[Data Science Case] Detection of Overhead Wire Anomalies Using Deep Learning

Examples of Data Science Utilization in the Construction and Social Infrastructure Sector! Case Study on Improving Inspection Efficiency

We will introduce a case study of improvement in abnormal detection of overhead wires using deep learning. At Tokyo Electric Power Grid, abnormality detection was performed by humans visually inspecting aerial footage as part of overhead wire inspection and equipment maintenance work. The visual confirmation involved slow playback of the aerial video at one-tenth speed, making the inspection process very costly, and there were also issues with accuracy due to missed detections. To address this, a model was developed using deep learning to determine abnormalities and normal conditions from image data. By quantifying and visualizing the "abnormality" of the overhead wires, AI was able to identify areas that required human visual confirmation, resulting in a reduction in the costs associated with visual inspections.

Image analysis software