Research On Signature Quantization Of Key Components’ Defect Of Flying Robot Inspection Image

Flying Robot with high-definition cameras has been applied to the electrical inspection. The robots collect image information of tower and tower’s components periodically and send results to ground station which extracts the inspection object from these images. This method is a powerful support for the defect diagnosis. However, due to the speciality of the location of the transmission line, these images contain complicated background and high-dimensional data information, which brings a huge challenges to the defect extraction from inspection objects rapid and accurate, and the accuracy of extraction is meaningful to the following image processing.For this reason, the thesis proposes a more effective method to extract defect signature from inspection images. This method is composed of three important parts. Firstly, on the image preprocessing phase, a new index based on the number of gradation conversion function of image contrast enhancement method is mentioned; Secondly, an improved spatial clustering inspection image density algorithm is depicted which combines the SLIC algorithm, DBSCAN space algorithm, and slope distance method. Compared to the traditional way, this algorithm is more effective and accurate to determine whether there is a defect signature in inspection image by separating object from background and keeping the edges intact and continuous. Finally, a method to find the flaws of inspection image is proposed. This method solves the problem of defect detection of an image by representing the image into over-complete dictionary sparse matrix. And a sparse evaluation function to calculate the sparse degree which could be used to detect the defect signature in inspection image is presented. And it is shown how to detect and quantize defect signature of the inspection image by using the theory of low rank matrix and sparse matrix. Experiment shows that the method this paper proposed can extract and quantize the defect signature in inspection image more effectively and rapidly than traditional way.