基于深度学习算法的航拍绝缘子检测

Aerial Insulator Detection Based on Deep Learning Algorithm

针对无人机航拍图像中绝缘子等设备因背景复杂、过度遮挡而产生的检测准确性差的问题,提出了一种基于CornerNet-Lite网络模型的改进算法。该算法应用LeakyReLU函数设计了更为合理的损失函数,在coco数据集和自建绝缘子数据集上进行试验。对比原网络模型检测结果可知,该方法将绝缘子检测准确率提升至93%,有效解决了绝缘子目标被塔架等间断性遮挡及多目标聚集时模型出现漏检现象的问题,尤其是在目标被过度遮挡时更能体现优势。

Aiming at the problem of poor detection accuracy of insulators and other equipment in UAV aerial images due to complex background and excessive occlusion,an improved algorithm based on the CornerNet-Lite network model is proposed.The LeakyReLU function was used by the algorithm to design a more reasonable loss function and the test is carried out on coco data set and self-built insulator data set.Comparing the detection results of the original network model,it can be seen that this method improves the accuracy of insulator detection to 93%,effectively solving the problem that the insulator target is intermittently obscured by towers and other objects and the model is missed when the target is gathered,especially when the target is It is more advantageous when over occlusion.