Power transmission line foreign matter and environment abnormal state detection method

Abstract

The invention discloses a transmission line foreign matter and environment abnormal state detection method. According to the method, aiming at various conditions, such as various suspended foreign matters, bird nests and the like, which may cause abnormity of the power transmission line, on the power transmission line, effective detection of the foreign matters and abnormal states of the power transmission line is realized by embedding a Coordinate Attention module on a fine-tuned YOLOv4 model, so that the accuracy of target detection is ensured, and the speed of model training test is also ensured. The model fine tuning enables the network to be more efficient when detecting the foreign matter and abnormal state data set of the power transmission line, and the embedding of the Coordinate Attention module helps the model to pay more attention to the target feature area when extracting the target feature, so that the final output result is more accurate.

Description

Technical field [0001] The present application relates to the field of image recognition, and more particularly to a transmission line foreign body and an environmental abnormal state detection method. Background technique [0002] my country's power grid is the world's largest interoperable Internet power system. With the construction of the grid, the scale of transmission lines is expanding, and the challenges are constantly upgrading. Common foreign objects including the transmission line include a Bird's Nest, a kite, a kite, a plastic bag. These foreign objects can cause short circuits and trigger a hill, smoke, so that the power system has failed, bringing huge gab to the country and people. Economic loss and safety hazards, so it is important to detect and clean up foreign affixions. Since the distribution of transmission lines is wide, it is often distributed in some people's smoke, environmentally complex areas, and the traditional artificial inspection work is low, and ...

AI Technical Summary

Problems solved by technology

At present, the two-stage algorithm with good effect is to use the generated candidate target area to locate and classify the target, such as Fast RCNN, Faster RCNN, RFCN, etc., but generating candidate areas will consume a lot of computing resources and reduce the detection speed, which is not conducive to Real-time monitoring of transmission lines

Images