Abstract: To address the difficulty of blasthole detection during the charging phase of drill-and-blast tunnelling, which is aggravated by dust interference and insufficient illumination, this study proposes an intelligent blasthole detection model based on a hybrid neural network. First, a multi-class classification module accurately categorises blasthole images acquired in complex environments; a feature transformation module then converts these images into equivalent ones with a clear background. Subsequently, a dedicated blasthole detection module identifies the blastholes and localises their positions. By strengthening the feature-extraction capability of deformable convolutions, introducing a triple-attention mechanism, and refining the loss function, the model achieves a significant improvement in detection accuracy under adverse conditions. Experimental results demonstrate that, in complex environments, the proposed model attains a detection precision of 94.47 % and a recall of 86.32 %. Compared with state-of-the-art deep-learning object detectors, the proposed model exhibits superior robustness and blasthole detection capability, reliably identifying blasthole locations that conventional models often miss, thereby providing a solid foundation for intelligent charging in tunnelling excavation.