UAV Multi-Camera Auto-Return Using Selected Visual Features

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Aerial vehicles using a single camera face challenges in auto-returning due to differences in image recording when changing directions, which can prevent effective visual cognition for navigation.

Innovation Solution

Equipping aerial vehicles with multiple cameras facing different directions to provide comprehensive 360-degree coverage, allowing for reliable image data collection and navigation using selected features for auto-return, reducing processing and storage demands by storing only relevant image data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a single camera is used for aerial photography, then the device complexity is reduced, but the auto-return reliability deteriorates due to inability to capture images in all directions

Engineering Contradiction:
Improveauto-return reliabilityVSAvoiddevice complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent divides the single camera system into multiple cameras, each responsible for capturing images in specific directions (front, back, left, right). This segmentation allows the system to overcome the limitation of a single camera's fixed field of view and enables reliable auto-return by providing comprehensive environmental coverage from all orientations

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent makes the camera system multi-functional by equipping the aerial vehicle with multiple cameras that can capture images in various directions simultaneously. This universal coverage ensures that regardless of the vehicle's orientation during return flight, the appropriate camera will capture the necessary visual information for navigation

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Reliability

If multiple cameras are equipped on the aerial vehicle, then the auto-return reliability is improved through comprehensive image coverage, but the processing and storage burden increases

Engineering Contradiction:
Improvevisual cognition reliabilityVSAvoidprocessing efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent extracts only the essential visual information from the multiple camera feeds by identifying and storing key feature points rather than processing entire images. This extraction approach maintains the reliability benefits of multiple cameras while dramatically reducing the processing and storage burden by focusing only on salient navigational features

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies partial action by selectively processing only certain images or image regions from the multiple cameras rather than analyzing all image data in full detail. This approach maintains sufficient visual information for reliable navigation while reducing overall computational load

Inventive Principle:
Principle #16Partial or excessive action

3Measurement precision

If all image data from multiple cameras is stored, then the visual cognition accuracy is improved, but the memory storage requirement increases

Engineering Contradiction:
Improvevisual cognition accuracyVSAvoiddata storage volume
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts only the critical visual information needed for navigation by identifying feature points in the images captured by multiple cameras. Instead of storing complete images, only these extracted feature points are stored, maintaining visual cognition accuracy while minimizing storage requirements

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the parameter of data representation from storing complete image data to storing condensed feature point data. This parameter transformation preserves the essential visual information needed for accurate navigation while dramatically reducing the quantity of stored data

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3835912B1Systems and methods for auto-return
Publication Date: 2025.06.25 SZ DJI TECH CO LTD
  • EP3835912B1 patent drawingFigure 1
  • EP3835912B1 patent drawingFigure 2
  • EP3835912B1 patent drawingFigure 3

AI summary

Systems and methods for vision-based auto-return are provided for an unmanned aerial vehicle (UAV) (100, 410) traversing an environment. The UAV (100, 410) may comprise multiple cameras (120a, 120b, 120c, 120d, 220, 415) with different orientations that may capture different fields of view (130a, 130b, 415). Image data collected by the multiple cameras (120a, 120b, 130c, 120d, 220, 415) may be stored or used for navigational aid, such as performing the auto-return. A selected portion of the image data, such as selected image features, may be stored or used for auto-return.