Insulator location for low-light-level image based on SVM

Aiming at the difficulty of a single method to detect the insulator edges from the colour low light level image with unclear salient features and low signal-to-noise ratio under low illumination. A new method which combines Histogram of directional gradient and other algorithms is designed. Firstly, in order to evenly distribute the light and shade of the image and enhance image texture characteristics, a three channel histogram equalization algorithm is used to extract the insulator outline, at the same time, Hough algorithm is used to correct the tilt of the image. Secondly, the directional gradient features of positive and negative samples are computed respectively, and PCA algorithm is used to reduce the dimension of high-dimensional sample eigenvector. Thirdly, the support vector machine is trained with the feature vectors, and using DCWPSO algorithm optimized parameters. Finally, the slider box is used to find the area, which was predicted to the insulator target. Sklearn’s SVM tool library was used in this research, and C-SVM tool box was selected. The experimental results show that the method is robust and accurate, it is also suitable for insulator detection under various lighting conditions. The support vector machine can reach 92.4% recognition accuracy under the condition of small sample training, compared with HOG-SVM algorithm, it is improved by 6.3%.