Method and device for the detection and classification of an object

A method for detecting and classifying a not-known object (4) by a device (1) having a two-dimensional field of view, including:

• a learning step (1F) comprising:
• positioning a first known object (4a) in the field of view of the device (1) having a first plurality of visual elements and obtaining a first image thereof;
• positioning a second known object (4b) having a second plurality of visual elements in the field of view of the device (1), and obtaining a second image thereof, said second known object having a visual element of said second plurality distinct from the first plurality and the remaining visual elements of the second plurality equal to those of the first plurality;
• identifying a distinct characteristic between said first image of said first object and said second image of said second object and associating thereto a difference descriptor (9) allowing to associate the first known object and the second known object to two distinct classes;
• and
• identifying a common characteristic between said image of said first object and said image of said second object and associating thereto an equality descriptor (8) relative to common elements between said first and second plurality of visual elements;
• a detection step (2F) including:
• detecting an image (10) of a third not-known object having the first or the second plurality of visual elements;
• processing said detected image;
• and
• accepting said image as a complete image (10′) including said first or second plurality of visual elements based on said equality descriptor (8);
• an inspection step (3F), including:
• classifying said third not-known object (4) based on an analysis of said complete image (10′) and of said difference descriptor (9).

A device (1) for detecting and classifying objects (4, 4a, 4b), including:

• a two-dimensional emitter (2) of luminous radiation adapted to illuminate a region in the space;
• a two-dimensional receiver (3), including an array of photo sensitive cells, adapted to detect luminous energy spread by an object (4, 4a, 4b) illuminated by said emitter and adapted to output a signal in form of function images (10) of said luminous energy;
• a processing unit (5) of the signal outputting from the receiver (3), said unit being programmed for being alternatively in a learning mode (1F) and in a detection and inspection mode (2F, 3F), wherein in the learning mode (1F) it is adapted to:
• acquire a first image of a first known object (4a) having a plurality of visual elements placed in the field of view of the receiver;
• acquire a second image of a second known object (4b) having one second plurality of visual elements, said second known object having a visual element of said second plurality distinct from the first plurality and the remaining visual elements of the second plurality equal to those of the first one;
• identify a distinct characteristic between said image of said first object and said image of said second object and associate thereto a difference descriptor (9) allowing to associate the first known object and the second known object to two distinct classes;
• identify a common characteristic between said image of said first object and said image of said second object and associate thereto an equality descriptor (8) relative to common elements between said first and second plurality of visual elements;
• and in the detection and inspection steps (2F, 3F) it is adapted to:
• acquire an image of a third not-known object (4) having the first or the second plurality of visual elements;
• accept said image as a complete image (10′) including said first or second plurality of visual elements based on said equality descriptor (8);
• classify said third not-known object (4) based on an analysis of said complete image (10′) and of said difference descriptor (9).