A novel helicopter flight action recognition method based on flight parameter data processing

doi:10.1016/j.engappai.2025.113522

A novel helicopter flight action recognition method based on flight parameter data processing

2025-12-17

0

PRE

AI

DOI：10.1016/j.engappai.2025.113522

Original

Original request for help

Save

Abstract

En 中文

Helicopter flight requires a large number of complex manipulation and coordination actions. Accurately recognizing helicopter flight actions is a fundamental step for advanced applications such as flight safety assessment and pilot training. However, current helicopter flight action recognition methods generally face challenges such as excessive reliance on expert experience, insufficient multidimensional data fusion capabilities and poor recognition accuracy. For these reasons, this paper proposes a helicopter flight action recognition method based on flight parameter data. The method adopts a data-driven approach as the core, using importance ranking and feature filtering to preprocess flight parameter data and build a systematic framework for flight action recognition, aiming to achieve high-precision, automated annotation of flight maneuvers. The data quality is effectively improved by combining the Random Forest and variance filtering method. Then, the stepwise feature selection method is adopted to explore the influence mechanism of flight parameter feature combinations on the recognition of flight actions. By integrating Bidirectional Long-Short Term Memory (BiLSTM), PatchMixer and gated Multilayer Perceptron (gMLP), the Bidirectional PatchMixer-Gated Multilayer Perceptron Network (BiPatch-GNet) is constructed to provide a more reliable and efficient solution for helicopter flight action recognition. The experimental results show that the helicopter flight action recognition method proposed in this paper can achieve 97.18% accuracy on five common flight actions, which are Wing-level flight, Rise, Glide, Turn and Hover. This work provides a reliable technical foundation for generating large-scale, accurately labeled flight datasets, which is a prerequisite for future safety-oriented applications.

A novel helicopter flight action recognition method based on flight parameter data processing

Abstract

Journal

Organization

Cited Papers