Object Detection Using Symmetry Evaluation for Reduced Processing Load

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing techniques face challenges in accurately detecting the position and size of objects in images while minimizing arithmetic processing load, which is essential for various applications.

Innovation Solution

An object detection apparatus that includes an image input unit, an image feature quantity extraction unit, and a symmetry evaluation unit, which evaluates symmetry by calculating weighted correlation values across a central axis, allowing for the detection of object size and position with high accuracy using less arithmetic processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional object detection techniques are used to obtain both position and size information, then detection accuracy is improved, but arithmetic processing load increases

Engineering Contradiction:
Improveobject position and size detection accuracyVSAvoidarithmetic processing load
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

The patent extracts only the essential symmetry information from images by calculating correlation values between symmetrical pixel regions around a central axis. Instead of processing entire images or using complex detection algorithms, the method extracts symmetry correlation values for candidate object regions, significantly reducing arithmetic processing while maintaining detection accuracy for position and size.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the detection parameter from comprehensive image analysis to symmetry-based correlation calculation. By varying the size of symmetry evaluation areas and calculating correlation coefficients between symmetrical pixel groups, the method transforms the detection problem into a parameter optimization task that requires fewer computational resources while achieving accurate object position and size detection.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If symmetry evaluation with varied area sizes is performed, then object detection accuracy is improved, but processing complexity increases

Engineering Contradiction:
Improvesymmetry evaluation accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the image processing task into distinct stages: first identifying candidate object regions, then evaluating symmetry for each candidate by calculating correlation values between symmetrical pixel groups. This segmentation allows the system to apply symmetry evaluation only where needed rather than processing the entire image, reducing overall complexity while maintaining accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs symmetry evaluation with varied area sizes for candidate object regions rather than for the entire image. By applying the symmetry evaluation process selectively to potential object locations and using a limited range of area sizes, the method achieves sufficient detection accuracy without the excessive complexity of exhaustive evaluation across all possible regions and scales.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP2830020B1Object detection device and program
Publication Date: 2019.05.01 MEGACHIPS
  • EP2830020B1 patent drawingFigure 1
  • EP2830020B1 patent drawingFigure 2
  • EP2830020B1 patent drawingFigure 3(a)~3(g)

AI summary

Important information about an object is detected using less arithmetic processing. An object detection unit (22) generates an edge image from a color image. The object detection unit (22) evaluates symmetry of an image included in the edge image. The object detection unit (22) identifies a symmetry center pixel forming an object having symmetry. The object detection unit (22) detects an object width for each symmetry center pixel. The object detection unit (22) identifies the width of the object in the vertical direction based on the width of the symmetry center pixels in the vertical direction, and identifies the width of the object in the horizontal direction based on the object width identified for each symmetry center pixel.