Hand-Side Motion Region Extraction for Occluded Object Recognition

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image recognition techniques struggle to accurately extract target objects when images are captured from above, as target objects may be hidden or obscured by other objects, leading to incorrect recognition.

Innovation Solution

A region extraction device and method that utilizes motion information to detect the positions of a human body's elbow and wrist, divides the motion region into hand and arm sides using the wrist as a boundary, and extracts the region corresponding to the hand side, employing color comparison and joint detection to refine the target object's region, expanding it based on finger joint and fingertip positions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If image is captured from above, then overview of the scene is obtained, but target object may be hidden and cannot be recognized

Engineering Contradiction:
Improvecoverage areaVSAvoidobject recognition accuracy
Core Design Contradiction:
Area of stationary objectVSMeasurement precision

Solution Approach 1:

The motion region is segmented into multiple parts using the wrist position as a reference point. By dividing the motion region into hand-side and arm-side portions, the system can focus on the hand-side region where the target object is most likely located, improving recognition accuracy while maintaining comprehensive coverage through multi-frame analysis.

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If image is captured from store shelf or outlet vicinity, then target object visibility is improved, but various other objects will be captured causing incorrect recognition

Engineering Contradiction:
Improvetarget object visibilityVSAvoidfalse positive recognition
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The system extracts only the hand-side portion of the motion region by using the wrist position as a reference point and dividing the motion region accordingly. This extraction isolates the region containing the target object from other objects in the scene, eliminating false positives while maintaining target visibility.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of analyzing the entire motion region, the system applies local quality analysis by focusing specifically on the hand-side portion. This localized approach improves recognition accuracy by concentrating computational resources on the most relevant area where the target object is held, while ignoring other regions that may contain distracting objects.

Inventive Principle:
Principle #3Local quality

3Productivity

If motion region is used for object detection, then dynamic object tracking is achieved, but hand and arm regions are mixed making precise hand object extraction difficult

Engineering Contradiction:
Improveobject tracking capabilityVSAvoidhand object region extraction accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The motion region is divided into multiple segments using the wrist position as a reference point. By segmenting the motion region into hand-side and arm-side portions, the system maintains the dynamic tracking capability of the entire motion region while enabling precise extraction of the hand object region through selective processing of the hand-side segment.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3958212B1Region extraction device, region extraction method, and region extraction program
Publication Date: 2026.03.11 RAKUTEN GROUP INC
  • EP3958212B1 patent drawingFigure 1
  • EP3958212B1 patent drawingFigure 2
  • EP3958212B1 patent drawingFigure 3

AI summary

An object is to extract a region in which a target object is shown from an image more accurately. A region extraction device acquires a first frame image and a second frame image which are temporally successive. The region extraction device acquires, based on the acquired first frame image and second frame image, motion information indicating a region where a motion exists in the first frame image. The region extraction device detects, based on the acquired first frame image, positions of an elbow and a wrist of a human body from the region where the motion exists indicated by the acquired motion information. The region extraction device extracts, based on the detected positions, a region corresponding to a part, of the region where the motion exists indicated by the acquired motion information, located in a human body hand's side of the wrist.