MOVING OBJECT PREDICTION DEVICE, HYPOTHETICAL MOVABLE OBJECT PREDICTION DEVICE, PROGRAM, MOVING OBJECT PREDICTION METHOD AND HYPOTHETICAL MOVABLE OBJECT PREDICTION METHOD

• US 20130223686A1
• Filed: 09/08/2011
• Published: 08/29/2013
• Est. Priority Date: 09/08/2010
• Status: Active Grant

• Current Assignee: Toyota Jidosha Kabushiki Kaisha (Toyota Motor Corporation)
• Original Assignee: Toyota Jidosha Kabushiki Kaisha (Toyota Motor Corporation)
• Inventors: Nagata, Shinichi, Shimizu, Masayuki, Sakugawa, Jun, Shimizu, Tsukasa, Ohama, Yoshihiro
• Granted Patent: US 9,424,468 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06T 2207/20076 Probabilistic image processing
• CPC Class Codes: G06T 2207/30261 Obstacle
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments
• CPC Class Codes: G06V 20/58 Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
• CPC Class Codes: G08G 1/166 for active traffic, e.g. moving vehicles, pedestrians, bikes

Abstract

A position, behavior state and movement state of a moving object are detected, together with plural categories of track segment region and stationary object regions, using an environment detection section. A presence probability is applied to the detected track segment regions and stationary object regions and a presence probability map is generated, using a map generation section. A moving object position distribution and movement state distribution are generated by a moving object generation section based on the detected moving object position, behavior state and movement state, and recorded on the presence probability map. The moving object position distribution is moved by a position update section based on the moving object movement state distribution. The moved position distribution is changed by a distribution change section based on the presence probabilities of the presence probability map, and a future position distribution of the moving object is predicted on the presence probability map. Consequently, the future position of the moving object can be predicted with good precision under various conditions.

Claims

1. A moving object prediction device comprising:

a moving object detection section that detects a position, and a behavior state or a movement state, of a moving object in a detection target range;

a region detection section that detects a plurality of categories of regions of track segment and regions of stationary object in the detection target range;

a map generation section that generates a presence probability map in which a presence probability expressing a presence likelihood or presence unlikelihood of a moving object is applied to the track segment regions and the stationary object regions detected by the region detection section;

a moving object recording section that generates a position distribution and a movement state distribution of the moving object based on the position and the behavior state or the movement state of the moving object detected by the moving object detection section, and records the distributions on the presence probability map; and

a prediction section that moves the moving object position distribution based on the moving object movement state distribution, changes the moved position distribution based on presence probabilities of the presence probability map, and predicts a future position distribution of the moving object on the presence probability map.

2. The moving object prediction device of claim 1, wherein:

the position distribution is represented by a plurality of moving object particles; and

the prediction section moves each of the plurality of moving object particles representing the moving object position distribution, based on the moving object movement state distribution, changes the moving object position distribution by eliminating the moved moving object particles and replicating other moving object particles based on presence probabilities of the presence probability map, and predicts a future position distribution of the moving object on the presence probability map.

3. The moving object prediction device of claim 1, wherein:

the position distribution is represented by a probability distribution; and

the prediction section moves the probability distribution representing the moving object position distribution based on the moving object movement state distribution, changes the moving object position distribution by weighting the moved probability distribution based on presence probabilities of the presence probability map, and predicts a future position distribution of the moving object on the presence probability map.

4. The moving object prediction device of claim 1, wherein:

the region detection section further detects categories of the track segment and a height of the stationary object; and

the map generation section generates the presence probability map by applying the presence probability corresponding to the track segment category to the track segment region, and applying the presence probability corresponding to the stationary object height to the stationary object region.

5. The moving object prediction device of claim 1, wherein:

the moving object detection section detects a category of the moving object together with the position of the moving object in the detection target range and the behavior state or the movement state of the moving object; and

the map generation section generates the presence probability map with the presence probability applied for each of the moving object categories.

6. The moving object prediction device of claim 1, wherein:

the prediction section predicts the moving object future position distribution on the presence probability map by moving the moving object position distribution, based on the moving object movement state distribution and changing the moved position distribution based on the magnitude of the presence probability applied to the region corresponding to the position distribution, or based on the difference or ratio between the presence probability applied to the region corresponding to the position distribution and the presence probability applied to the region for the position distribution prior to moving.

