Differential twin network transformer substation image difference and difference detection method
The invention discloses a difference twin network transformer substation image similarities and differences detection method, which comprises end-to-end target and area detection, and the implementation process of the method comprises data acquisition, data annotation, deep learning network feature...
Gespeichert in:
Hauptverfasser: | , , |
---|---|
Format: | Patent |
Sprache: | chi ; eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | The invention discloses a difference twin network transformer substation image similarities and differences detection method, which comprises end-to-end target and area detection, and the implementation process of the method comprises data acquisition, data annotation, deep learning network feature extractor model design, deep learning network error design and model training. According to the method, the deep twin network is applied, normal pictures and abnormal pictures are jointly detected, the detection speed is greatly increased, a reference data set is composed of a public data set and a substation acquisition data set, and a training data set randomly generates training samples at different angles through reference data rotation and illumination scrambling; the method gets rid of instability caused by influence of noise light distortion and the like in a traditional algorithm, and accuracy, robustness and usability are improved.
本发明公开了一种差分孪生网络变电站图像异同检测方法,所述方法包含端到端的目标与区域检测,所述方法的实施过程包括数据采集,数据标注,深度学习网络特诊提取 |
---|