Safety Monitoring Method for Live Working Based on Deep Learning and Binocular Vision

As the power industry continuously raises safety requirements for live-line working, traditional safety monitoring methods can no longer meet the demands for real-time and precise monitoring. This paper proposes a safety monitoring method for live-line operations based on deep learning and binocular vision technology, aiming to enhance safety at the operation site through intelligent means. The method uses binocular vision to obtain two-dimensional and three-dimensional image data of the site, and generates accurate spatial information through image registration and depth computation. By training a deep neural network model, the system can automatically recognize various behaviors of workers, such as operation posture, distance assessment, and hazard area identification, providing real-time feedback on relevant safety risks. Experimental results show that the proposed safety monitoring method can maintain high-precision detection capabilities in complex live-line operation scenarios, with strong adaptability and efficiency. It provides technical support for safety management in live-line operations in the power industry and has broad application prospects.