Object Recognition Using Filter Information for Markerless AR

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional Augmented Reality (AR) recognition techniques face challenges in achieving high recognition rates using markerless-based methods, struggling to detect objects of interest from images with multiple objects and often incorrectly identifying unwanted objects, leading to inefficient processor usage and time consumption.

Innovation Solution

An object recognition apparatus and method utilizing filter information, which includes an object information acquiring unit, a filter information input unit, a controller, and storage units to acquire, process, and output image information, allowing users to input filter information to recognize objects of interest by comparing image data with reference characteristic information.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If markerless-based recognition techniques (edge and boundary detection) are used to recognize objects, then the system can operate without pre-designated markers, but the recognition rate is low and unreliable

Engineering Contradiction:

Improvemarkerless operation capabilityVSAvoidrecognition rate

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent segments the object recognition process into multiple stages: initial object detection from image data, extraction of characteristic information, comparison with stored reference data, and verification. This multi-stage segmentation allows the system to maintain markerless operation while improving recognition reliability through systematic processing and validation at each stage.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary actions by pre-storing characteristic information of objects in a database before recognition occurs. During operation, the system retrieves and compares this pre-prepared reference data with detected object characteristics, enabling reliable recognition without markers by leveraging pre-processed reference information.

Inventive Principle:

Principle #10Preliminary action

2Productivity

If conventional recognition techniques process images with multiple objects, then all objects in the image are analyzed, but objects of interest cannot be easily detected and unwanted objects are incorrectly identified

Engineering Contradiction:

Improvecomprehensive object analysisVSAvoidobject of interest detection accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by allowing different regions or objects in the image to be processed with different recognition criteria and priorities. The system can assign different weights or thresholds to different detected objects based on their relevance, enabling precise identification of objects of interest while maintaining comprehensive analysis of all objects in the scene.

Inventive Principle:

Principle #3Local quality

3Productivity

If conventional techniques recognize all objects in an image, then complete object detection is achieved, but processor resources and time are consumed by unwanted objects

Engineering Contradiction:

Improvecomprehensive detection coverageVSAvoidprocessing time

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The patent implements partial action by performing complete object detection initially, then applying selective filtering and prioritization to focus processing resources on objects of interest. The system performs excessive detection of all objects but then reduces actual processing effort by identifying and concentrating resources on relevant objects, achieving both comprehensive coverage and efficient resource utilization.

Inventive Principle:

Principle #16Partial or excessive action

4Measurement precision

If a large amount of time is spent finding desired objects among similar prestored objects, then accurate identification is achieved, but time consumption is excessive

Engineering Contradiction:

Improveobject identification accuracyVSAvoidobject search time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary organization and indexing of prestored object characteristic information before recognition queries occur. By pre-structuring the reference database with efficient retrieval mechanisms, the system reduces the time required to search and compare objects while maintaining accurate identification through systematic comparison of characteristic information.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS9405986B2Apparatus and method for recognizing objects using filter information

Publication Date: 2016.08.02 APPLE INC

AI summary

An object recognition method using filter information includes acquiring object image information including an object of interest, acquiring filter information for recognizing the object of interest from the object image information, and recognizing the object of interest using the filter information. An object recognition apparatus using filter information including an object information acquiring unit to acquire object image information comprising an object of interest, a filter information input unit to acquire filter information, an output unit to output the image information and the filter information, and a controller to recognize the object of interest in the object image information using the filter information.