Bad State Detection Of Support And Suspension Device In Catenary Based On Image Processiing

The support and suspension devices of catenary play an important role in supporting the contact network, whose state can affect the stability of whole suspension system and the performance of contact line. Its failure may result into that the pantograph and catenary cannot contact with each other well, thus to affect the quality of current collection. At present, the efficiency of manual detection as the main mode in detection of the state of support and suspension devices is very low. The existing intelligent detection methods focus on the pantograph location, slide wear overrun inspection and catenary parameters identification etc. The bad state of the device can lead to potential safety hazard, so it is necessary to study the real-time and intelligent detection method.The detection was based on the catenary support and suspension image, and applied the image processing technology in this paper. Then it proposed some methods of the catenary support and suspension bad state detection through image processing, and realized the detection of insulator and rotary double-ear. Firstly, the images were pre-processed. Secondly, the target images were recognized and extracted with the feature-based image matching method. Finally, the local images of insulator and rotary double ear were analyzed and processed, and realized the detection of its state.This paper respectively used the affine moment invariant, SIFT and SURF algorithms to recognize the insulators and rotary double ears. And then it was concluded that the SURF algorithm is the best method to recognize the support and suspension device of catenary through comparing the identification effect.In the insulator state detection, it was necessary to take some pretreatments for the local insulator image like angle correction, dilation operation and contrast adjustment. Then the state of insulator tablet deletion and foreign body exists was identified and located with the minimum value of gray statistics and the band width of the gray statistical curves as detection criterions. Lastly, the paper compared the recognition effect and obtained the detection criterions of minimum value of gray statistics.In the rotary double-ear state detection, the local images needed to be detected and positioned accurately with the circle and rectangular detection of Hough transform. And ear surface state was initially recognized wiht the gray variance of local image. Then based on the primarily identification, the ear with the crack characteristics was detected further, and it realized the discrimination of ear fracture and ear crack based on the differential values of two kinds of ears’upper boundary bending.The recognition and state detection algorithm were realized by programming under MATLAB, and the images with different angles and exposures were analyzed and processed by the selected algorithm. The experimental results show that the target recognition based on SURF feature detection, the insulator state detection based on the minimum value of gray statistics and rotary double-ear state detection based on differential values of boundary bending have achieved good results and had high higher precision and speed. Using these methods synthetically, it can realize the target recognition and state detection of insulator and rotary double-ear in the support and suspension device of catenary.