|
The geographical environments the lines go through are more complex,which makes it necessary to use helicopter when patrol the lines to meet the need of the electrical development.The electric overhead transmission line are made of ground wires,insulators,hardware fittings,towers,foundations,grounding devices and so on, which had severe corrosion damage in the variety of harsh environment.Because it operated in the field long times,the safe operation would have an important impact.The rust defection is an important aspect.In recent years with the constant development of computer technology,the demanding in intelligent inspection by the helicopter have become higher and higher. The defect recognition in transmission line based on image processing technique is development trend of polling by helicopter.In this paper,it use the defect image which was provided by transmission line group to view of rusty defection,analysis of rust characteristics and backgrounds,make use of the rust characteristics under different spatial images,then analysis of image features,determine whether there is rust.For the complexity of aerial images and the characteristics of rust,it make the color image of existence of rust into gray image,contrasting of different methods of image gray processing,it is better that using Excess Red method to gray-scale image.For the error in gray image segmentation,using least squares fitting to fit the gray-scale images,the scope of rusty threshold is determined,which bound the Otsu threshold method,then the gray-scale image is binary processed,it make morphology processing to the segmentation image,extracting of geometric characteristics of the segmentation region, remove the non-rust region,the rusty regions marked in red on the original color image, it recognize the rusty components by the geometric characteristics of rusty regions.The rusty recognition system is implementation.Analysis of the identification of rusty defection and the efficiency of algorithm,components identification accuracy is more than 75%,the running time of algorithms is about 11 seconds.
|