & SENSOREN GMBH
  • EP3934929B1 patent drawingFigure 1
  • EP3934929B1 patent drawingFigure 2
  • EP3934929B1 patent drawingFigure 3

AI summary

The invention relates to a method for classifying objects (2) within the interior (3) of a vehicle (1), having the following steps: providing at least one 3D image-capturing unit (4) which is designed to capture 3D images of a scene, providing at least one processing unit (5) which is coupled to the 3D image-capturing unit (4), said processing unit (5) being designed to receive 3D images from the 3D image-capturing unit (4); detecting 3D images of a scene, each of the 3D images comprising at least one 2D intensity image and at least one 3D depth map of the scene; calculating a reflectivity map from the at least one 2D intensity image and the at least one 3D depth map using the processing unit (5); and classifying at least one object (2) of the scene using the reflectivity of the object. The invention additionally relates to a device for classifying objects (2) within the interior (3) of the vehicle (1), comprising a 3D image-capturing unit (4) and a processing unit (5) coupled to the 3D image-capturing unit (4) for carrying out the method. The invention additionally relates to a driver assistance system comprising a device for carrying out the method and to a computer program product comprising commands which prompt a computer to carry out the method when the program is ran by said computer.