Object identification apparatus and method for identifying object

A computer-implemented method for identifying an object, the method comprising:

• detecting an object from captured image data;
• generating a feature vector from each of a plurality of partial areas of the object detected from the captured image data;
• estimating an attribute of an appearance of the object detected from the captured image data;
• converting the feature vector into object identification data by a converting method corresponding to the estimated attribute of the appearance of the object and the partial area from which the feature vector has been generated, wherein one of a plurality of converting methods is defined in correspondence with each attribute of the object with respect to each of the plurality of partial areas;
• acquiring, from an object dictionary data storage unit, object dictionary data for each of the plurality of partial areas of the object obtained by one of the plurality of converting methods corresponding to the estimated attribute of the appearance of the object detected from the captured image data and respective partial areas, wherein object dictionary data storage unit stores object dictionary obtained by converting each feature vector generated from each registered object by respective ones of the plurality of converting methods;
• identifying a class to which the object detected from the captured image data belongs, based on the object identification data obtained by converting the feature and the acquired object dictionary data;
• and
• the object identification unit configured to use a support vector machine previously trained to identify classes of variation between objects to identify the class.

A non-transitory computer-readable storage medium storing a program that causes a computer to execute a method for identifying an object, the method comprising:

• detecting an object from captured image data;
• generating a feature vector from each of a plurality of partial areas of the object detected from the captured image data;
• estimating an attribute of an appearance of the object detected from the captured image data;
• converting the feature vector into object identification data by a converting method corresponding to the estimated attribute of the appearance of the object and the partial area from which the feature vector has been generated, wherein one of a plurality of converting methods is defined in correspondence with each attribute of the object with respect to each of the plurality of partial areas;
• acquiring, from an object dictionary data storage unit, object dictionary data for each of the plurality of partial areas of the object obtained by one of the plurality of converting methods corresponding to the estimated attribute of the appearance of the object detected from the captured image data and respective partial areas, wherein object dictionary data storage unit stores object dictionary obtained by converting each feature vector generated from each registered object by respective ones of the plurality of converting methods;
• identifying a class to which the object detected from the captured image data belongs, based on the object identification data obtained by converting the feature and the acquired object dictionary data;
• and
• the object identification unit configured to use a support vector machine previously trained to identify classes of variation between objects to identify the class.