Active audio-visual surveillance system

A method for identifying events occurring at a location-of-interest comprising:

• capturing continuously, using a Fisheye image capturing device provided at the location-of-interest, wide-angle image frames and converting the captured wide-angle image frames into non-distorted perspective images using a set of calibration parameters generated by a fisheye correction scheme;
• capturing, using an audio module provided at the location-of-interest, audio signals and determining a Direction of Arrival (DOA) of the audio signals;
• receiving, using a computing module communicatively coupled to the audio module and the Fisheye image capturing device, a notification from the audio module when the audio module detects an event having an audio signal at the location-of-interest, whereby the notification includes data of a DOA of the audio signal;
• instructing, using the computing module, the Fisheye image capturing device to generate, using a trained single-image-super-resolution (SISR) scheme, high-resolution perspective images based on the non-distorted perspective images associated with the DOA of the audio signal,
• wherein the step of detecting the event having an audio signal at the location-of-interest comprises:
• generating, using the audio module, a discrete set of DOA points that cover a DOA region from which the audio signal of the event was detected;
• and
• embedding, using the audio module, the discrete set of DOA points in the notification that is sent to the computing module
• wherein the step of generating the high-resolution perspective images based on the non-distorted perspective images by the Fisheye image capturing device comprises:
• extracting, using the Fisheye image capturing device, the discrete set of DOA points received from the computing module, wherein the discrete set of DOA points are provided in the instructions sent by the computing module;
• generating, using the Fisheye image capturing device, a unified region of interest of the non-distorted perspective images that corresponds to the extracted set of DOA points;
• and
• generating, using the Fisheye image capturing device and the trained single-image-super-resolution (SISR) scheme, high-resolution perspective images of the unified region of interest.