To provide an art capable of easily generating accurate teaching data from a small number of teacher images.The image of a standard object is captured to extract a characteristic value from the teacher image of the standard object through image processing while making a user input the already-known knowledge for a configuration of an object and an object such as a tolerance and entering the knowledge in a knowledge base. Then, context data in which various attributes which can be objects are described together with their semantics is generated in accordance with the characteristic value extracted from the teacher image of the standard object and the knowledge entered in the knowledge base. Moreover, attributes corresponding to types of characteristic values used for recognition processing are extracted from the context data entered in the knowledge base to generate the teaching data for the recognition processing.