Feature Point Tracking for UAV Frame Extraction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for extracting still images from moving images captured by a UAV with a moving viewpoint are ineffective, as they rely on fixed viewpoints and cannot detect object movement or specific motions in uniform infrastructure structures.

Innovation Solution

An image processing device that detects feature points in a moving image, tracks their movement, and outputs a frame as a clipped frame when a predetermined displacement threshold is met, using a feature point detection tracking unit, movement calculation unit, and determination unit.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a fixed viewpoint capturing method is used, then object movement detection is simplified, but it cannot be applied to moving viewpoint scenarios

Engineering Contradiction:

Improveapplicability to moving viewpointVSAvoidcomplexity of movement detection

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent transforms the moving viewpoint problem into a solvable form by calculating viewpoint movement parameters (translation vector and rotation angle) from captured images. By compensating for these parameters in the coordinate system transformation, the system adapts moving viewpoint captures to the fixed viewpoint analysis framework, enabling applicability without excessive complexity

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces an intermediary coordinate transformation process that acts as a bridge between moving viewpoint captures and fixed viewpoint analysis. By transforming images to a standardized coordinate system that compensates for viewpoint movement, the system enables movement detection algorithms designed for fixed viewpoints to work effectively with moving viewpoint data

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If frame extraction is performed at fixed time intervals, then processing is simple, but important movement moments may be missed

Engineering Contradiction:

Improveprecision of movement detection timingVSAvoidprocessing efficiency

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent implements feedback-based frame extraction by continuously monitoring displacement amounts between consecutive frames. When the displacement exceeds a predetermined threshold, the system triggers frame extraction. This feedback mechanism ensures that important movement moments are captured with high precision while avoiding unnecessary extractions during static periods, thus maintaining processing efficiency

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent transitions from static, fixed-interval frame extraction to dynamic, event-driven extraction. By adjusting the extraction timing based on actual movement conditions (displacement thresholds), the system achieves both high measurement precision for important moments and maintains productivity by extracting frames only when necessary

Inventive Principle:

Principle #15Dynamics

3Reliability

If all frames are extracted and stored, then no movement information is lost, but storage requirements and processing load increase significantly

Engineering Contradiction:

Improvecompleteness of movement informationVSAvoidstorage capacity requirement

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential information (frames where displacement exceeds threshold) from the complete moving image data, rather than storing all frames. This selective extraction maintains reliability for movement analysis by capturing all significant movement moments while dramatically reducing the quantity of stored data compared to storing every frame

Inventive Principle:

Principle #2Taking out (Extraction)

Data Source

PatentUS20250005768A1Image processing device, image processing method, and program

Publication Date: 2025.01.02 NIPPON TELEGRAPH & TELEPHONE CORP

AI summary

An image processing device (10) according to the present disclosure includes a feature point detection tracking unit (12) that detects a feature point in a first frame that is a frame at a first time, tracks the feature point from the first time to a second time later than the first time, and calculates a two-dimensional vector indicating a motion of the feature point, a feature point movement amount calculation unit (13) that calculates a movement amount of the feature point based on the calculated two-dimensional vector, and a determination unit (14) that determines whether the calculated movement amount is equal to or greater than a predetermined threshold value, and outputs a second frame as the clipped frame, the second frame being a frame at the second time, when the movement amount is determined to be equal to or greater than the threshold value.