Estimating rainfall intensity using surveillance audio and deep-learning

Surveillance cameras generate both video and audio outputs. Unlike video images recorded, the audio can be supplemented reliably as audio sources resist background interference and lighting variability. Creating a reliable way to use these audio sources to estimate the intensity of rainfall could open a new chapter in rainfall intensity estimation.

In a study published in Environmental Science and Ecotechnology, researchers created an audio dataset of six real-world rainfall events, named the Surveillance Audio Rainfall Intensity Dataset (SARID). This dataset's audio recordings were segmented into 12,066 pieces and annotated with rainfall intensity and environmental information, such as underlying surfaces, temperature, humidity, and wind.

The researchers developed a deep learning-based baseline to estimate rainfall intensity from surveillance audio. Validated from ground truth data, the research baseline from the system deployed achieved a root mean absolute error of 0.88 mm h-1 and a coefficient of correlation of 0.765.

These findings demonstrate the potential of surveillance audio-based models as practical and effective tools for rainfall observation systems, initiating a new chapter in rainfall intensity estimation.

The work offers a new approach for high-resolution hydrological sensing and contributes to the broader landscape of urban sensing, emergency response, and environmental resilience.

END