Object recognition device and object recognition method

An object recognition method by an object recognition device for recognizing an object, comprising the steps of:

• allowing a memory to store a reference image of an object of a recognition candidate, each feature point in the reference image, and a feature quantity at each feature point;
• acquiring a scene image that is an image of a recognition processing target;
• detecting, via a processor, a definition indicating a degree of sharpness in each region of the acquired scene image;
• detecting, via the processor, a feature point in the scene image to perform a process of matching with the feature point,
• executing, via the processor, by different methods, in the detection of the feature point and the process of matching with the feature point, at least one of an extraction method of feature points in a first region of the scene image where the definition is a first range and in a second region of the scene image where the definition is a second range lower than the first range, and a detection method of feature quantities in the first region and the second region,
• storing, in the memory, a first feature quantity that is high in expressing capability and a second feature quantity that is high in a degree of robustness for each feature point of the reference image,
• performing, via the processor, a matching process of a feature point using the first feature quantity for the first region of the scene image and performs a matching process of a feature point using the second feature quantity for the second region,
• storing, in the memory, a third feature quantity indicating a three-dimensional position information of an image for each region of the reference image, and
• performing, via the processor, a process of matching with a feature point of the reference image using the third feature quantity for a third region of the scene image of which the definition is a third range lower than the second range.