Image recognition device
An image recognition device includes: a luminance image generator and a distance image generator that generate a luminance image and a distance image, respectively, based on an image signal of an imaging target object output from a photoreceptor element; a target object recognition processor that extracts a target-object candidate from the luminance image using a machine learning database; and a three-dimensional object determination processor that uses the distance image to determine whether the extracted target-object candidate is a three-dimensional object. If it is determined that the target-object candidate is not a three-dimensional object, the target-object candidate extracted from the luminance image is prevented from being used, in the machine learning database, as image data for extracting a feature value of a target object.
1. An image recognition device, comprising:
a light source that emits modulated light which is temporally intensity-modulated;
a photoreceptor element that receives reflected light from an imaging target object irradiated with the modulated light;
a luminance image generator that generates a luminance image based on an image signal of the imaging target object, the image signal being output from the photoreceptor element;
a distance image generator that generates a distance image based on the image signal of the imaging target object, the image signal being output from the photoreceptor element;
a target object recognition processor that extracts a target-object candidate from the luminance image using a machine learning database storing image data for recognizing a feature value of a target object; and
a three-dimensional object determination processor that determines, based on the distance image, whether the target-object candidate extracted by the target object recognition processor is a three-dimensional object,
wherein, when the three-dimensional object determination processor determines that the target-object candidate is not a three-dimensional object, the target-object candidate extracted from the luminance image is prevented from being used, in the machine learning database, as image data for extracting a feature value of a target object.
2. The image recognition device according to claim 1 ,
wherein the three-dimensional object determination processor
calculates, for pixels in a plurality of lines in the distance image of the target-object candidate, distances between the target-object candidate extracted by the target object recognition processor and the photoreceptor element,
calculates an average and a dispersion of the distances for each of the plurality of lines, and
determines whether the target-object candidate is a three-dimensional object, based on a proportion of pixels whose deviation from the average is a predetermined value or more with respect to the dispersion to all pixels in the line.
3. The image recognition device according to claim 2 ,
wherein the three-dimensional object determination processor determines that the target-object candidate is a three-dimensional object when pixels whose deviation from the average is two times or more than the dispersion constitute 10% or more of all pixels in the line.
4. The image recognition device according to claim 1 ,
wherein the modulated light emitted from the light source has a wavelength in a range from 800 nm to 1100 nm, inclusive.