Real-Time Moving Target Detection Using Vision Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional image detection algorithms fail to effectively detect moving objects when the sensor or platform is in motion, requiring complex processing and large memory resources.

Innovation Solution

A system using registered frame differencing with self-adjusting noise thresholds to generate a field-by-field motion pixel map, followed by an efficient two-step pixel grouping process that writes links into a Group Table in real-time, reducing memory requirements and iterations through the data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional segmentation algorithms (thresholding, histogram-based methods, edge detection, clustering, region growing) are used to detect moving objects, then detection capability is provided, but processing complexity increases and the system breaks down when sensor or platform is in motion

Engineering Contradiction:
Improvedetection capabilityVSAvoidprocessing complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the image processing task into distinct phases: motion detection phase (comparing current frame with reference frame to identify moving pixels) and region labeling phase (grouping adjacent moving pixels). This segmentation allows each phase to be optimized independently, reducing overall processing complexity while maintaining detection reliability

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary motion detection by comparing the current frame with a reference frame before performing region labeling. This preliminary action identifies candidate moving pixels, allowing the subsequent region labeling to focus only on relevant areas rather than processing the entire image, thus reducing processing complexity

Inventive Principle:
Principle #10Preliminary action

2Reliability

If region labeling is performed using multiple passes through the data, then complete region identification is achieved, but memory requirements and processor intensity increase significantly

Engineering Contradiction:
Improveregion identification completenessVSAvoidmemory requirements
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent divides region labeling into two separate passes: first pass establishes initial region assignments and creates a region table, second pass resolves ambiguities and merges regions. This segmentation of the labeling process reduces the memory burden compared to traditional multi-pass approaches while maintaining complete region identification

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts and stores region information in a separate region table during the first pass, removing the need to repeatedly access and process the entire image data in subsequent passes. This extraction of essential region attributes reduces memory requirements by avoiding redundant data storage

Inventive Principle:
Principle #2Taking out (Extraction)

3Measurement precision

If complex processing algorithms are used to handle platform and sensor motion, then detection accuracy is maintained, but processing speed decreases

Engineering Contradiction:
Improvedetection accuracyVSAvoidprocessing speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent performs preliminary motion detection by comparing current frame with reference frame to identify moving pixels before performing detailed region analysis. This preliminary action filters out static background, allowing subsequent processing to focus only on moving regions, thus maintaining detection accuracy while improving processing speed

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent applies processing only to portions of the image containing moving pixels rather than the entire image. By performing region labeling only on identified moving regions and using a compact region table, the system achieves partial processing that maintains accuracy for moving objects while significantly improving overall processing speed

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3213297B1System for real-time moving target detection using vision based image segmentation
Publication Date: 2020.05.27 RAYTHEON CO
  • EP3213297B1 patent drawingFigure 1
  • EP3213297B1 patent drawingFigure 2A~2B
  • EP3213297B1 patent drawingFigure 3~4

AI summary

A system for processing video signal information to identify those pixels associated with a moving object in the presence of platform and/or sensor pointing induced motion. Frame differencing with self-adjusting noise thresholds is implemented to detect pixels associated with objects that are in motion with respect to the background and a field-by-field motion pixel map of pixels associated with the moving object is generated. A two (2) step pixel grouping process is used where the first pass runs in real-time as the video signal is received and writes the links between pixel groups into entries in a table. The second pass operates on a smaller set of link data and only needs to reorder entries in the table.