Apparatus and method of pedestrian recognition

1. A pedestrian recognition apparatus comprising:

a radar unit for emitting a radar beam therefrom and receiving a reflected beam that is reflectively redirected by a reflective object when scanning of a detection area by the radar beam is performed in plural scan lines at a predetermined detection interval;

a reflective object detection unit for detecting a position of the reflective object as two-dimensional coordinates in a predetermined coordinate system based on a result of emission of the radar beam and reception of the reflected beam by the radar unit;

an object determination unit for determining whether the reflective object is a stopping object or a moving object based on a positional change of the reflective object represented by the two-dimensional coordinates in the predetermined coordinate system derived from plural times of position detection of the reflective object at the predetermined detection interval; and

a pedestrian recognition unit for recognizing a pedestrian, wherein pedestrian recognition is performed as formation of the object group of close-by reflective objects for forming an object group of the reflective objects in the two-dimensional coordinates in the predetermined coordinate system out of the reflective objects determined as the moving object based on a determination result by the object determination unit, the formation of the object group succeeded by determination of a group size of the object group based on a predetermined criterion of the group size for associating the object group with the pedestrian.

2. The pedestrian recognition apparatus as in claim 1 further comprising:

a motion detection unit for detecting a motion of a subject vehicle,

wherein the pedestrian recognition apparatus is disposed in the subject vehicle,

the object determination unit uses an absolute coordinate system with an arbitrarily determined origin as the predetermined coordinate system for defining the two-dimensional coordinates of an object position,

the position of the subject vehicle is calculated as coordinates in the absolute coordinate system based on a detection result of the motion detection unit, and

positional relationship of the subject vehicle relative to the position of the reflective object in the absolute coordinate system is used in association with the position of the subject vehicle for a calculation of the coordinates of the position of the reflective object.

3. The pedestrian recognition apparatus as in claim 2 ,

wherein the reflective object detection unit calculates the coordinates of the position of the subject vehicle by adopting a model that changes a slip angle according to a travel speed of the subject vehicle when the subject vehicle is in a turning motion.

4. The pedestrian recognition apparatus as in claim 2 ,

wherein the radar unit emits plural radar beams in vertically and horizontally different directions,

the plural radar beams in the vertically different directions are controlled to be reflected by the pedestrian, and

the reflective object detection unit determines a position of the reflective object detected by the plural radar beams in the vertically different directions by two-dimensional location coordinates that is commonly used to locate an object position on a road surface in the two-dimensional coordinate system.

5. The pedestrian recognition apparatus as in claim 1 ,

wherein the pedestrian detection unit calculates a predicted position of the object group of the reflective objects being determined as the pedestrian by applying Kalman filter that considers an acceleration component as the noise based on an assumption that the object group of the reflective objects is a body in a uniform motion, and

the pedestrian detection unit tracks the object group of the reflective objects based on a predetermined relationship between the predicted position and an actual measurement of the coordinates of the position of the object group of the reflective objects.

6. The pedestrian recognition apparatus as in claim 5 ,

wherein the pedestrian detection unit calculates the predicted position of the object group of the reflective objects being determined as the pedestrian even when the predetermined relationship between the predicted position and the actual measurement of the coordinates of the position of the object group of the reflective object is discrepant, and

the pedestrian detection unit continues to track the object group of the reflective objects based on the predetermined relationship between the predicted position and the actual measurement of the coordinates of the position of the object group of the reflective object at the predetermined detection interval for a predetermined number of times after the discrepancy is detected.

7. A pedestrian recognition method comprising:

emitting a radar beam and receiving a reflected beam that is derived as a result of reflective redirection of the radar beam by a reflective object when scanning of a detection area by the radar beam is performed in plural scan lines at a predetermined detection interval;

detecting a position of the reflective object as two-dimensional coordinates in a predetermined coordinate system based on a result of emission of the radar beam and reception of the reflected beam;

determining whether the reflective object is a stopping object or a moving object based on a positional change of the reflective object represented by the two-dimensional coordinates in the predetermined coordinate system as a result of plural times of scanning of the reflective object at the predetermined detection interval; and

performing pedestrian recognition, wherein the pedestrian recognition is performed as formation of an object group of close-by reflective objects for forming the object group of the reflective objects in the two-dimensional coordinates of the predetermined coordinate system out of the reflective objects determined as the moving object based on a determination result that each of the reflective objects is one of the stopping object and the moving object, the formation of the object group succeeded by determination of a group size of the object group based on a predetermined criterion of the group size for associating the object group with the pedestrian.

8. The method as in claim 7 further comprising:

detecting a motion of a subject vehicle,

wherein the pedestrian recognition is performed in the subject vehicle for recognizing a pedestrian in a proximity of the subject vehicle,

the predetermined coordinate system for defining the two-dimensional coordinates of an object position is an absolute coordinate system having an arbitrarily determined origin,

the position of the subject vehicle is calculated as coordinates in the absolute coordinate system based on a detection result of the motion of the subject vehicle, and

positional relationship of the subject vehicle relative to the position of the reflective object in the absolute coordinate system is used in association with the position of the subject vehicle for a calculation of the coordinates of the position of the reflective object.

9. The method as in claim 8 ,

wherein calculation of the coordinates of the position of the subject vehicle is performed by adopting a model that changes a slip angle according to a travel speed of the subject vehicle when the subject vehicle is in a turning motion.

10. The method as in claim 8 ,

wherein plural radar beams are emitted in vertically and horizontally different directions,

the plural radar beams in the vertically different directions are controlled to be reflected by the pedestrian, and

a position of the reflective object detected by the plural radar beams in the vertically different directions is determined as plural sets of coordinates that is commonly used to locate an object position on a road surface in the two-dimensional coordinate system.

11. The method as in claim 7 ,

wherein calculation of a predicted position of the object group of the reflective objects being determined as the pedestrian is performed by applying Kalman filter that considers an acceleration component as a noise based on an assumption that the group of the reflective objects is a body in a uniform motion, and

tracking of the object group of the reflective object is performed based on a predetermined relationship between the predicted position and an actual measurement of the coordinates of the position of the object group of the reflective objects.

12. The method as in claim 11 ,

wherein calculation of the predicted position of the object group of the reflective objects being determined as the pedestrian is performed even when the predetermined relationship between the predicted position and the actual measurement of the coordinates of the position of the object group of the reflective object is discrepant, and

tracking of the object group of the reflective objects is continued based on the predetermined relationship between the predicted position and the actual measurement of the coordinates of the position of the object group of the reflective object at the predetermined detection interval for a predetermined number of times after the discrepancy is detected.