Automatic Electrical Wiring Inspection with Advanced Computer Vision Models

Abstract Safety and reliability are paramount in energy industries. In this context, guaranteeing precise electrical wiring connections within terminal blocks is crucial for maintaining quality standards and enabling efficient remote support for on-site installations. Unfortunately, the current wiring quality checks are time-consuming and repetitive, and they are prone to human errors. This situation has led to an increasing interest in leveraging the capabilities of Artificial Intelligence (AI) to expedite the inspection process. AI assistants can either help in or completely automate the tedious checking operations, thereby enhancing operational efficiency. In this work, we introduce a novel approach that can be used to automatically verify the correctness of wiring connections to a terminal block. Our proposed methodology consists of a series of sequential operations, each contributing to the overall effectiveness of the system. First, we use an instance segmentation model to identify and isolate the connections between PINs and wires on a terminal block image. Then, we use text detection and recognition models to translate the pixels into an alphanumeric string. Lastly, we parse the obtained strings and compare results with the expected wiring connections in a reference database. Our proposed approach is not only simple and modular but also highly effective. It only requires a camera and a mobile device, and it swiftly generates a comprehensive report about the wire connections that are under inspection. This method is not only fast, but it also demonstrates promising performance levels (e.g. detecting connections with Precision and Recall > 96%), thereby confirming the potential of AI-based systems in revolutionizing the inspection process. With our work, we underscore the potential of AI-based systems in advancing traditional processes and setting new standards for efficiency and accuracy in the energy sector.