Image Processing Device Principal Face Detection via Apex Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current object recognition systems, particularly in automotive applications, face challenges in accurately identifying the principal face of an object, such as distinguishing between the back face and lateral faces, which is crucial for vehicle control systems to make informed decisions during collisions.

Innovation Solution

An image processing device that includes a first extracting unit for extracting a first area representing an object, a second extracting unit for extracting contour directions, a face detecting unit for detecting a first face, and a decider to determine the principal face based on the detected face and apex information.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If only size information is extracted from parallax images, then object recognition is simplified, but orientation information is lost making it difficult to distinguish between different types of objects

Engineering Contradiction:
Improveobject recognition processingVSAvoidorientation information
Core Design Contradiction:
Device complexityVSLoss of information

Solution Approach 1:

The patent segments the object detection process into multiple stages: first extracting size information from parallax images, then separately extracting orientation information by detecting apices and face directions. This segmentation allows the system to handle complex objects like vehicles by breaking down the recognition task into manageable components, resolving the contradiction between simplified processing and information completeness.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from two-dimensional size measurement to three-dimensional orientation analysis by detecting apices (boundary points) and face directions. This dimensional expansion enables the system to distinguish between objects of the same size but different orientations, such as identifying whether a detected object is a vehicle facing forward or sideways, thereby preventing information loss.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Loss of information

If apex detection is performed to identify boundary points, then orientation information is obtained, but it becomes difficult to determine which face is the principal face

Engineering Contradiction:
Improveorientation informationVSAvoidprincipal face identification
Core Design Contradiction:
Loss of informationVSDifficulty of detecting and measuring

Solution Approach 1:

The patent introduces a decider component as an intermediary that receives apex detection results and face direction information, then synthesizes this data to identify the principal face. This intermediary layer processes the raw apex data and contextual information to make intelligent decisions about which face should be considered the principal one, resolving the difficulty in face identification.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system employs feedback mechanisms where the decider uses apex information and face direction data to determine the principal face, and this determination then feeds back into the object recognition process. This feedback loop allows the system to continuously refine its understanding of object orientation and principal face identification, making the complex task more manageable.

Inventive Principle:
Principle #23Feedback

3Loss of information

If multiple faces are detected for a single object, then comprehensive object information is obtained, but processing complexity increases

Engineering Contradiction:
Improvecomprehensive object informationVSAvoidface processing complexity
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent extracts and separates the principal face information from the complete set of detected faces using the decider. By isolating only the essential principal face data for further processing while discarding redundant information about non-principal faces, the system maintains comprehensive object information where needed while reducing processing complexity for control decisions.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP3392830B1Image processing device, object recognition device, apparatus control system, image processing method and program
Publication Date: 2023.09.06 RICOH CO LTD
  • EP3392830B1 patent drawingFigure 1~2(b)
  • EP3392830B1 patent drawingFigure 3
  • EP3392830B1 patent drawingFigure 4~4(b)

AI summary

An image processing device includes a first extracting unit configured to extract, based on distance information regarding an object, a first area representing the object; a second extracting unit configured to extract contour directions of a contour of the first area; a face detecting unit configured to detect a first face in the first area, based on the contour directions extracted by the second extracting unit; a first apex detecting unit configured to detect a first apex indicating a boundary of the first face; and a decider configured to decide on a principal face of the object represented by the first area, based on at least the first face and the first apex.