Prediction of illuminance on the work plane using low dynamic unprocessed image data

Light level measurement is crucial in applications involving lighting control, circadian lighting, and light booths. The paper proposes a unique camera-based approach for measuring illuminance on a work plane. A machine learning-based model is suggested that takes the RAW pixel data of standard neutral colour patches placed on the work plane as inputs and predicts the illuminance level on the work plane. The training data is populated using a DSLR camera, calibrated illuminance meter, standard neutral patches, and controlled lighting in the darkroom. The RAW image files are processed using MATLAB, and the App Designer tool is used to develop the user interface. The model's performance is accessed under various illuminance levels and colour temperatures. The percentage error of prediction during testing by the model was less than 5. Since the proposed technique incorporates the camera as the light sensor, simultaneous illuminance measurement at multiple target points present in the camera vision can be realised.