Research on Intelligent Cognition Method of Missing status of Pins Based on attention mechanism

Abstract

The state of the pins on the transmission tower has an important impact on the safety of the transmission line. However, the size of the pin itself is small, the characteristics of different angles vary greatly, and the surrounding environment may be complicated. The traditional manual inspection method is no longer suitable for the increasingly large transmission network. In response to the above problems, this paper proposes a detection method for bolt bolts of transmission towers based on the attention mechanism. First, preprocess the input pin image, and then scan the image with a perturbation neural network that includes an attention mechanism to get the most likely area, and cut and save the area for further processing. This can reduce The useless information on the image increases the proportion of the pins on the image. After multiple rounds of focusing, a pin image with obvious characteristics can be obtained. Secondly, the deconvolutional perturbation neural network is used to establish the feature mapping relationship from the whole to the partial pin image. Thirdly, according to the closed-loop control principle, the distinguishability evaluation index is used to obtain the difference of the extracted features in the feature space. According to the evaluation results, the feature space is layered, the images with large feature deviations are separated and recombined into a new data set, which is then used as the training set of the next level model to retrain the new model, and then based on the evaluation The result of the decision determines whether to continue to layer the training set, realize the multi-level feature space separation and self-optimization adjustment mechanism of the recognition structure, and establish a multi-level model recognition system with stronger generalization ability.