UAV Small-Object Recognition With Staged Image Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing systems struggle to efficiently detect and recognize small and ultra-small objects in images using machine learning algorithms on unmanned aerial vehicles (UAVs) due to limited computing power, leading to reduced recognition accuracy and increased processing time, which is particularly challenging in real-time applications.

Innovation Solution

A system and method utilizing a computing module with a machine learning algorithm, such as a convolutional neural network, on board the UAV for real-time detection and recognition of small objects, enabling classification and communication of results to a ground station via radio or cellular networks.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If machine learning algorithms are used for object detection on UAVs, then recognition accuracy improves, but processing time increases significantly

Engineering Contradiction:
Improveobject recognition accuracyVSAvoidimage processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The image processing is divided into multiple stages: initial filtering to identify potential small objects, followed by focused machine learning analysis only on those regions. This segmentation allows the system to maintain high recognition accuracy for small objects while avoiding the computational burden of applying complex algorithms to the entire image, thus reducing overall processing time.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs preliminary processing steps such as image enhancement, contrast adjustment, and preliminary object candidate identification before applying the full machine learning algorithm. This preliminary action prepares the data in advance, making the subsequent recognition process more efficient and accurate while reducing the time required for the main computational task.

Inventive Principle:
Principle #10Preliminary action

2Productivity

If simplified classification is used to reduce processing time, then productivity improves, but measurement precision deteriorates

Engineering Contradiction:
Improveprocessing speedVSAvoidclassification accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The classification system dynamically adjusts its complexity based on the specific application context and available computational resources. For time-critical applications, simplified classification is used with acceptable accuracy, while for applications requiring high precision, more complex classification algorithms are deployed. This dynamic adaptation allows the system to optimize the trade-off between processing speed and classification accuracy according to real-time needs.

Inventive Principle:
Principle #15Dynamics

3Loss of time

If computing resources are increased for real-time processing, then processing time decreases, but device complexity increases

Engineering Contradiction:
Improvereal-time processing capabilityVSAvoidcomputing module complexity
Core Design Contradiction:
Loss of timeVSDevice complexity

Solution Approach 1:

The system implements partial processing by focusing computational resources only on regions of the image that contain small objects, rather than processing the entire image at full resolution. This approach achieves real-time processing capabilities without requiring proportionally large increases in computing resources, thereby avoiding excessive device complexity while maintaining acceptable processing speeds.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4250251B1System and method for detecting and recognizing small objects in images using a machine learning algorithm
Publication Date: 2025.12.24 AO KASPERSKY LAB
  • EP4250251B1 patent drawingFigure 1
  • EP4250251B1 patent drawingFigure 2
  • EP4250251B1 patent drawingFigure 3A

AI summary

Disclosed are system and method for detecting small-sized objects based on image analysis using an unmanned aerial vehicle (UAV). The method includes obtaining object search parameters, wherein the search parameters include at least one characteristic of an object of interest; generating, during a flight of the UAV, at least one image containing a high-resolution image; analyzing the generated image using a machine learning algorithm based on the obtained search parameters; recognizing the object of interest using a machine learning algorithm if at least one object fulfilling the search parameters is detected in the image during the analysis; and determining the location of the detected object, in response to recognizing the object as the object of interest.