Data generation device, image identification device, data generation method, and recording medium

The data generation device according to claim

• wherein the first radar parameter includes:
• an incident angle that indicates an angle formed by a vertical direction of the target object and a straight line that connects the radar and the target object;
• and
• an irradiation direction that indicates an angle formed by a travelling direction of the target object and a direction from the radar to the target object.

A data generation method that a processing device executes, the method including:

• a step of setting a radar parameter that indicates a specification of a radar that is a synthetic aperture radar or an inverse synthetic aperture radar;
• a step of setting a three-dimensional model that indicates a shape of a target object to identify;
• a step of calculating a simulation image based on the three-dimensional model and the radar parameter;
• a step of outputting a difference data that indicates a difference between a radar image of the target object acquired by the radar and the simulation image;
• a step of changing the three-dimensional model based on model correction data inputted based on the difference data;
• a step of setting a plurality of radar parameters when the three-dimensional model is determined to be decided;
• a step of calculating a plurality of simulation images respectively corresponding to the plurality of radar parameters;
• and
• a step of outputting training data in that the plurality of simulation images and types of the target object are associated to each other,
• wherein the training data are data for an image identification device, configured to identify a type of an object taken in an inputted image, to learn.

A recording medium that stores a program to make a processing device execute:

• a step of setting a radar parameter that indicates a specification of a radar that is a synthetic aperture radar or an inverse synthetic aperture radar;
• a step of setting a three-dimensional model that indicates a shape of a target object to identify;
• a step of calculating a simulation image based on the three-dimensional model and the radar parameter;
• a step of outputting a difference data that indicates a difference between a radar image of the target object acquired by the radar and the simulation image;
• a step of changing the three-dimensional model based on a model correction data inputted based on the difference data;
• a step of setting a plurality of radar parameters when the three-dimensional model is determined to be decided;
• a step of calculating a plurality of simulation images respectively corresponding to the plurality of radar parameters;
• and
• a step of outputting training data in that the plurality of simulation images and types of the target object are associated to each other,
• wherein the training data are data for an image identification device, configured to identify a type of an object taken in an inputted image, to learn.