| Summary: | 碩士 === 國立高雄應用科技大學 === 電子工程系碩士班 === 101 === Purpose: In this study, we constructed an insulator spark detection system, which integrated visual image monitoring devices and an analysis system to reduce the complexity of the insulator spark detection system.
Materials and Methods: The image of an insulator spark was captured by a video capture device, such as a video monitoring device or webcam. The captured images were used as the input for image pre-processing, insulator spark characteristic detection, and area and brightness analysis. The result was then applied to the construction of a logistic regression model of current leakage for identifying the relationship between the brightness of the insulator spark and current leakage, which is used in warning operations of current leakage at different degrees of leakage.
Results: The characteristics of insulator spark were extracted and identified clearly after the pre-processing of the insulator spark image. The analysis of the logistic regression model of current leakage showed that the variables chosen by Omnibus examination were significant (p-value of <0.05). Hosmer-Lemeshow examination was suitable when the p-value was >0.05. The estimated area under the receiver operation characteristic curve (AUC) was 0.997, which is above the average value. The prediction of current leakage of an insulator spark and warning operations can be implemented based on the cut point, spark brightness, and the relationships that are induced from the logistic regression model of insulator spark. Consequently, reliable and usable insulator spark detection can be constructed for operators.
Conclusion: This paper demonstrates that the accuracy of prediction of current leakage using a logistic regression model is acceptable. Nevertheless, more studies are needed for practical application.
|
|---|
