B&P Project 23-4: Autonomous Measurement Readout of Multiple, Non-instrumented, Analog Gauges Using Artificial Intelligence

This project developed a system that can read analog gauges using a video source. The low TRL work implemented a combination of algorithms that included artificial intelligence (AI), computer vision, and statistics. This work developed a framework for the processes of collecting and pre-processing gauge image data, selecting the structure and parameters of a convolutional neural network, and optimizing them for deep learning. The work also included the development of algorithms and processes to recognize a particular gauge, from others, on a ground support equipment (GSE) panel visible from a single camera field of view. In addition to the implementation of the gauge system in MATLAB, this work implemented the machine learning and AI algorithms in C-language for deployment in small form, low-cost embedded systems such as a raspberry pi single board computer.