Detection of Bird's Nest in High Power Lines in the Vicinity of Remote Campus Based on Combination Features and Cascade Classifier

Abstract

High-voltage transmission towers are built to supply electricity for campuses and local residents. In order to guarantee the power supply and safety for remote campus, the unmanned aerial vehicles (UAVs) are used to take the images of high power lines to alarm potential malfunction. A novel method of bird's nest images detection based on cascade classifier and combination features is proposed. Different features of the bird's nest and iron tower are analyzed, and the following four novel features are proposed: proportion of white area (PWA), ratio of white pixels (RWP) in each lap, projection feature (PF), and improved burr feature (IBF). The combined features are used to describe the characteristics of the bird's nest backbone area and the edges, respectively. Furthermore, the cascade classifier combined with the four proposed features is used for the further classification of bird's nest region. The proposed detection process mainly consists of three stages. First, the suspected bird's nest region is obtained by template convolution. Second, PWA and RWP with low dimensionality and high discrimination are used to classify the sample set of suspected nest region. Third, based on the previous classification results with positive and negative samples, PF and IBF are adopted to further conduct the secondary classification in order to reduce the misclassified samples, and the final classification label is determined by the second classification results. Experimental results show that the proposed algorithm can accurately detect the nest and achieve good performance.