Infrared Target Detection of High Voltage Insulation Bushing Based on Textural Features

In infrared image target detection based on the traditional image segmentation method, when the background color and the color of the detected object are similar, it is often difficult to identify the detected object effectively in the infrared image. Therefore, to further improve the recognition accuracy of insulating bushings in infrared images, this paper proposes a target detection method based on the texture features of insulation bushings. First, to enhance the texture of the image, bilateral filtering is used to replace the Gaussian convolution filtering in the traditional Laplacian of Gaussian, and image filtering and enhancement are performed through Laplace of bilateral filtering. Then, based on the special texture of the outer sheds and insulation bushing, a descriptor reflecting the periodic distribution of sheds was established and rough identification was performed using the image scanning method. Finally, based on the DBSCAN clustering algorithm, a method for solving its hyper parameters was established to achieve outlier elimination and feature clustering, and to complete the fine identification of the high-voltage insulation bushing. By experimentally comparing other recognition algorithms for infrared images of insulating bushings, the algorithm in this study can effectively segment the insulation bushing main body and overcome the shortcomings of traditional image segmentation methods. The recognition rate on the dataset reached over 85%.