Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Learning Based Multi-Level Classification for Aviation Safety

Elaheh Sabziyan Varnousfaderani, Syed A. M. Shihab, Jonathan King

TL;DR

The paper tackles the risk of bird strikes in aviation by addressing the lack of species-level information in current avian-radar systems. It proposes two CNN-based strategies—Cascade Classification Approach (CCA) and Unified Classification Approach (UCA)—to identify bird species and extract flock characteristics, using a dataset of the top 33 US strike species and synthetic flock imagery, with ResNet50V2 as the core architecture. Results show CNNs outperform traditional baselines (SVM, RF, KNN) across size- and species-specific tasks, with UCA achieving 92.80% overall accuracy and CCA achieving 90.77% end-to-end accuracy; robustness tests reveal degradation under heavy rain, snow, noise, and low light, while flock-type and flock-size classifiers provide additional actionable insights. The study demonstrates a practical pathway to species-aware flight-path prediction and mitigation, though it acknowledges the need for real-world validation of flock data and future refinements to improve robustness and integration with predictive safety systems.

Abstract

Bird strikes pose a significant threat to aviation safety, often resulting in loss of life, severe aircraft damage, and substantial financial costs. Existing bird strike prevention strategies primarily rely on avian radar systems that detect and track birds in real time. A major limitation of these systems is their inability to identify bird species, an essential factor, as different species exhibit distinct flight behaviors, and altitudinal preference. To address this challenge, we propose an image-based bird classification framework using Convolutional Neural Networks (CNNs), designed to work with camera systems for autonomous visual detection. The CNN is designed to identify bird species and provide critical input to species-specific predictive models for accurate flight path prediction. In addition to species identification, we implemented dedicated CNN classifiers to estimate flock formation type and flock size. These characteristics provide valuable supplementary information for aviation safety. Specifically, flock type and size offer insights into collective flight behavior, and trajectory dispersion . Flock size directly relates to the potential impact severity, as the overall damage risk increases with the combined kinetic energy of multiple birds.

Deep Learning Based Multi-Level Classification for Aviation Safety

TL;DR

The paper tackles the risk of bird strikes in aviation by addressing the lack of species-level information in current avian-radar systems. It proposes two CNN-based strategies—Cascade Classification Approach (CCA) and Unified Classification Approach (UCA)—to identify bird species and extract flock characteristics, using a dataset of the top 33 US strike species and synthetic flock imagery, with ResNet50V2 as the core architecture. Results show CNNs outperform traditional baselines (SVM, RF, KNN) across size- and species-specific tasks, with UCA achieving 92.80% overall accuracy and CCA achieving 90.77% end-to-end accuracy; robustness tests reveal degradation under heavy rain, snow, noise, and low light, while flock-type and flock-size classifiers provide additional actionable insights. The study demonstrates a practical pathway to species-aware flight-path prediction and mitigation, though it acknowledges the need for real-world validation of flock data and future refinements to improve robustness and integration with predictive safety systems.

Abstract

Bird strikes pose a significant threat to aviation safety, often resulting in loss of life, severe aircraft damage, and substantial financial costs. Existing bird strike prevention strategies primarily rely on avian radar systems that detect and track birds in real time. A major limitation of these systems is their inability to identify bird species, an essential factor, as different species exhibit distinct flight behaviors, and altitudinal preference. To address this challenge, we propose an image-based bird classification framework using Convolutional Neural Networks (CNNs), designed to work with camera systems for autonomous visual detection. The CNN is designed to identify bird species and provide critical input to species-specific predictive models for accurate flight path prediction. In addition to species identification, we implemented dedicated CNN classifiers to estimate flock formation type and flock size. These characteristics provide valuable supplementary information for aviation safety. Specifically, flock type and size offer insights into collective flight behavior, and trajectory dispersion . Flock size directly relates to the potential impact severity, as the overall damage risk increases with the combined kinetic energy of multiple birds.
Paper Structure (25 sections, 6 equations, 17 figures, 13 tables)

This paper contains 25 sections, 6 equations, 17 figures, 13 tables.

Table of Contents

  1. Introduction
  2. Bird Strikes
  3. How Bird Strikes Would Be Actively Prevented?
  4. Outline of the Paper
  5. Literature Review
  6. Problem Statement
  7. Methodology
  8. Convolutional Neural Networks
  9. Benchmark Models: SVM, RF, KNN
  10. Results
  11. Species Profiling, Size Classification, Experimental Setup, and Performance Metrics
  12. Size Categorization
  13. Experimental Configuration
  14. Performance Metrics
  15. The Performance of the Cascade Classification Approach
  16. ...and 10 more sections

Figures (17)

  • Figure 1: Total number of annual bird strikes with civil aircraft in the United States, 1990-2023 faa2023wildlife
  • Figure 2: Annual number of commercial flights and bird strike rates (2003–2023) (Adapted from varnousfaderani2025bird)
  • Figure 3: Workflow for bird strike prevention using avian sensors (e.g., radars, cameras), a classifier and predictive models (adapted from sabziyan2025deep)
  • Figure 4: Flowchart illustrating the two proposed approaches for bird species identification. The CCA decomposes the problem into three stages: bird detection, size classification, and species classification using size-specific models. The UCA uses a single classifier to simultaneously identify bird species or detect aircraft inputs.
  • Figure 5: Two-stage classification framework for flock type identification. Bottom-view images are first used to classify the horizontal flock formation. If a column formation is detected, side-view images are used to determine the vertical alignment (ascending, descending, or level).
  • ...and 12 more figures