Computer Vision-Based Information Extraction from Electrical Assets: A Systematic Literature Review

Automated inspection of electrical equipment has become increasingly important as utilities seek to replace inefficient manual processes in infrastructure management. Electrical-equipment nameplates contain critical technical information, but automatically extracting this information remains challenging. Although this topic has attracted growing research interest, the literature has not yet been systematically synthesized. This review addresses three research questions: (1) What real-world challenges affect the inspection and maintenance of electrical equipment, and what research gaps and opportunities remain? (2) Which computer vision and deep learning algorithms and architectures are used to extract text and digits from electrical equipment images? (3) Which metrics, datasets, and evaluation methodologies are used to assess these approaches? We conducted a systematic literature review in accordance with PRISMA 2020. Peer-reviewed English-language studies published between 2020 and 2026 were retrieved from IEEE Xplore, Scopus, ScienceDirect, SpringerLink, the ACM Digital Library, and complementary semantic-search sources; the final searches were conducted in May 2026. The screening and eligibility process reduced 19,760 initial records to 66 primary studies. Owing to substantial heterogeneity in tasks, datasets, and metrics, the evidence was synthesized narratively and in structured tables rather than through meta-analysis. DBNet and CRNN were among the most frequently used architectures for text detection and recognition, respectively, whereas integrated end-to-end systems emerged as a promising research direction. The scarcity of standardized public datasets remains a major barrier to objective comparison and cumulative progress in the field.