Power line detection using spatial contexts for low altitude environmental awareness

Abstract

The environmental awareness of ground obstacles is posed with vital significance to ensure the low altitude flight safety. This paper focuses on the detection of power lines, a kind of hazardous small ground obstacle. Traditional power line detection (PLD) methods mainly utilize the line features of the power lines. These methods face difficulty in complex circumstance where some background objects may also have strong line features. Noticed that the spatial contexts of power lines can provide discriminative features, a novel spatial contexts based PLD method is proposed in this paper. Two kinds of spatial contexts, the pylonline spatial context and the line-line spatial context, are utilized to support the detection. Firstly, a line segment pool is generated to get power line segment candidates as much as possible. Secondly, the pylonline spatial context is utilized to find the power line segments linking with the pylons. Thirdly, to achieve better detection effects, a line-line spatial context based detection algorithm is designed to find more power line segments using the well-identified power line segments. Experimental results show that the proposed method obtains higher detection rates and lower false alarm rates comparing to comparative line feature based PLD methods.