Object identification system
An object identification system includes a camera and a processing device. The processing device includes a classifier subjected to machine learning based on the output image of the camera so as to allow it to identify an object. A gating camera divides a field of view in the depth direction into multiple ranges, and captures an image while changing the time difference between light projection and exposure for each range. The classifier is subjected to machine learning using multiple images IMG 1 through IMG N generated by the gating camera as the learning data.
1. An object identification system comprising:
a camera; and
one or more processors comprising a classifier subjected to machine learning so as to be capable of identifying one or more objects, which are separate from each other, based on an output image of the camera,
wherein the classifier is subjected to machine learning based on learning data configured as a plurality of images generated by a gating camera structured to divide a field of view in a depth direction into a plurality of ranges, and to capture an image while changing a time difference between light projection and exposure for each range, and
wherein the plurality of images generated by the gating camera includes the one or more objects and the plurality of images is input into the classifier as separate items of learning data,
wherein the learning data is the plurality of images being generated as separate images in a state in which the one or more objects overlap each other in the depth direction and are positioned in different ranges.
2. The object identification system according to claim 1 , wherein the camera is structured as a monocular camera.
3. The object identification system according to claim 1 , wherein the camera is structured as a gating camera.
4. An automotive lamp comprising the object identification system according to claim 1 .
5. An automobile comprising the object identification system according to claim 1 .
6. One or more processors comprising:
a classifier subjected to machine learning so as to be capable of identifying one or more objects, which are separable from each other, based on an output image of a camera,
wherein the classifier is subjected to machine learning based on learning data configured as a plurality of images generated by a gating camera structured to divide a field of view in a depth direction into a plurality of ranges, and to capture an image while changing a time difference between light projection and exposure for each range, and
wherein the plurality of images generated by the gating camera includes the one or more objects and the plurality of images is input into the classifier as separate items of learning data,
wherein the learning data is the plurality of images being generated as separate images in a state in which the one or more objects overlap each other in the depth direction and are positioned in different ranges.
7. A learning method for a classifier structured to identify one or more objects, which are separate from each other, based on an output image of a camera, the learning method comprising:
dividing a depth direction into a plurality of ranges, and acquiring a plurality of images using a gating camera while changing a time difference between light projection and exposure for each range; and
applying machine learning to the classifier using learning data configured as a plurality of images generated by the gating camera,
wherein the plurality of images generated by the gating camera includes the one or more objects and the plurality of images is input into the classifier as separate items of learning data,
wherein the learning data is the plurality of images being generated as separate images in a state in which the one or more objects overlap each other in the depth direction and are positioned in different ranges.
8. An object identification system comprising:
a gating camera structured to divide a field of view in a depth direction into a plurality of ranges, and to capture an image while changing a time difference between light projection and exposure for each range; and
one or more processors configured to be capable of identifying a kind of one or more objects, which are separate from each other, based on a plurality of images that correspond to a plurality of ranges generated by the gating camera,
wherein the one or more processors comprises a plurality of classifiers structured to have different algorithms used for the separate objects of the one or more objects, and to select the classifier that corresponds to the range so as to identify the one or more objects,
wherein the one or more processors employs a classifier with a You Only Look Once (YOLO) algorithm for image processing in a first-distance range,
wherein the one or more processors employs a classifier with a Single Shot MultiBox Detector (SSD) algorithm for image processing in a second-distance range,
wherein the first-distance range is one or more distance ranges closer to the gating camera than the second-distance range, and
wherein the first-distance range and the second-distance range overlap at each respective boundaries in the depth direction.
9. An automobile comprising the object identification system according to claim 8 .
10. An automotive lamp comprising the object identification system according to claim 8 .
11. One or more processors that forms an object identification system together with a gating camera, wherein the gating camera is structured to divide a field of view in a depth direction into a plurality of ranges, and to capture an image while changing a time difference between light projection and exposure for each range,
wherein the one or more processors is structured to be capable of identifying a kind of one or more objects, which are separate from each other, based on a plurality of images that correspond to a plurality of ranges generated by the gating camera,
wherein the one or more processors comprises a plurality of classifiers structured to have different algorithms used for the separate objects of the one or more objects, and to select a classifier that corresponds to the range so as to identify the one or more objects,
wherein the one or more processors employs a classifier with a You Only Look Once (YOLO) algorithm for image processing in a first-distance range, and
wherein the one or more processors employs a classifier with a Single Shot MultiBox Detector (SSD) algorithm for image processing in a second-distance range,
wherein the first-distance range is one or more distance ranges closer to the gating camera than the second-distance range, and
wherein the first-distance range and the second-distance range overlap at each respective boundaries in the depth direction.