A Hybrid Neural Architecture Search Algorithm Optimized via Lifespan-PSO for Coal Mine Image Recognition

Coal mine scene image recognition is vital for safety monitoring and equipment detection, but traditional methods often rely on manually designed neural network architectures. These models face difficulties in handling the complex backgrounds, low illumination, and diverse objects typical in coal mine environments. Manual designs are not only inefficient but also limit the exploration of optimal architectures, leading to subpar performance. To address these challenges, we propose using Neural Architecture Search (NAS) to automate the design of neural networks. While traditional NAS methods are computationally intensive, we enhance the process by incorporating Particle Swarm Optimization (PSO), a scalable algorithm known for its ability to balance global and local search. To further improve PSO’s efficiency, we integrate the Lifespan mechanism, which prevents premature convergence and ensures more thorough exploration of the search space. Our proposed method defines a flexible search space that includes various types of convolutional layers, activation functions, pooling operations, and network depths, allowing for a comprehensive optimization process. Extensive experiments demonstrate that the Lifespan-PSO NAS method outperforms traditional manually designed networks and standard PSO-based NAS approaches, offering significant improvements in both recognition accuracy and computational efficiency. This makes it a highly effective solution for real-world coal mine image recognition tasks via a PSO-optimized approach in terms of performance and efficiency.