Vehicle Seat Image Comparison for Child Seat Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current seat occupancy classification systems in vehicles face challenges in reliably distinguishing between a child seat occupied by a large child and an unoccupied seat, due to limited training data and similarities in image features, leading to potential misclassification.

Innovation Solution

A computerized method and system that utilizes a pre-trained machine learning classification network, specifically a Siamese neural network, to process current and reference images of vehicle seats, determining the seat state by comparing embeddings and confidence scores to accurately identify whether a child seat is mounted, even with limited data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional machine learning algorithms are used with limited training data, then the system can operate with available data, but misclassification errors occur between child seats and unoccupied seats

Engineering Contradiction:
Improveclassification accuracyVSAvoidtraining data volume
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The system performs preliminary actions by capturing and storing reference images of unoccupied seats before actual classification tasks. These reference images are stored in a database and used during runtime comparison, allowing the system to achieve reliable classification without requiring extensive training data during the machine learning training phase

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system creates copies of seat images at different states (occupied and unoccupied) and stores them as reference images. During classification, the current seat image is compared against these stored reference images using image processing techniques, enabling accurate differentiation without needing large volumes of diverse training data

Inventive Principle:
Principle #26Copying

2Measurement precision

If image processing is used to detect child seats, then visual information can be analyzed, but misclassification occurs when large children resemble unoccupied seats

Engineering Contradiction:
Improveseat state detection accuracyVSAvoidimage processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the classification task into distinct phases: capturing reference images of unoccupied seats, processing current seat images, comparing the two sets of images, and making classification decisions. This segmentation allows each phase to be optimized independently, improving detection accuracy while managing complexity through modular processing steps

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system introduces reference images as an intermediary element between the raw current seat image and the final classification decision. By comparing the current image against stored reference images of known unoccupied seats, the system creates an intermediate comparison stage that enhances detection precision and reduces misclassification

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4303822A1Child seat detection for a seat occupancy classification system
Publication Date: 2024.01.10 APTIV TECHNOLOGIES AG
  • EP4303822A1 patent drawingFigure 1~2
  • EP4303822A1 patent drawingFigure 3
  • EP4303822A1 patent drawingFigure 4A~4D

AI summary

A computerized method of child seat detection for a seat occupancy classification system in a vehicle is presented. The method comprises receiving a current image showing a region of a vehicle seat currently captured inside of the vehicle and retrieving one or more reference images, wherein at least one reference image is a previously stored first reference image of the vehicle seat being unoccupied. The method further comprises determining a seat state of the vehicle seat by processing the current image and one or more reference images with a pre-trained machine learning classification network and determining, based on an output from the machine learning classification network, the seat state indicating at least whether the vehicle seat in the current image comprises a mounted child seat. Finally, the method forwards the seat state to the seat occupancy classification system. A corresponding system, vehicle, and computer program are also presented.