Aerostat main cable surface defect detection method and system based on small sample learning

Abstract

The invention provides an aerostat main cable surface defect detection method and system based on small sample learning. The method comprises the following steps: setting a system software and hardware environment; aerostat main cable surface images are collected and processed through preset adversarial network expansion, cable surface enhanced image data are obtained and marked, and a defect sample library is constructed according to the cable surface enhanced image data; designing a network model according to a DenseNet network and main cable defect characteristics, constructing the network model by using small sample learning, and training the network model according to a defect sample library; processing the query set in the defect sample library by using the trained network model and using a yolov4 single-stage detection algorithm so as to obtain shallow texture features and high-level semantic features; the structure of a DenseNet network of a network model is utilized, shallow texture features are transmitted to high-level semantic features through jump connection, and the high-level semantic features are processed through metric logic; and selecting and acquiring surface defect detection data of the main cable of the aerostat in different modes at the terminal. The technical problems of high defect omission ratio, high false detection rate and low detection efficiency are solved.

Description

technical field

[0001] The invention relates to a cable defect detection technology of an aerostat, in particular to a method and system for detecting surface defects of a main cable of an aerostat based on small sample learning. Background technique

[0002] The outer sheath on the surface of the cable is a structural part that prevents external factors from eroding the cable insulation layer. Its main function is to improve the mechanical strength of the cable and prevent chemical corrosion, water, and combustion. However, in the production process of cables, surface defects such as creases, scratches, small holes, bulges, and damaged insulation will inevitably occur due to factors such as processing equipment, production technology, and production materials. These defects not only damage product performance and affect commercial use, but also serious apparent quality may even cause safety hazards in later use.

[0003] The invention patent with the application number CN2...

Examples

Embodiment 1

[0068] like figure 1 As shown, a small-sample learning-based surface defect detection system for the main cable of the aerostat is provided in the following steps:

[0069] S1. Set the software and hardware operating environment according to the demand analysis data. Optionally, based on the requirements of the main cable defect detection: conduct relevant operating environment research, realize system function decomposition, and clarify system development content;

[0070] S2. Design image acquisition equipment. Optionally, in the hardware design stage, design cable image acquisition facilities according to the environment in which the main cable operates and the cable surface conditions;

[0071] S3, collecting images, optional, image collection and preprocessing: using collection equipment for image collection;

[0072] S4, establishing a sample library;

[0073] S5. Enhance the image. Optionally, complete the necessary image preprocessing work, and use the image enhancem...

Embodiment 2

[0080] like figure 2 It can be seen from the figure that the method of small sample metric learning mainly consists of feature encoding module and metric module. The support set in the figure indicates that the sample category in the data set is known to us, and the query set indicates that the samples in the data set belong to the samples to be tested.

[0081] Sample Data Augmentation

[0082] like image 3 As shown, in view of the challenges of unbalanced defect samples, difficulty in manual labeling, and diverse on-site environment changes, this project not only uses image enhancement to expand the diversity of existing samples, but also expands samples by building an adversarial neural network. In the deep neural network In order to further alleviate the impact of category imbalance during training, focal loss and label smoothing strategies are further used.

[0083] The measurement method used in this paper can be expressed by the following formula:

[0084]

[0...

technical field

[0001] The invention relates to a cable defect detection technology of an aerostat, in particular to a method and system for detecting surface defects of a main cable of an aerostat based on small sample learning. Background technique

[0002] The outer sheath on the surface of the cable is a structural part that prevents external factors from eroding the cable insulation layer. Its main function is to improve the mechanical strength of the cable and prevent chemical corrosion, water, and combustion. However, in the production process of cables, surface defects such as creases, scratches, small holes, bulges, and damaged insulation will inevitably occur due to factors such as processing equipment, production technology, and production materials. These defects not only damage product performance and affect commercial use, but also serious apparent quality may even cause safety hazards in later use.

[0003] The invention patent with the application number CN2...

Examples

Embodiment 1

[0068] like figure 1 As shown, a small-sample learning-based surface defect detection system for the main cable of the aerostat is provided in the following steps:

[0069] S1. Set the software and hardware operating environment according to the demand analysis data. Optionally, based on the requirements of the main cable defect detection: conduct relevant operating environment research, realize system function decomposition, and clarify system development content;

[0070] S2. Design image acquisition equipment. Optionally, in the hardware design stage, design cable image acquisition facilities according to the environment in which the main cable operates and the cable surface conditions;

[0071] S3, collecting images, optional, image collection and preprocessing: using collection equipment for image collection;

[0072] S4, establishing a sample library;

[0073] S5. Enhance the image. Optionally, complete the necessary image preprocessing work, and use the image enhancem...

Embodiment 2

[0080] like figure 2 It can be seen from the figure that the method of small sample metric learning mainly consists of feature encoding module and metric module. The support set in the figure indicates that the sample category in the data set is known to us, and the query set indicates that the samples in the data set belong to the samples to be tested.

[0081] Sample Data Augmentation

[0082] like image 3 As shown, in view of the challenges of unbalanced defect samples, difficulty in manual labeling, and diverse on-site environment changes, this project not only uses image enhancement to expand the diversity of existing samples, but also expands samples by building an adversarial neural network. In the deep neural network In order to further alleviate the impact of category imbalance during training, focal loss and label smoothing strategies are further used.

[0083] The measurement method used in this paper can be expressed by the following formula:

[0084]

[0...

Images