Visual and semantic fusion fault identification algorithm for power inspection robot

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.