Dictionary learning device, dictionary learning method, and program storage medium

A dictionary learning method, with a computer, comprising:

• extracting an image of a reference image area from a captured image based on reference extraction information, the reference image area including a detection target in an object to be recognized and serving as a reference, the captured image being captured by an imaging device that images the object to be recognized;
• extracting an image of an enlarged image area from the captured image based on enlargement extraction information, the enlarged image area including the reference image area and being larger than the reference image area, and including a background in which an object other than the object to be recognized is shown;
• extracting an image of a reduced image area from the captured image based on reduction extraction information, the reduced image area including the detection target and being smaller than the reference image area, and being an area in which the object to be recognized is partially outside the reduced image area;
• and
• learning a dictionary including reference data to be referred to in detection processing of detecting the detection target from the captured image by using the images of the reference image area, the enlarged image area, and the reduced image area.

A non-transitory computer readable storage medium containing program instructions for causing a computer to execute:

• extracting an image of a reference image area from a captured image based on reference extraction information, the reference image area including a detection target in an object to be recognized and serving as a reference, the captured image being captured by an imaging device that images the object to be recognized;
• extracting an image of an enlarged image area from the captured image based on enlargement extraction information, the enlarged image area including the reference image area and being larger than the reference image area, and including a background in which an object other than the object to be recognized is shown;
• extracting an image of a reduced image area from the captured image based on reduction extraction information, the reduced image area including the detection target and being smaller than the reference image area, and being an area in which the object to be recognized is partially outside the reduced image area;
• and
• learning a dictionary including reference data to be referred to in detection processing of detecting the detection target from the captured image by using the images of the reference image area, the enlarged image area, and the reduced image area.