Application and Research of Small Object Detection in Oil Field

With the construction of intelligent oilfield, in the current operation site of oilfield, the computer has been able to analyze some violations intelligently through the deep neural network, while the recognition of small objects (small objects refer to less than 32 × 32 pixel points) such as playin...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2022 5th International Conference on Advanced Electronic Materials, Computers and Software Engineering (AEMCSE) s. 492 - 496
Hlavní autoři: Yang, Huaizhou, Liu, Chenzhuo
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 01.04.2022
Témata:
Feature extraction
Neural networks
Object detection
Oils
Semantics
small object detection
Software algorithms
Target recognition
transformer algorithm
Yolov5 object detection algorithm
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:With the construction of intelligent oilfield, in the current operation site of oilfield, the computer has been able to analyze some violations intelligently through the deep neural network, while the recognition of small objects (small objects refer to less than 32 × 32 pixel points) such as playing cell phone and smoking is still an unsolved problem. Small object detection has been a core task in computer vision, which is more difficult to recognize because it contains less semantic information and is not easy to distinguish from the background. In this paper, we analyze and experimentally apply the classical small target recognition methods and propose the Yolo-SO network, which is tested and verified in the dataset, and under the evaluation criteria of Ap75, the detection accuracy of Yolo-SO network is 2.6% higher than the existing classical network EFPN.
DOI:10.1109/AEMCSE55572.2022.00102