The research on binocular ranging technology for transmission lines based on two - dimensional line matching

Using binocular distance measurement technology to obtain the distance between the dangerous object and the transmission line, which provides powerful data information for the judgment of the risk degree of external damage and the assessment of the probability of the line suffering from external damage, so as to ensure the safety of the line. In order to solve the problem of large epipolar error caused by serious image distortion under wide-angle lens, local two-dimensional line matching scanning technology, which based on the object position information obtained by target detection, reduced the influence of epipolar error and increase the robustness of matching. At the same time, the dense calculation of global image distance is degraded to the local dense calculation of target, which improves the real-time performance of binocular ranging algorithm in high-resolution monitoring image. Finally, the image segmentation technology is used to improve the mismatching problem at the depth discontinuity to further improve the accuracy of binocular ranging.

I. Introduction

With the development of social economy, the total mileage of transmission lines in China is increasing. By 2018, the total length of 35kV and above transmission lines in China has reached 1.892 million kilometers, ranking first in the world. However, with the increase of transmission line length, the risk of external damage is also increasing, especially in some “three span” environment. Once the transmission line accident occurs, it will have a significant impact on the surrounding systems and economy. In recent ten years, with the development of artificial intelligence technology, more and more equipment equipped with real-time monitoring and image recognition technology appear in the transmission line monitoring system [1], [2]. Through the real-time acquisition of transmission line surrounding images and video information, artificial intelligence technology realizes the early warning of external damage [3], [4]. However, artificial intelligence technology only provide target category and position information on the image, and can not further obtain important information such as the distance between dangerous objects and transmission lines. Therefore, it needs to be realized with the help of multi domain technology of comprehensive image recognition.