Detection of Bird's Nest in Real Time Based on Relation with Electric Pole Using Deep Neural Network

Abstract

In order to guarantee the safety of the electric transmission systems from the bird's nest, it is necessary to detect bird's nest. However, it is difficult for human to monitor the electric power system one by one whether the bird's nest exists or not. Therefore, the detection of the bird's nest is necessary. We suggest the bird's nest detection model in real time with keeping high detection performance. The reason why we use real time detection model is to be able to using pictures taken in the driving car. When we detect the nest through just only nest detector, there exists failure cases that the whole shape of the tree is observed as the nest. In order to deal with this problem, we use both the nest detector and the electric pole detector to revise the detection results. To solve this problem, we construct nest dataset directly.