An image recognition method for bird species related to bird-related faults in transmission lines

Abstract

The invention discloses a method for recognizing images of bird species related to bird-related faults in transmission lines. Firstly, by collecting bird species information around the transmission line, an image database of bird species related to bird-related faults is established, and the bird species images are analyzed by a method based on a category activation map. The background preprocessing is removed; then, four kinds of deep convolutional neural networks are used to establish a learning model, and they are pre-trained through the ImageNet dataset, the network structure of the pre-trained model is fine-tuned, and the pre-processed bird species images are used for training. The fine-tuned model is retrained on the set, and the test set is classified and identified. Finally, according to the classification accuracy of the four network models, a linear weighting method is used to establish a bird species image related to bird wading faults fused with multi-convolution networks. Recognition model to classify and recognize bird images. The invention can provide the transmission line operation and maintenance personnel with methods and means for correct bird recognition, helps to realize the differentiated prevention and control of the bird's fault, and reduces the trip rate of the bird's fault.

Description

technical field [0001] The invention relates to the field of power transmission lines, in particular to a method for identifying images of bird species related to bird-related faults of power transmission lines. Background technique [0002] With the large-scale construction of the power grid and the improvement of the ecological environment, the contradiction between bird activities and transmission lines has become increasingly prominent, and the bird fault has become an important reason for line tripping. Due to the sudden nature of bird failures, it is often impossible to determine what kind of bird caused the failure after the failure, and it is difficult to install anti-bird measures in a targeted manner. Although the power grid operation unit has counted the list of bird species related to the transmission line fault and the types of faults it may cause, due to the lack of knowledge of ornithology by the operation and maintenance personnel, although the line patrol pr...

AI Technical Summary

Problems solved by technology

[0003] At present, the research on image recognition of bird species related to transmission lines is mainly limited to the coarse-grained binary classification problem of bird detection, and there are few studies on multi-classification recognition of bird species that are endangered by bird faults on transmission lines. The number of bird images collected is limited, and it is easy to have insufficient generalization ability when used to train a brand new network, and the network trained by using a specific bird image data set is difficult to match the bird image recognition related to bird-related faults task

Images