7. The moving object prediction device of claim 1, further comprising a dangerousness computation section that computes a danger of collision between a computation subject moving object and a moving object other than the computation subject moving object, based on a position distribution of the computation subject moving object and a position distribution of the moving object other than the computation subject moving object, which have been predicted by the prediction section.

8. The moving object prediction device of claim 1, wherein the moving object recording section:

identifies a blind-spot region formed by a stationary object as viewed from the moving object installed with the device itself based on the detected stationary object region and the position of the moving object installed with the device itself;

generates a position distribution and a movement state distribution of a moving object postulated to be present in the identified blind-spot region; and

records the generated moving object position distribution and movement state distribution in the identified blind-spot region on the presence probability map.

9. The moving object prediction device of claim 8, wherein the moving object recording section:

identifies the blind-spot region based on the detected stationary object region and the position of the moving object installed with the device itself;

estimates the category of a moving object postulated to be present in the identified blind-spot region based on the movement environment at the periphery of the identified blind-spot region, out of movement environments detected by a movement environment detection section that detects movement environments of moving objects in the detection target range, or based on a position of the identified blind-spot region, and generates a position distribution and movement state distribution of the moving object in the blind-spot region; and

records the generated position distribution and the movement state distribution of the moving object in the identified blind-spot region on the presence probability map.

10. The moving object prediction device of claim 1, further comprising:

an object position detection section that detects a position of an object present in the detection target range as viewed from a moving object installed with the device itself;

a motion estimation section that estimates the motion of the moving object installed with the device itself;

a map update section that repeatedly updates map data as currently viewed from the moving object installed with the device itself by updating the map data from a previous update time based on a current motion of the moving object installed with the device itself estimated by the motion estimation section; and

a stationary object recording section that, each time the map data is updated by the map update section, records in the updated map data a stationary object presence in a block corresponding to a position of an object currently detected by the object position detection section, and reduces a record of stationary object presence in each block corresponding to between the moving object installed with the device itself and the current detected object position, wherein

the region detection section detects the stationary object region based on the map data.

11. A hypothetical movable object prediction device comprising:

a region detection section that detects a plurality of categories of regions of track segment and regions of stationary object in a detection target range;

a map generation section that generates a presence probability map in which a presence probability expressing a presence likelihood or presence unlikelihood of a moving object is applied to the track segment regions and the stationary object regions detected by the region detection section;

a blind-spot region identification section that identifies a blind-spot region in the detection target range where a movable object might be present;

a hypothetical movable object generation section that generates a hypothetical movable object postulated to be present in the blind-spot region identified by the blind-spot region identification section;

a hypothetical movable object recording section that, for the hypothetical movable object generated by the hypothetical movable object generation section, postulates a position and a behavior state or movement state of the hypothetical movable object, generates a position distribution and a movement state distribution of the hypothetical movable object, and records the distributions on the presence probability map; and

a prediction section that moves the hypothetical movable object position distribution based on the hypothetical movable object movement state distribution, changes the moved position distribution based on presence probabilities of the presence probability map, and predicts a future position distribution of the hypothetical movable object on the presence probability map.

12. The hypothetical movable object prediction device of claim 11, wherein the blind-spot region identification section together with identifying the blind-spot region also sets a presence probability of the blind-spot region in the presence probability map based on a distance of the blind-spot region.

13. The hypothetical movable object prediction device of claim 11, wherein, for the hypothetical movable object generated by the hypothetical movable object generation section, the hypothetical movable object recording section postulates a position of the hypothetical movable object and a behavior state or movement state of the hypothetical movable object, and also postulates the category of the hypothetical movable object based on a distance of the blind spot region in which the hypothetical movable object might be present, generates a position distribution and a movement state distribution for the hypothetical movable object and records the distributions on the presence probability map.

14. The hypothetical movable object prediction device of claim 11, wherein for the hypothetical movable object generated by the hypothetical movable object generation section, the hypothetical movable object recording section postulates a position of the hypothetical movable object and a behavior state or a movement state of the hypothetical movable object, such that the hypothetical movable object will collide with the moving object installed with the device itself when the hypothetical movable object jumps out from the blind-spot region, generates a position distribution and a movement state distribution for the hypothetical movable object and records the distributions on the presence probability map.

