Safety Rope State Detection in Electricity Field Construction

In the high-altitude operation in electric power field construction, safety rope abidance is an important prevention to practitioners. It is an intensive enforcement of on-site inspection as well. In recent years, computer vision technology has been applied to industrial safety supervision and has become a research hotspot. However, in the scenario of high-altitude operation in an electric construction field, cameras are generally at a far distance from practitioners and the similarity in the appearance of safety rope and cable poses the challenge to target detection. This paper proposes a novel algorithm for safety rope state check. Firstly, horizontal and vertical poles are identified in the image as the associated target that safety rope should be locked on; secondly, the mobile object detection and Canny edge detection are carried out through the Gaussian Mixture Model to localize practitioners as candidate locations of horizontal/vertical poles; Eight-Neighbor algorithm is introduced to extract lines from an image, and folding times, length and curvature are serving as auxiliaries for precise recognition. Finally, safety rope abidance is determinate according to the spatial position of horizontal/vertical poles and the safety rope. The efficiency of our method is evaluated and analyzed in carefully designed experiments with on-site practical data from a field aerial work video data set.