Deep Learning–Based Object Detection for Automated Solar Panel Inspection: A Comprehensive Review

Abstract

With the proliferation of photovoltaic (PV) installations, there is an emerging need for low cost and robust inspection systems to safeguard PV functionality over time and minimize costs associated with maintenance. Traditional inspection methods are time and labor-consuming, and lack in scalability. Nowadays, automatic defect detection using deep learning-based object detection models has been possible thanks to the development of artificial intelligence and computer vision. This paper offers a thorough summary of typical one-stage, twostage and transformer-based detection models for solar panel inspection. The paper studies the typical defect types and imaging modalities of the data, model performances and deployment constraints in embedded/edge platforms. We give a detailed review and discuss the significance of these challenges to practical implementation. In addition, we discuss emerging research directions on multi-modal learning, model optimization, and decision-support integration. In this paper, efficient and practical design guides for high-scalable, fault-tolerant intelligent PVMS are given.