Object Recognition Using Predicted Appearance Areas

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing object recognition systems on devices with low computational capability, such as vehicles and CCTV systems, face prolonged latency due to high computational demands, limiting their ability to quickly perform object recognition on arbitrary images.

Innovation Solution

An object recognition method using a machine learning model that predicts and defines a candidate area for object appearance on arbitrary images, allowing for efficient object recognition by integrating results with previous recognition data, and utilizing a separate server for dataset generation and machine learning operations to reduce computational load.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If deep neural network with high inference accuracy is used, then object recognition accuracy is improved, but computation time is prolonged

Engineering Contradiction:

Improveobject recognition accuracyVSAvoidcomputation time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent divides the image into multiple regions of interest based on detected target objects, and performs object recognition only on these segmented regions rather than the entire image. This segmentation approach maintains recognition accuracy for relevant objects while significantly reducing the computational burden by excluding irrelevant areas from processing.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies different processing quality levels to different regions of the image. Regions containing detected objects receive full-depth object recognition processing, while other regions are excluded or processed with reduced depth. This local quality differentiation maintains accuracy where needed while reducing overall computation time.

Inventive Principle:

Principle #3Local quality

2Productivity

If fixed regions of interest are established for object recognition, then computation amount is reduced, but adaptability to different situations is limited

Engineering Contradiction:

Improveprocessing speedVSAvoidapplicability to different situations

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent dynamically determines regions of interest based on detected target objects rather than using fixed predetermined regions. The system adapts the recognition areas to the actual content of each image by identifying objects and expanding around them, making the approach versatile across different situations while maintaining processing efficiency through focused regional analysis.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system automatically identifies target objects and generates appropriate regions of interest without requiring manual configuration or prior knowledge of specific photographing situations. This self-service capability enables the system to adapt to various scenarios autonomously while maintaining efficient processing by focusing only on relevant areas.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS11620360B2Methods and systems for recognizing object using machine learning model

Publication Date: 2023.04.04 RES & BUSINESS FOUND SUNGKYUNKWAN UNIV

AI summary

In accordance with an aspect of the present disclosure, there is provided an object recognizing method using a machine learning model performed by an object recognizing apparatus including, obtaining an image for recognizing an object, predicting an object appearance area on the obtained image by using a machine learning model that is trained to learn a feature of an area on which the object appears in advance and recognizing the object on the obtained image by using the predicted object appearance area.