Aviation Plug Clustering Based Fault Detection Method Using Hyperspectral Image

Through clustering the hyperspectral image and registering with plug standard template for the fault plug pin detection is proposed in this paper. This method firstly searches for the center and radius of the hyperspectral image of plug and obtains the image only containing the chassis and the pin. In this paper, k-means clustering is used to finely cluster the plug pin. After registering with the standard plug template, the fault pin is detected. The key of this method is to use the spectral resolution of hyperspectral which can distinguish different substances accurately. It can also solve the problems that the pin and the chassis are similar in color and the plug frame is too high causing a lot of shadow on the image. The clustering accuracy is greatly improved in this paper. For the plug detection, the traditional method is to edge detect the aviation plug square frame. The fault pin is detected by identifying the parameters and registering with the plug standard template. However, since the position of the square frame is estimated, there is a large error in the detection results. In the experiments, two main detection methods are compared and the results show that the proposed method in the paper has higher accuracy.