Study On Pantantograph Porcelain Insulator Location Detection Methods Based On Learning Algorithm

The pantograph porcelain insulator of the electronic locomotive is a key component that ensures the insulation of the electronic locomotive roof and the safety of locomotive running. Location detection of the pantograph porcelain insulator is an important part of locomotive detection. Nowadays manual method is mostly used to detect the pantograph porcelain insulator. It is extremely arduous and has low detection efficiency and the existence of security risks. Therefore, it is necessary to provide a new detection method to achieve high accuracy of location detection and greatly improve working efficiency.This paper mainly studies the location detection algorithm of the pantograph porcelain insulator. Considering the diversity of the pantograph porcelain insulator, BP neural net-work algorithm and Adaboost algorithm that are both based on the training and learning are discussed for the location detection of the different pantograph porcelain insulator.In the process of the pantograph porcelain insulator locating detection by the BP neural net-work, this paper firstly makes use of the algorithm of K-means clustering image segmentation to segment the pantograph porcelain insulator based on its color feature, then trains the sample features composed by seven shape parameters of the insulator by way of the BP neural net-work algorithm, lastly, successfully detects several insulators in one image and improves the accuracy of location detection comparing to the algorithm based on visual analyze and gradient feature.In order to obtain the image of the pantograph porcelain insulator from the running locomotive at high speed that is up to 160km/h, the image acquisition system of high-speed locomotive pantograph porcelain insulator is built. The images acquired by the system are dealt with by the algorithm of location detection based on Adaboost. The experimental results show that this algorithm improves the stability and adaptability comparing to the algorithm of location detection based on template matching...