Perspective Image Retrieval via Semantic Region Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image retrieval technologies for perspective images, such as those used in customs clearance and security services, struggle with inefficiencies due to the lack of consideration for semantic elements, leading to inaccurate and rough retrieval of objects within closed spaces.

Innovation Solution

A semantic-based method and apparatus that divides perspective images into visually separable semantic regions, utilizes deep learning for feature extraction, and retrieves similar regions from a history image database, providing detailed information for each region.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional image retrieval methods are used for perspective images, then the retrieval process is simple, but the retrieval accuracy is low and inefficient

Engineering Contradiction:
Improveretrieval accuracyVSAvoidretrieval system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the perspective image into multiple semantic regions based on visual separability and spatial relationships. Each semantic region is then processed and retrieved independently, which significantly improves retrieval accuracy by focusing on specific objects of interest rather than the entire image. This segmentation approach resolves the contradiction by making the retrieval process more precise without requiring complete analysis of the whole image.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different processing strategies to different semantic regions based on their local characteristics. Each semantic region is extracted and retrieved according to its specific visual features and spatial attributes, allowing the system to adapt to the local quality of different regions. This local quality approach enables high-accuracy retrieval for specific objects while maintaining overall system efficiency.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If the entire perspective image is retrieved as a whole, then the retrieval process is fast, but the detail and precision of object identification is insufficient

Engineering Contradiction:
Improveobject identification precisionVSAvoidretrieval time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

By segmenting the perspective image into multiple semantic regions, the system can retrieve only the relevant regions rather than processing the entire image. This segmentation enables high-precision object identification for specific regions of interest while reducing the overall retrieval time by avoiding unnecessary processing of irrelevant areas.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts and retrieves only the necessary semantic regions from the perspective image based on visual separability and spatial relationships. This extraction approach allows the system to provide detailed and precise object identification for important objects while minimizing the time spent processing the entire image, thus resolving the contradiction between precision and time.

Inventive Principle:
Principle #2Taking out (Extraction)

3Measurement precision

If semantic region division is performed, then the retrieval accuracy improves, but the processing complexity increases

Engineering Contradiction:
Improveretrieval accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements semantic region division by analyzing visual separability and spatial relationships in the perspective image. This segmentation improves retrieval accuracy by creating meaningful regions that can be retrieved independently. The processing complexity is managed through automated detection algorithms that efficiently identify semantic regions without requiring manual intervention.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs self-service by automatically detecting and dividing semantic regions based on visual separability and spatial relationships. The algorithm autonomously identifies appropriate regions without requiring external intervention or complex manual processing. This self-service approach improves retrieval accuracy while keeping the processing complexity manageable through automated detection and classification mechanisms.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3882816B1Semantic-based method and apparatus for retrieving perspective image
Publication Date: 2025.09.24 NUCTECH CO LTD
  • EP3882816B1 patent drawingFigure 1
  • EP3882816B1 patent drawingFigure 2
  • EP3882816B1 patent drawingFigure 3~4

AI summary

A semantic-based method and apparatus for retrieving a perspective image, an electronic device and a computer-readable storage medium are provided. The method includes: obtaining a perspective image for a space containing an inspected object therein; performing a semantic division on the perspective image using a first method, to obtain a plurality of semantic region units; constructing a feature extraction network using a second method; extracting, based on the perspective image and each of the plurality of semantic region units, a feature of each semantic region unit using the feature extraction network; and retrieving, based on the feature of each semantic region unit, an image most similar to the semantic region unit from an image feature database, to assist in determining an inspected object in the semantic region unit.