Detecção de obstruções na faixa de domínio em linhas de transmissão de energia elétrica usando a técnica de classificação hierárquica

Abstract

Currently Remote Sensing has varied methodologies that enable acquire quickly and need information about the surface. One technique is the use of digital mapping technology LASER or LIDAR (Light Detection And Ranging) as it is also known. This technique has been highlighted as an option for low cost and high productivity when compared with other more conventional methods of acquiring three-dimensional data. The Brazilian energy matrix in the last year reached the mark of 121,100 megawatts produced, these being from generating units in 2809, with 100,000 miles of transmission lines, which is expected to grow in the coming years, thus highlighting the need for techniques that allowing the data acquisition to accurately and efficiently. The methodology consists of the use of supervised classification techniques through segmentation models and pixel by pixel by integrating spatial and spectral data of an orthophoto with altimetry data from the Laser Scanner. Analyzing the results is noted that the data Laser System contributed to both methods of classification, but the classification technique hierarchical data presented above when compared with the traditional method (pixel by pixel). Finally there is an improvement in classification results after implementation of the data Laser System, and was obtained kappa index of 0.951 and an overall accuracy of 95.80% for the best result for classification.