Faster R-Transformer: An efficient method for insulator detection in complex aerial environments

Abstract

Insulator inspection is important for high-voltage transmission line safety. However, existing object-detection algorithms cannot complete the task robustly in complex aerial environments. To improve the performance of the detection task, this paper proposes an algorithm, called Faster R-Transformer, which combines the advantage of Convolutional neural networks and the self-attention mechanism for aerial insulator detection. First, the D-ResNet performs feature extraction to obtain a larger receptive field. Then, the RPN module is used for region recommendation, and the Align Pooling module is used for region Pooling. Finally, the self-attention mechanism is adopted in the proposed region to perform local region feature weighting. The Faster R-Transformer was evaluated in complex aerial environments, and the evaluation showed that the proposed method was capable of detecting insulators of various sizes under different illuminations and viewpoints. We also explored hyperparameters of Faster R-Transformer to meet the requirements of various scenarios.