Template Creation Apparatus for Robust 3D Object Recognition

Solution Overview

Problem

Conventional template matching methods for 3D object recognition suffer from reduced accuracy due to noise and measurement errors affecting normal vector directions that form small angles with the camera optical axis, leading to incorrect feature quantization and matching.

Innovation Solution

A template creation apparatus that uses a central reference region near the axis and peripheral regions to quantify normal vectors, allowing for robust feature extraction and matching by setting a central reference region based on the angle between the normal vector and the axis, and allowing additional peripheral regions for quantization, thereby reducing the impact of noise and measurement errors.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional template matching uses normal vector directions for feature quantization, then object recognition can be performed, but recognition accuracy deteriorates due to noise and measurement errors affecting normal vectors that form small angles with the camera optical axis

Engineering Contradiction:

Improverecognition accuracyVSAvoidrobustness to noise and measurement errors

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent applies local quality by differentiating the treatment of normal vectors based on their orientation relative to the camera optical axis. Normal vectors forming small angles with the optical axis (within the predetermined angle threshold) are excluded from feature quantization, while other normal vectors are processed normally. This localized differentiation improves recognition accuracy by eliminating the specific source of error (small-angle normal vectors) without affecting the overall recognition process.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent changes the parameter of angle thresholding to filter out problematic normal vectors. By introducing a predetermined angle threshold relative to the camera optical axis, the system dynamically selects which normal vectors to include in feature quantization based on their angular parameter, thereby improving robustness against noise and measurement errors.

Inventive Principle:

Principle #35Parameter changes

2Loss of information

If all normal vectors are used for feature quantization, then comprehensive feature extraction is achieved, but noise and measurement errors in small-angle normal vectors reduce recognition accuracy

Engineering Contradiction:

Improvefeature extraction completenessVSAvoidrecognition accuracy

Core Design Contradiction:

Loss of informationVSMeasurement precision

Solution Approach 1:

The patent applies local quality by differentiating the treatment of normal vectors based on their orientation relative to the camera optical axis. Normal vectors forming small angles with the optical axis (within the predetermined angle threshold) are excluded from feature quantization, while other normal vectors are processed normally. This localized differentiation improves recognition accuracy by eliminating the specific source of error (small-angle normal vectors) without affecting the overall recognition process.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent converts the harmful effect of small-angle normal vectors (prone to noise and measurement errors) into a beneficial filtering criterion. By using the angle with the camera optical axis as a selection criterion, the system automatically excludes problematic vectors, transforming what was previously a source of error into a useful feature for improving recognition accuracy.

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

Data Source

PatentEP3460715B1Template creation apparatus, object recognition processing apparatus, template creation method, and program

Publication Date: 2023.08.23 OMRON CORP

AI summary

Provided is a technique related to an image processing apparatus that is unlikely to be affected by a change in the vector direction of a normal vector that forms a small angle with a camera optical axis. A template creation apparatus includes a three-dimensional data acquisition unit that acquires three-dimensional data of an object that is a recognition target, a normal vector calculation unit that calculates a normal vector at a feature point of an object viewed from a predetermined viewpoint that is set for the object, a normal vector quantization unit that quantizes a normal vector by mapping the normal vector to a reference region on a plane orthogonal to an axis that passes through the viewpoint, so as to acquire a quantized normal direction feature amount, the reference region including a central reference region corresponding to the vicinity of the axis and a reference region in the periphery of the central reference region, a template creation unit that creates a template for each viewpoint based on the quantized normal direction feature amount, and a template information output unit that outputs the template.