Representation of binary feature pooling for detection of insulator strings in infrared images

Insulators are the most common equipment in power systems, and the failure of insulators would be the direct threat to the stability and safety of the system. With the advantages of being non-contact and non-destructive, infrared imaging technology is efficient for monitoring and evaluating the thermal condition of insulators. Detecting the position of the insulator string in an infrared image is a crucial step for automatic diagnosis, thus an efficient and robust representation of insulator string is necessary. This paper proposes a novel method for insulator strings detection, and presents an appearance representation for insulator strings in infrared image based on Binary Robust Invariant Scalable Keypoints (BRISK) and Vector of Locally Aggregated Descriptors (VLAD). We name this feature generation method Binary Feature Pooling. A classification model based on Support Vector Machine (SVM) is integrated into multi-scale sliding window framework for locating insulator string in infrared image. Then redundant regions are merged by non-maximum suppression and shape prior knowledge constraints. The classification accuracy of the proposed method is 89.1026% on our standard infrared insulator dataset, and insulator strings under various conditions can be successfully detected. The results show that this method can detect multiple insulator strings in infrared images with low resolution and complex background.