|
The maintenance and inspection of power lines is key to ensuring their normal operation and maintaining an uninterrupted power supply for various human activities. Traditional methods to detect power line assets usually detect only a small number of assets, and face the challenge of low inspection accuracy and high computational resources. This study proposes a power line asset detection algorithm designed based on the YOLOv8 to address these issues. Firstly, the C2f-DC module is added to the algorithm’s backbone network, enhancing the feature extraction capability of the network. Then, the MSCA module is incorporated into the algorithm’s neck network. This module effectively captures multi-scale contextual information, thereby enhancing the feature extraction capability of the algorithm. Finally, the DyHead module and the Inner-IoU loss function are used to replace the original detection head and the original loss function, respectively. The Dyhead module dynamically adjusts feature representations, thus boosting the detection accuracy of the algorithm. The Inner-IoU loss function addresses the issues of poor generalization and slow convergence associated, thereby improving the performance of the algorithm. Experimental results demonstrate that the power line asset detection algorithm designed based on YOLOv8 achieves a mean average precision (mAP) of 90%, accurately detecting power line asset.
|
