Bolt two-dimensional visual structure clustering method based on morphological optimization depth features

Abstract

The invention discloses a bolt two-dimensional visual structure clustering method based on morphological optimization depth features. The method comprises the following steps: S1, extracting image sample depth features in a data set by adopting a pre-trained deep convolutional neural network model, and calculating a morphological ratio of each target image; s2, combining the morphological ratio obtained in the step S1 with the image sample depth feature obtained in the step S1 to obtain a morphological optimization depth feature; s3, calculating the aspect ratio distribution trend of each image sample in the data set so as to obtain the number of clustering centers; and S4, dividing the image samples into clusters based on the morphological optimization depth features obtained in the stepS2 and the clustering center obtained in the step S3, and performing cluster optimization selection on the target image samples through a minimum Euclidean distance principle so as to obtain a clustering result of the two-dimensional visual structure of the bolt. By the adoption of the bolt two-dimensional visual structure clustering method based on the morphological optimization depth characteristics, the problem of clustering analysis of various visual structures presented when a three-dimensional entity is represented by a two-dimensional image is solved, a clustering center does not need to be defined manually, and the method has the advantages of being high in accuracy, high in generalization capacity and the like.

Description

technical field [0001] The invention relates to an image analysis technology, in particular to a bolt two-dimensional visual structure clustering method based on shape optimization depth features. Background technique [0002] Bolts are extremely important fasteners that exist in large quantities in transmission lines. They play the role of connecting components and fastening connections on transmission lines. They are fault-prone components, so they need to be periodically repaired. At present, the use of aircraft to inspect lines has become a routine method for inspections of transmission lines. First, during the inspection, the key parts are photographed to obtain a large number of aerial images, and then the clustering of the two-dimensional visual structure of the bolts is realized. Finally, the bolts are automatically identified and diagnosed. , where clustering is an important prerequisite for the automatic identification and diagnosis of bolts in massive aerial image...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a bolt two-dimensional visual structure clustering method based on morphological optimization depth features, which solves the problem of cluster analysis of various visual structures presented when three-dimensional entities are represented by two-dimensional images without artificial Define the cluster center, which has the advantages of high accuracy and strong generalization ability

Images