System and method for location recognition and learning utilizing convolutional neural networks for robotic exploration

A computer program product for location recognition of a mobile platform, the computer program product comprising:

• a non-transitory computer-readable medium having executable instructions encoded thereon, such that upon execution of the instructions by one or more processors, the one or more processors perform operations of:
• causing a forward-looking video camera mounted on the mobile platform to capture images of a scene proximate the mobile platform;
• converting images captured by the camera into a high-dimensional feature vector
• identifying a candidate location identification of the scene;
• storing the candidate location identification in a history buffer;
• and
• upon receiving a cue, determining if a location of the scene is a known location or a new location by performing operations of:
• upon receiving the cue, processing images to generate feature vectors;
• filling the history buffer with new location identifications, such that after Lbuf image frames has passed, the history buffer is filled with all new location identifications;
• determining a most frequent location identification, idmax, exclusive of “−1”, in the history buffer and comparing its frequency with a threshold, Tmin, such that if the frequency is equal to or greater than Tmin, then designating the location of the scene as idmax as a known location and ending, otherwise continuing;
• generating a new unique location identification, idnew for the location of the scene;
• replacing the location identification of each feature vector in the history buffer with idnew;
• and
• adding all the feature vectors in the history buffer into a feature database and designating the location of the scene as a new location.

A computer implemented method for location recognition of a mobile platform, the method comprising an act of:

• causing one or more processers to execute instructions encoded on a non-transitory computer-readable medium, such that upon execution, the one or more processors perform operations of:
• causing a forward-looking video camera mounted on the mobile platform to capture images of the scene proximate the mobile platform;
• converting images captured by the camera into a high-dimensional feature vector
• identifying a candidate location identification of the scene;
• storing the candidate location identification in a history buffer;
• and
• upon receiving a cue, determining if a location of the scene is a known location or a new location by performing operations of:
• upon receiving the cue, processing images to generate feature vectors;
• filling the history buffer with new location identifications, such that after Lbuf image frames has passed, the history buffer is filled with all new location identifications;
• determining a most frequent location identification, idmax, exclusive of “−1”, in the history buffer and comparing its frequency with a threshold, Tmin, such that if the frequency is equal to or greater than Tmin, then designating the location of the scene as idmax as a known location and ending, otherwise continuing;
• generating a new unique location identification, idnew for the location of the scene;
• replacing the location identification of each feature vector in the history buffer with idnew;
• and
• adding all the feature vectors in the history buffer into a feature database and designating the location of the scene as a new location.