An optimized detection model for micro-terrain around transmission lines

Abstract

Abstract Detecting micro-terrain is essential for the effective layout and maintenance of transmission lines. To address the issues of detection incompleteness, classification ambiguity, and inefficiency in traditional methods, particularly the challenge of distinguishing between saddle and canyon micro-terrain, this paper optimizes the calculation of micro-terrain features and the strategy of micro-terrain detection, and explores a detection method of micro-terrain around transmission lines based on the GPU parallel random forest. This paper employs the GPU parallel random forest model as the extraction framework, leveraging the computational speed advantage of GPU parallel technology for handling large datasets and the robustness inherent in the ensemble approach of random forests. The DEM data of 49 transmission lines in the study area was used for micro-terrain detection experiments. Most of these 49 routes are situated in mountainous regions with complex terrain and contain diverse micro-terrain categories along their paths, rendering them highly representative. The experimental results demonstrate that the proposed method effectively identifies atypical micro-terrain types and four typical micro-terrain types—saddle, canyon, alpine watershed, and uplift—with a classification accuracy of 97.96% and a Kappa coefficient of 0.974. Compared to the traditional method, which achieves a classification accuracy of 75.19% and a Kappa coefficient of 0.642, the proposed method demonstrates a clear improvement in performance. Moreover, by employing the parallel model, the acceleration ratios for training and classification reach 50.57 and 109.06, respectively, significantly improving the efficiency of micro-terrain detection for large-scale regions. These findings could significantly enhance transmission line maintenance and layout planning by providing more accurate micro-terrain data, enabling better decision-making and resource allocation for infrastructure development and disaster risk mitigation.