Subject Detection Accuracy Across Varying Capture Angles

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional subject detection technologies based on image features face reduced accuracy when the capturing angle is large due to the scarcity of training data, making it difficult to detect subjects in regions with high capturing angles effectively.

Innovation Solution

A system that divides a captured image into regions where subject detection is performed using image features and regions where detection is based on alternative factors, allowing for simultaneous subject detection and correction of detection accuracy by switching between methods based on capturing angle and previous frame analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If image feature-based detection is used for large capturing angles, then detection coverage is improved, but detection accuracy deteriorates due to scarcity of training data

Engineering Contradiction:
Improvedetection coverageVSAvoiddetection accuracy
Core Design Contradiction:
Area of stationary objectVSMeasurement precision

Solution Approach 1:

The patent segments the captured image into multiple regions based on capturing angle characteristics. A first region (small capturing angle) uses image feature-based detection, while a second region (large capturing angle) uses moving object detection. This segmentation allows each region to use the most appropriate detection method, maintaining high accuracy while covering the entire image area.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent dynamically switches between different detection methods based on the capturing angle of each region. The system adjusts the detection approach in real-time according to the regional characteristics, using image feature detection for small angles and moving object detection for large angles, thereby optimizing detection accuracy across varying conditions.

Inventive Principle:
Principle #15Dynamics

2Measurement precision

If image feature-based detection is used for small capturing angles, then detection accuracy is improved, but detection capability is lost in regions with large capturing angles

Engineering Contradiction:
Improvedetection accuracyVSAvoiddetection capability across regions
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent implements a multi-functional detection system that can perform both image feature-based detection and moving object detection. The system universally applies both methods across different regions, selecting the appropriate one based on capturing angle characteristics. This multi-functionality ensures high detection accuracy is maintained while adapting to various capturing angle scenarios.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent changes the detection parameter (detection method) based on the capturing angle parameter. For small capturing angles, image feature detection is applied; for large capturing angles, moving object detection is applied. This parameter change allows the system to maintain optimal detection accuracy across different regional conditions.

Inventive Principle:
Principle #35Parameter changes

3Device complexity

If a single detection method is used for the entire image, then system complexity is reduced, but detection accuracy varies significantly across different regions

Engineering Contradiction:
Improvedetection system complexityVSAvoidregional detection accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent segments the image into regions with different capturing angle characteristics and applies different detection methods to each segment. This segmentation strategy maintains relatively simple system architecture while significantly improving regional detection accuracy by matching detection methods to regional characteristics.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3496390B1Information processing device, information processing method, and storage medium
Publication Date: 2023.04.26 CANON KK
  • EP3496390B1 patent drawingFigure 1
  • EP3496390B1 patent drawingFigure 2A~2B
  • EP3496390B1 patent drawingFigure 3A~3B

AI summary

An information processing device according to one embodiment of the present invention includes: an acquisition unit that acquires a captured image; and an output unit that outputs a detection result of a subject detection process based on an image feature, which is for a region where subject detection based on an image feature of the captured image is performed, and a detection result of a subject detection process based on a factor other than an image feature, which is for a region other than the region where the subject detection based on the image feature of the captured image is performed.