Object Mask Merging for Multi-Resolution Video Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current image processing methods for monitoring cameras face challenges in accurately detecting and tracking objects across varying image resolutions, leading to inefficient data analysis and representation.

Innovation Solution

A method that combines object masks from sequences of image frames with different resolutions to create a common scene description, merging detailed and less detailed parts of object masks to provide a unified representation, thereby improving object tracking and reducing data analysis complexity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If object detection is performed on high-resolution image frames, then object recognition accuracy is improved, but data processing complexity and bandwidth requirements increase

Engineering Contradiction:
Improveobject recognition accuracyVSAvoiddata processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the image processing task into two segments: low-resolution processing for broad scene coverage and object detection, and high-resolution processing only for detected objects of interest. This segmentation allows the system to maintain low overall processing complexity while achieving high recognition accuracy for relevant objects.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different image qualities (resolutions) to different regions of the scene. High-resolution images are used only for detected objects where detailed analysis is needed, while the rest of the scene uses low-resolution images. This local quality approach reduces overall data processing complexity while maintaining high object recognition accuracy where required.

Inventive Principle:
Principle #3Local quality

2Loss of information

If multiple sequences of image frames with different resolutions are processed separately, then detailed object information is obtained, but data redundancy and bandwidth consumption increase

Engineering Contradiction:
Improvedetailed object informationVSAvoiddata redundancy
Core Design Contradiction:
Loss of informationVSQuantity of substance

Solution Approach 1:

The patent merges detection results from low-resolution processing with detailed information from high-resolution processing into a unified object description. This combining approach eliminates data redundancy by integrating information from both sequences rather than processing them separately, while preserving all relevant object details.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent uses detection results from low-resolution frames as an intermediary to guide selective processing of high-resolution frames. This intermediary step identifies which objects require detailed analysis, allowing the system to obtain comprehensive object information while minimizing unnecessary data transmission and processing.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Loss of information

If high-resolution images are transmitted for all objects, then complete object description is achieved, but bandwidth requirements and data transmission volume increase

Engineering Contradiction:
Improvecomplete object descriptionVSAvoidbandwidth consumption
Core Design Contradiction:
Loss of informationVSLoss of energy

Solution Approach 1:

The patent transmits only the necessary portion of high-resolution data - specifically, detailed information only for detected objects of interest rather than entire high-resolution frames. This local quality approach ensures complete object descriptions are achieved for relevant targets while significantly reducing bandwidth consumption compared to transmitting all high-resolution images.

Inventive Principle:
Principle #3Local quality

4Productivity

If low-resolution images are used for all objects, then data processing efficiency is improved, but object recognition accuracy deteriorates

Engineering Contradiction:
Improvedata processing efficiencyVSAvoidobject recognition accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent segments the processing pipeline into a first stage using low-resolution images for efficient scene-wide object detection, and a second stage using high-resolution images only for identified objects of interest. This segmentation maintains high data processing efficiency in the first stage while ensuring accurate object recognition in the second stage.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary detection using low-resolution images to identify objects of interest before conducting detailed analysis. This preliminary action filters out irrelevant data, allowing the system to maintain high processing efficiency while directing high-resolution processing only where needed for accurate object recognition.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP2662827B1Video analysis
Publication Date: 2016.01.13 AXIS
  • EP2662827B1 patent drawingFigure 1~3

AI summary

A method (200) and an object analyzer (104) for analyzing objects in images captured by a monitoring camera (100) uses a first and a second sequence of image frames, wherein the first sequence of image frames covers a first image area (300) and has a first image resolution, and the second sequence of image frames covers a second image area (302) located within the first image area (300) and has a second image resolution higher than the first image resolution. A common set of object masks is provided wherein object masks of objects (304) that are identified as being present in both image areas are merged.