Automatic power line extraction from high resolution remote sensing imagery based on an improved Radon transform

Highlights

• An improved Radom transform, Cluster Radon Transform, for weak linear and short segment feature extraction is proposed.

• A distinguish algorithm for power line extraction are developed.

• A scheme for power line detection is constructed.

Abstract

In this paper, we propose a new algorithm for power line identification and extraction from high resolution remote sensing images. Theoretically, it is difficult to detect power lines in satellite images due to some characteristics, such as sub-pixel, weak target, discrete and the complicated background. To our knowledge, the problem of extraction of the power lines from satellite images is faced for the first time. An improved Radon transform, Cluster Radon Transform (CRT), was developed to extract linear feature from satellite image. Compared with conventional Radon transform, CRT can efficiently avoid false alarm. After that, a set of rules of power lines was abstracted to distinguish power lines from other linear feature, such as roads. The experimental results show that CRT not only has strong anti-noise capability to random noise, but also has strong anti-noise capability to system noise caused by non-linear feature. Furthermore, CRT also has the strong capability to detect short segment in an image. Finally, synthetic images and true images were used to verify the new approach. The achievement has potential to be applicable not only to the power line extraction, but also to other weak linear target detection.

Introduction

The ultrahigh voltage transmission lines play a key role in power transmission, and inspection management of them is a basic work to ensure the safety of power line and equipment. At present, in China and most other countries, the power line inspection work mainly depend on manual records, which have many disadvantages such as high cost, dangerous working condition and inspection missing. In recent two decades, aerial inspection was employed [1], [2], which can improve detection efficiency and precision greatly. But the method is restricted by flight safety, airline control, weather change and refueling. The most important problem of safety requires not only many security measures, but also well trained pilot. Therefore, the helicopter method is not generalized because of its high cost. Unmanned aerial vehicle (UAV) is also used to inspect power line recently [3]. But the safety issue and shortage of endurance restrain its further application.

As space technology developed rapidly in the past ten years, the ability to obtain spatial surface information has improved greatly. The spatial resolution of many commercial satellite had already achieved to sub-meter, and the revisit cycle had also shortened to one day. The most commonly used commercial satellites with high resolution include QuickBird, GeoEye-1 and Worldview, and the highest resolution has reached 0.41 m. These make remote sensing technology apply to power line inspection possible.

Usually, it is an important task to extract linear features from remotely sensed imagery. There are many linear targets in imagery, such as roads, railroads, streams, coastline, etc. Because of significant of roads and the need of GIS updates, scientists and researchers pay more attention to road extraction [4]. Research into automated road extraction from imagery dates back to 1980s in last the century. Tavakoli [5] utilized edge detection, gray level and geometric information to extract building and road from high-resolution aerial photographs. Tupin [6] developed a two-step algorithm to detect road networks from synthetic aperture radar (SAR) image. In his study, a Markov random field and contextual knowledge were employed to identify the real roads from road segment candidates. Baumgartner [7] developed an automatic road extraction method from digital aerial imagery. In his research, roads were regarded as a network of intersections and links between these intersections. In the last two decade, more studies have been focused on road extraction from satellite images. Stoica [8] used a probabilistic model to extract road from SPOT, ERS and aerial images. The model was based on a Gibbs point process framework, and road network was found by minimizing an energy function.

A few researches had also been done to study on power line extraction from aerial SAR or optical image. Sarabandi used a statistical polarimetric detection algorithm to improve the signal-to-clutter ratio, and then utilized the algorithm to detect power line from aerial SAR image. The result was used to prevent collision with power line for helicopters [9]. Blazquez used an airborne Probe Eye Scanner/Normal Color (NC) Video System (PESNVS) to detect the problems in power line manually [10]. Yan developed an automatic method to extract power line from helicopter aerial image based on Radon transform [11]. Deng detected power lines from SAR images based on Hough transform [12]. Recently, scientists began to utilize unmanned aerial vehicles (UAVs) to inspect power lines. Some methods based on Hough transform were used to detect the power lines from UAV optical images [3], [13]. To our knowledge, there is no research on power line extraction from satellite images in literature. This could be due to the image resolution and power line׳s relative fine diameter.

The main contribution of this research is that it solves the problem of extracting very weak linear features from a complex background. In order to handle this issue, we proposed an improved Radon transform, which is called Cluster Radon Transform, to extract linear feature and avoid false alarm caused by complex environment noise. The new method can efficiently erase false alarm. After that, power line knowledge is extracted, which is then used to distinguish power lines from other linear features. In additional, in order to anti-noise, we use Curvelet transform before edge detection to enhance the image at first. It is proved to be useful.

The remainder of the paper is outlined as follows. Section 2 briefly introduces the theoretical basis of sub-pixel object, which proved the feasibility of power line extraction from satellite images. In Section 3 the Radon transform based detection algorithm for linear feature extraction is introduced, and our proposed approach is described in detail. Section 4 presents the rules of power lines, and the algorithm to distinguish power lines from other linear features. Section 5 presents the experiments and discusses the experimental results. In Section 6 conclusions are drawn and further research perspective was addressed.