A method of insulator detection from video sequence

We present a very efficient approach for insulators detection. Unlike previous texture-based approach, we directly search insulators location in the images by using Profile projection. For overcoming the negative effect of image noise on object detection, we preprocess insulators image by thresholding method. To make insulators detection more effective and efficient, we design a tilt correction method based on principal component analysis. The correction enables our method to derive accurate feature extraction curve from insulators image, then we extract five features from the feature curve, which are related to the number of binary sequence and the normalized variance of the binary sequence length. After obtaining the insulators feature from an image, we apply SVM to identify insulators with the five features. Some experiments on video frame images show that our approach significantly outperforms the state-of-the-art in term of both accuracy and efficiency