Intelligent identification and monitoring method of substation pyrotechnics based on deep learning

Abstract

The invention discloses a method for intelligent identification and monitoring of pyrotechnics in a substation based on deep learning. By improving and optimizing the video recognition model and the image recognition model for the actual scene of the substation, and then fusing the two improved model frameworks, avoiding the two as much as possible. Based on the disadvantages of the former, give full play to the respective advantages of the two, and design a more reasonable and flexible detection method: usually use the image recognition model for monitoring, after detecting smoke, automatically call the video recognition model for a second re-inspection, after the verification is accurate Then send an alarm signal to the monitoring platform, which can effectively complete the detection and early warning work.

Description

technical field [0001] The invention relates to a method for intelligent identification and monitoring of pyrotechnics in substations based on deep learning, and belongs to the technical field of artificial intelligence monitoring and identification for substation safety. Background technique [0002] Fire has always been one of the great threats to the safety of people's lives and properties. The suddenness, frequency and high destructive power of fires seriously threaten people's lives, properties and the natural environment. The substation is the hub of the power system. Once a fire breaks out, it may cause the collapse of the entire power grid system, seriously endangering the reliability of power supply. Therefore, doing a good job of fire prevention measures in substations is a crucial task to ensure the safe and stable operation of the power grid. The main reasons for fires in substations are as follows: electrical equipment causes fires, and substations contain a la...

AI Technical Summary

Problems solved by technology

The traditional sensor-based smoke detection method is easily affected by the environment, has a certain alarm delay and is not suitable for a wide range of environments such as outdoors

With the rapid development of intelligent video technology, research on smoke detection using image recognition or video recognition technology has begun. Most current applications use images alone for detection or video alone for detection, but these two methods have their limitations. Responsibility: Image recognition takes up less resources but is more likely to generate false positives; video recognition has higher accuracy but takes up too many resources. When network bandwidth resources are tight and multi-channel video real-time transmission is detected, the server load is heavy

Images