CN-122023363-A - Crimping defect detection method based on improved target detection model, computer equipment and storage medium

CN122023363ACN 122023363 ACN122023363 ACN 122023363ACN-122023363-A

Abstract

The invention relates to the technical field of industrial detection, and discloses a crimping defect detection method, computer equipment and storage medium based on an improved target detection model, wherein the extraction capability of the model to multiscale defect characteristics (such as tiny gaps and uneven indentation) of a crimping terminal is obviously enhanced by introducing a C2f-LSK module, and the spatial positioning precision and channel characteristic expression efficiency of the model to a key area are effectively improved by combining a CA and ECA attention mechanism, so that the detection precision and environmental robustness of the crimping defect are greatly improved on the basis of keeping the original real-time detection advantage of YOLOv, the problems of strong subjectivity of traditional manual detection and high false leakage detection rate of the traditional visual method are effectively solved, the quick, accurate and automatic evaluation of the crimping quality of the cable terminal can be realized, potential safety hazards such as virtual connection, loosening and the like are timely discovered, the running fault risk of a power grid is obviously reduced, and the reliability and safety of a power system are improved.

Inventors

YANG YUFU
WU JIANJIE
LIU GUOBING
HE XIAOJING

Assignees

广东电网有限责任公司东莞供电局

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (10)

1. A method of crimp defect detection based on an improved target detection model, the method comprising: Acquiring an image to be detected after crimping of a cable terminal; The method comprises the steps of inputting an image to be detected into an improved target detection model to detect the crimping defect of a cable terminal in the image to be detected, wherein the improved target detection model is based on a YOLOv model, replacing a third C2f module and a fourth C2f module in a backbone network of the YOLOv model with C2f-LSK modules, fusing a CA attention mechanism with the SPPF module in the backbone network to form an SPPF+CA module, replacing a C2f module between a Concat module and a Upsample module in a feature fusion network with a C2f-LSK module, and replacing a C2f module between a Concat module and a Conv module in the feature fusion network with a C2f-LSK module and fusing an ECA attention mechanism to form a C2f-LSK+ECA module.
2. The improved target detection model-based crimp defect detection method of claim 1, further comprising: Constructing an improved target detection model according to the YOLOv model, the C2f-LSK module, the CA attention mechanism and the ECA attention mechanism; And training the improved target detection model to obtain the trained improved target detection model.
3. The crimping defect detection method based on the improved target detection model according to claim 2, wherein the step of constructing the improved target detection model according to YOLOv model, C2f-LSK module, CA attention mechanism, and ECA attention mechanism is: based on YOLOv model, replacing the third C2f module and the fourth C2f module in the backbone network of YOLOv model with C2f-LSK modules; Merging the SPPF module in the backbone network with a CA attention mechanism to form an SPPF+CA module; Replacing a C2f module between Concat modules and Upsample modules in the feature fusion network with a C2f-LSK module; And replacing a C2f module between the Concat module and the Conv module in the feature fusion network with a C2f-LSK module and fusing an ECA attention mechanism to form a C2f-LSK+ECA module.
4. The method of claim 2, wherein the step of training the improved target detection model to obtain a trained improved target detection model comprises: Acquiring cable terminal images under different crimping states, shooting angles and illumination conditions, marking crimping areas and crimping defects in the cable terminal images, and constructing a crimping detection data set; preprocessing the crimping detection data set; inputting the processed crimping detection data set into an improved target detection model for training, and obtaining the trained improved target detection model.
5. The improved target detection model based crimp defect detection method of claim 4, wherein the preprocessing comprises image size normalization and data enhancement.
6. The improved target detection model based crimp defect detection method of claim 5, wherein the data enhancements comprise random flipping, random scaling, mosaic enhancement, and MixUp enhancement.
7. The method for detecting crimp defects based on an improved target detection model according to claim 4, wherein in the training process, parameters of each layer of the network are updated through a back propagation algorithm based on a loss function, iterative updating is performed by an optimizer, and convergence speed and accuracy of the model are improved through a learning rate adjustment strategy.
8. The improved target detection model-based crimp defect detection method of claim 1, wherein the YOLOv model's detection head network employs a joint detection head structure comprising three parallel branches, namely a region localization branch, a defect classification branch, and a confidence branch.
9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor, when executing the computer program, implements the improved target detection model-based crimp defect detection method of any one of claims 1-8.
10. A computer-readable storage medium having stored thereon computer-executable instructions for execution by a computer processor to implement the improved target detection model-based crimp defect detection method of any of claims 1-8.

