Object Detection Device Multi-Resolution ROI Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Objects appearing deep in the background of image data become excessively small when reduced, making them difficult to detect using conventional object detection models.

Innovation Solution

An object detection device that inputs both target and partial data to an object detection model, where the partial data is enlarged to a specific size, allowing for the detection of small objects by integrating results from both data sets.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If image data is reduced to a predetermined size for object detection, then processing efficiency is improved, but detection accuracy of small objects deteriorates

Engineering Contradiction:

Improveprocessing efficiencyVSAvoiddetection accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The image data is divided into multiple regions of interest (ROIs) based on detection results from reduced-size image data. Each ROI is then extracted and processed at full resolution, allowing small objects to be detected accurately while maintaining overall processing efficiency through selective high-resolution analysis.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system processes image data at multiple resolution levels - first at reduced size for overall scene understanding, then extracts specific regions and processes them at full resolution. This multi-scale approach allows the system to handle both large-scale scene analysis and small-object detection effectively.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Power

If image data is reduced to a predetermined size, then computational load is reduced, but the ability to detect small objects deteriorates

Engineering Contradiction:

Improvecomputational loadVSAvoiddetection reliability

Core Design Contradiction:

PowerVSReliability

Solution Approach 1:

The detection process is segmented into two stages: first, reduced-size image data is processed to identify potential regions of interest, which significantly reduces computational load; second, only the extracted ROI data at full resolution is processed for final detection, ensuring high detection reliability for small objects while maintaining low overall computational requirements.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system performs preliminary detection on reduced-size image data to identify candidate regions before processing the full-resolution ROI data. This preliminary action filters out most of the image data that doesn't contain small objects, reducing computational load while preserving detection reliability for actual small objects.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentEP4024333B1Object detection device, object detection method, object detection program, learning device, learning method and learning program

Publication Date: 2023.10.11 MITSUBISHI ELECTRIC CORP

Figure 1
Figure 2
Figure 3

AI summary

A data extraction unit (23) extracts, out of image data obtained by photographing a photographing region with a photographing device (41), image data of a region including a detection target region, as target data, and extracts, out of the target data, image data of an enlarging region, as partial data. A size modification unit (24) size-modifies each of the target data and the partial data to a request size requested by an object detection model being a model that detects an object from image data. An object detection unit (25) inputs each of the size-modified target data and the size-modified partial data to the object detection model, and detects a target object from each of the target data and the partial data.