Reduction of Number of Patrols Around Transmission Facility using Deep Learning

One of the purposes of patrols is to check whether construction is taking place near the transmission lines. In rural area, a patrol person cannot find a sign of under construction because there are few constructions, and it is wasteful for people to patrol. Thus, we have attempted to reduce the number of patrols using a surveillance camera system. When using a surveillance camera, it is necessary to confirm the content of the images that are taken. To avoid watching a large number of images, we have attempted to use Deep Learning, for which many data is required for learning. Therefore, we have developed a scheme for reducing the number of confirmation tasks. The scheme creates a combined image which is the sum of all images using a weighting coefficient that is determined by the magnitude of scene change every hour. We confirm the effectiveness of the scheme. Finally, we apply Deep Learning to detect heavy machines using the result of learning based on images gathered by the scheme and confirm that Deep Learning is effective for the detection of heavy machines.