15. The hypothetical movable object prediction device of claim 11, wherein, for the hypothetical movable object generated by the hypothetical movable object generation section, the hypothetical movable object recording section:

postulates as a position of the hypothetical movable object a position in the blind-spot region closest to the position of postulated collision with the moving object installed with the device itself when the hypothetical movable object jumps out from the blind-spot region, and postulates a behavior state or a movement state of the hypothetical movable object such that the hypothetical movable object will collide with the moving object installed with the device itself when the hypothetical movable object jumps out from the blind-spot region; and

generates a position distribution and a movement state distribution for the hypothetical movable object and records the distributions on the presence probability map.

16. A non-transitory computer-readable storage medium that stores a program that causes a computer to function as:

a moving object detection section that detects a position and a behavior state or a movement state, of a moving object in a detection target range;

a region detection section that detects a plurality of categories of regions of track segment and regions of stationary object in the detection target range;

a map generation section that generates a presence probability map in which a presence probability expressing a presence likelihood or presence unlikelihood of a moving object is applied to the track segment regions and the stationary object regions detected by the region detection section;

a moving object recording section that generates a position distribution and a movement state distribution of the moving object based on the position and the behavior state or the movement state of the moving object detected by the moving object detection section, and records the distributions on the presence probability map; and

a prediction section that moves the moving object position distribution based on the moving object movement state distribution, changes the moved position distribution based on presence probabilities of the presence probability map, and predicts a future position distribution of the moving object on the presence probability map.

17. A moving object prediction method comprising:

a step of, by a moving object detection section, detecting a position, and a behavior state or a movement state, of a moving object and, by a region detection section, detecting a plurality of categories of regions of track segment and regions of stationary object in the detection target range;

a step of, by a map generation section, generating a presence probability map in which a presence probability expressing a presence likelihood or presence unlikelihood of a moving object is applied to the detected track segment regions and the stationary object regions; and

a step of, by a moving object recording section, generating a position distribution and a movement state distribution of the moving object based on the position and the behavior state or the movement state of the detected moving object and recording the distributions on the presence probability map, and also, by a prediction section, moving the moving object position distribution based on the moving object movement state distribution, changing the moved position distribution based on presence probabilities of the presence probability map, and predicting a future position distribution of the moving object on the presence probability map.

18. A non-transitory computer-readable storage medium that stores a program that causes a computer to function as:

a region detection section that detects a plurality of categories of regions of track segment and regions of stationary object in a detection target range;

a map generation section that generates a presence probability map in which a presence probability expressing a presence likelihood or presence unlikelihood of a moving object is applied to the track segment regions and the stationary object regions detected by the region detection section;

a blind-spot region identification section that identifies a blind-spot region in the detection target range where a movable object might be present;

a hypothetical movable object generation section that generates a hypothetical movable object postulated to be present in the blind-spot region identified by the blind-spot region identification section;

a hypothetical movable object recording section that, for the hypothetical movable object generated by the hypothetical movable object generation section, postulates a position, and a behavior state or movement state, of the hypothetical movable object, generates a position distribution and a movement state distribution of the hypothetical movable object, and records the distributions on the presence probability map; and

a prediction section that moves the hypothetical movable object position distribution based on the hypothetical movable object movement state distribution, changes the moved position distribution based on presence probabilities of the presence probability map, and predicts a future position distribution of the hypothetical movable object on the presence probability map.

19. A hypothetical movable object prediction method comprising:

a step of, by a region detection section, detecting a plurality of categories of regions of track segment and regions of stationary object in a detection target range;

a step of, by a map generation section, generating a presence probability map in which a presence probability expressing a presence likelihood or presence unlikelihood of a moving object is applied to the detected track segment regions and the detected stationary object regions;

a step of, by a blind-spot region identification section, identifying a blind-spot region in the detection target range where a movable object might be present;

a step of, by a hypothetical movable object generation section, generating a hypothetical movable object postulated to be present in the identified blind-spot region;

a step of, by a hypothetical movable object recording section, postulating for the generated hypothetical movable object a position and a behavior state or movement state of the hypothetical movable object, generating a position distribution and a movement state distribution of the hypothetical movable object, and recording the distributions on the presence probability map; and

a step of, by a prediction section, moving the hypothetical movable object position distribution based on the hypothetical movable object movement state distribution, changing the moved position distribution based on presence probabilities of the presence probability map, and predicting a future position distribution of the hypothetical movable object on the presence probability map.