Transmission Line Instance Segmentation Algorithm Based on YOLACT

In the field of intelligent power patrol inspection, the transmission line is an important identification and detection target. The measurement of line spacing and ground distance are key technologies in the field of inspection. Therefore, it is necessary to quickly and accurately segment transmissi...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Journal of physics. Conference series Ročník 2562; číslo 1; s. 12018 - 12023
Hlavní autoři: Chen, Yong, Wang, Yun-hui, Li, Song, Li, Meng
Médium: Journal Article
Jazyk:angličtina
Vydáno: Bristol IOP Publishing 01.08.2023
Témata:
Algorithms
Inspection
Instance segmentation
Physics
Target detection
Transmission lines
ISSN:1742-6588, 1742-6596
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:In the field of intelligent power patrol inspection, the transmission line is an important identification and detection target. The measurement of line spacing and ground distance are key technologies in the field of inspection. Therefore, it is necessary to quickly and accurately segment transmission lines. The transmission lines occupy a large span and vary widely in length. To improve the segmentation rate and accuracy of the transmission lines, we adopted EfficientNet as the main network. With the same accuracy, the number of parameters is reduced by 80% compared with ResNet 101. The network training cost is reduced, and the detection rate of the model is improved. To deal with the influence caused by the large change in transmission line length, we introduce adaptive anchor box calculation and the FPN + PAN structure. At the same time, multiple transmission lines are often overlapped, so the traditional NMS makes it easy to cause the omission or confusion of lines. We improve the NMS. Finally, we adopted the modified CIoU loss function to optimize the loss function. From the experimental results, our model has good performance for instance in segmenting transmission lines.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1742-6588
1742-6596
DOI:10.1088/1742-6596/2562/1/012018