Object Recognition Device Using Reliability-Based Grouping

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional pedestrian recognition devices often misrecognize multiple adjacent moving objects as a single group, leading to inaccurate object recognition.

Innovation Solution

An object recognition device that includes an object detection unit, a coordinate specifying unit, a reliability setting unit, and a grouping unit, which sets object presence reliability and groups detection points based on this reliability to avoid mistaken grouping, using a distribution map and point-to-point distance measurement to improve recognition accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If adjacent reflecting objects are grouped into a single group based on two-dimensional coordinate system, then the device complexity is reduced, but the measurement precision of individual objects deteriorates

Engineering Contradiction:
Improvegrouping complexityVSAvoidobject recognition accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by introducing reliability information specific to each detection point and the space between detection points. Instead of uniform grouping criteria, the system evaluates local reliability values that vary across different spatial locations, allowing adjacent objects to be distinguished based on their local reliability characteristics while maintaining efficient grouping processing.

Inventive Principle:
Principle #3Local quality

2Loss of information

If multiple adjacent moving objects are grouped as a single group, then the loss of information is reduced, but the measurement precision of individual objects deteriorates

Engineering Contradiction:
Improvedetection information completenessVSAvoidindividual object recognition accuracy
Core Design Contradiction:
Loss of informationVSMeasurement precision

Solution Approach 1:

The patent segments the grouping process by introducing reliability-based criteria that divide detection points into distinct groups. Instead of treating all adjacent objects uniformly, the system segments them based on reliability thresholds and reliability value comparisons, enabling individual object identification while maintaining complete detection information through the distribution map structure.

Inventive Principle:
Principle #1Segmentation

3Ease of operation

If object presence reliability is set uniformly for all detection points, then the ease of operation is improved, but the measurement precision of object boundaries deteriorates

Engineering Contradiction:
Improvereliability setting simplicityVSAvoidobject boundary detection accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The patent implements local quality by setting reliability values that vary according to spatial position and detection point characteristics. The reliability setting unit assigns different reliability values to different detection points based on local conditions, and the distribution map creation unit propagates these local reliability values to create a spatially varying reliability field that accurately reflects object boundaries while maintaining systematic operation.

Inventive Principle:
Principle #3Local quality

Data Source

PatentUS10614322B2Object recognition device
Publication Date: 2020.04.07 TOYOTA JIDOSHA KK
  • US10614322B2 patent drawing
  • US10614322B2 patent drawing
  • US10614322B2 patent drawing

AI summary

In an object recognition device, a reliability setting unit sets object presence reliability indicating the reliability of whether or not an object is present with respect to the surrounding area of a host vehicle. Then, a grouping unit groups the reflection points based on the object presence reliability in the area between the reflection points. For this reason, for example, even though two objects are adjacent to each other, in a case where the object presence reliability in the area between the reflection points of one object and the reflection points of the other object is low, the reflection points of the two objects can be separately grouped without being grouped into a single group. As a result, since it is possible to avoid mistakenly recognizing a plurality of objects in a single object, recognition accuracy of the object can be improved.