Description

Crimping defect detection method based on improved target detection model, computer equipment and storage medium Technical Field The present invention relates to the field of industrial detection technologies, and in particular, to a crimping defect detection method, a computer device, and a storage medium based on an improved target detection model. Background The electric wires and cables are used as 'nerves' and 'blood vessels' of national economy, and are basic materials for maintaining energy, industry and urban operation. The cable industry is not only an important matching industry for the healthy development of national economy, but also a key field of the first two in the mechanical and electrical industry of China, and occupies a significant position in the whole national economy system. Among the cable connection systems, crimping has been attracting attention due to its wide application, and particularly plays an indispensable role in large-scale power grid engineering. The traditional crimping operation relies mainly on hydraulic pliers, which work on the principle of placing the cable and terminal in the jaws of the hydraulic pliers and applying pressure, so that a firm and reliable connection is formed. The hydraulic pliers have the remarkable advantages of simple process, small influence by cable size and materials, simplicity and easiness in operation, high efficiency, good repeatability, high quality, wide application range and the like. However, in practical application, the hydraulic clamp crimping process also has some defects that factors such as experience difference of operators, uneven force application or unstable environment are easy to cause uneven indentation, virtual connection, loose terminal and the like, the hydraulic clamp jaws are usually polygonal dies, so that the terminal and the cable cannot be tightly attached after crimping, gaps exist, further accidents such as tip discharge, heating and even firing can be caused, and under a long-term vibration or high-load operation environment, the terminal is gradually loosened and separated from the cable due to the existence of the gaps, electrical faults are caused, and reliable operation of a power grid is seriously affected. Therefore, it is important to detect the crimp quality of the cable terminal. However, the existing detection method is based on manual detection or traditional machine vision technology, and has the problems of low detection precision, poor robustness, insufficient instantaneity and the like in the aspect of crimping defect detection. The above information is presented as background information only to aid in the understanding of the invention and is not intended to determine or acknowledge whether any of the foregoing is useful as prior art with respect to the invention. Disclosure of Invention The invention provides a crimping defect detection method, computer equipment and storage medium based on an improved target detection model, which are used for solving the problems in the prior art. In order to achieve the above object, the present invention provides the following technical solutions: In a first aspect, the present invention provides a method for detecting crimp defects based on an improved target detection model, the method comprising: Acquiring an image to be detected after crimping of a cable terminal; The method comprises the steps of inputting an image to be detected into an improved target detection model to detect the crimping defect of a cable terminal in the image to be detected, wherein the improved target detection model is based on a YOLOv model, replacing a third C2f module and a fourth C2f module in a backbone network of the YOLOv model with C2f-LSK modules, fusing a CA attention mechanism with the SPPF module in the backbone network to form an SPPF+CA module, replacing a C2f module between a Concat module and a Upsample module in a feature fusion network with a C2f-LSK module, and replacing a C2f module between a Concat module and a Conv module in the feature fusion network with a C2f-LSK module and fusing an ECA attention mechanism to form a C2f-LSK+ECA module. Further, the crimping defect detection method based on the improved target detection model further comprises the following steps: Constructing an improved target detection model according to the YOLOv model, the C2f-LSK module, the CA attention mechanism and the ECA attention mechanism; And training the improved target detection model to obtain the trained improved target detection model. Further, in the crimping defect detection method based on the improved target detection model, the steps of constructing the improved target detection model according to the YOLOv model, the C2f-LSK module, the CA attention mechanism and the ECA attention mechanism are as follows: based on YOLOv model, replacing the third C2f module and the fourth C2f module in the backbone network of YOLOv model with C2f-LSK modules; Merging the SPPF m