Study On Self-explosion Feature Of Insulator Recognition In Aerial Image

The increasingly development of China’s social needs toward power grid, brings a corresponding increase in electrical inspection tasks. However, the traditional way of human inspection is of great risk and low efficiency as well as high potential security risk. For the past few years, a new way of electrical inspection has come into being, that is the Unmanned Aerial Vehicle electrical inspection, which has been gradually applied and extensively promoted. Under such background, this research aims at studying the identification and detection of the self-explosion characteristic of electrical insulator based on transmission line of aerial imagery. The research into this technology would pave the way for Unmanned Aerial Vehicle electrical inspection. The research provides powerful technical support in the aspect of avoiding obstacle automatically, distance measurement and then realizing intelligent patrolling and checking in the future. The main job this research has done include following four parts:1. This thesis analyzes the current situation of domestic and foreign research.whitch towards the practical needs of engineering under the consideration of the characteristics of multi-rotor UAV electrical inspection. This thesis also analyzes the actual engineering requirement under the consideration of the characteristics of multi-rotor UAV electrical inspection. This study makes a comparison between two recognition algorithms of self-explosion characteristic of electrical insulator, which are respectively based on computer visual and pattern recognition.2. According to the self-explosion feature of electrical in aerial image, one recognition and detection algorithm is proposed based on the related computer vision and image processing technologies. The first step is to make threshold segmentation respectively with a spatial image transforming from LAB and grayscale image by LAB OTSU method. After that, AND Operation is used to segmentations, and then build a mathematic model on the basis of pure insulator string through morphologic processing and area filtering. In addition, mark the start point and center of the insulator string. Finally, make a comparison between white pixels in insulator string and threshold to judge whether the self-explosion feature of electrical insulator existing.3. The effect of insulator’s image segmentation by such segmentation method will be influenced by the complexity of image background. To solve this problem, this research proposes to utilize the recognition algorithm with related pattern recognition technology. On the basis of grayscale image, extract SURF features with invariant scale from the image, and recognize the insulator by applying the SVM(Support Vector Machine) method, and then exclude the mistakenly-recognized outlier which is similar to the insulator by RANSAC(random sample consensus algorithm) method. And the last step is to calculate the proportion of insulator pixels to judge the defect position.4. Test the two proposed methods in the effect of self- explosion, which one is employed in literatures and another is based on independent insulator segmentation. Then, carry out the qualitative and quantitative analyses about the test results and compare the results. In general, this research aims at providing proper guidance for the future researches after analyzing the defects and disadvantages existed in this algorithm according to the test results.