Visual and semantic fusion fault identification algorithm for power inspection robot
Authors:
- 23 May 2025
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 414-422
Abstract
In the process of power inspection, this paper aims at the problems such as low recognition rate and poor robustness of traditional fault identification methods in complex environments, aiming at improving the intelligence level and fault detection ability of inspection robots. The algorithm makes use of the complementarity of visual information and semantic information, and constructs a multi-modal information fusion framework to achieve accurate extraction and recognition of complex fault features in inspection scenes. Specifically, this paper firstly preprocesses the image data in the inspection scene, and extracts semantic information by combining natural language processing technology. A deep-learning-based fusion model was then developed to effectively combine and analyze visual and semantic features. Experimental results indicate that the algorithm achieves high accuracy in recognition and shows strong robustness in various fault scenarios. This advancement offers significant technical support for enhancing the intelligent fault detection capabilities of power inspection robots, contributing to improved operational reliability and maintenance effectiveness.Keywords
Funding Information
- State Grid Corporation of China
This publication has 13 references indexed in Scilit:
- GPT (Generative Pre-Trained Transformer)— A Comprehensive Review on Enabling Technologies, Potential Applications, Emerging Challenges, and Future DirectionsIEEE Access, 2024
- Natural language processing: state of the art, current trends and challengesMultimedia Tools and Applications, 2022
- Attention mechanism in neural networks: where it comes and where it goesNeural Computing and Applications, 2022
- A novel image processing algorithm to separate linearly clustered kiwifruitsBiosystems Engineering, 2019
- An improved Canny edge detection algorithm2014
- An improved Sobel edge detection2010
- A Leaf Recognition Algorithm for Plant Classification Using Probabilistic Neural Network2007
- A novel automatic image processing algorithm for detection of hard exudates based on retinal image analysisMedical Engineering & Physics, 2007
- The FERET database and evaluation procedure for face-recognition algorithmsImage and Vision Computing, 1998
- Natural Language Processing: A Historical Review1994