Method and apparatus of detecting object using event-based sensor

1. An object detection method comprising:

generating an event image based on an event signal output by an event-based sensor, the event image comprising a plurality of target pixels and a plurality of neighbor pixels;

determining a feature vector based on the plurality of target pixels and the plurality of neighbor pixels;

determining a target object corresponding to the plurality of target pixels based on the feature vector; and

verifying a type of the target object based on a valid range around a position of the target object, wherein the valid range is determined based on a previous position of the target object and a possible movable range of the type of the target object with respect to a neighbor object corresponding to the plurality of neighbor pixels.

2. The object detection method of claim 1 , wherein the determining the target object comprises:

inputting the feature vector into a classifier that is trained by a learning sample comprising a target area and a neighbor area adjacent to the target area, and determining the target object based on a result output by the classifier.

3. The object detection method of claim 1 , wherein the determining the target object comprises:

determining the type of the target object and the position of the target object.

4. The object detection method of claim 1 , wherein the determining the target object comprises:

determining the position of the target object based on positions of pixels corresponding to the target object.

5. The object detection method of claim 1 , wherein the determining the feature vector comprises:

segmenting the event image into a plurality of areas; and

sampling the plurality of neighbor pixels in a neighbor area adjacent to a target area that comprises at least one target pixel among the plurality of areas.

6. The object detection method of claim 5 , wherein the sampling the plurality of neighbor pixels comprises:

sampling a preset number of pixels in the neighbor area.

7. The object detection method of claim 1 , wherein the possible movable range of the type of the target object is determined based on a connection relationship between the target object and the neighbor object.

8. The object detection method of claim 1 , further comprising:

determining a motion trajectory of the target object based on the position of the target object; and

generating an action command corresponding to the motion trajectory.

9. The object detection method of claim 8 , wherein the generating the action command comprises:

segmenting the motion trajectory into a plurality of action segments;

extracting information about the plurality of action segments; and

generating the action command based on the information about the plurality of action segments,

wherein the information about the plurality of action segments comprises at least one of position information, route information, movement direction information, speed information, and acceleration information.

10. The object detection method of claim 8 , wherein the generating the action command comprises:

combining a plurality of different objects into a combined object and determining a motion trajectory of the combined object based on a motion trajectory of each of the plurality of different objects;

extracting information about the motion trajectory of the combined object; and

generating the action command based on the information about the motion trajectory of the combined object,

wherein the information about the motion trajectory of the combined object comprises at least one of position information, route information, movement direction information, speed information, and acceleration information.

11. A learning method comprising:

generating a learning sample comprising a target area and a neighbor area adjacent to the target area, wherein a learning target type of the learning sample comprises a type of a target object corresponding to the target area and a type of a neighbor object corresponding to the neighbor area;

training a classifier to identify the type of the target object and the type of the neighbor object based on the learning sample,

wherein the training the classifier comprises adjusting a parameter of the classifier based on the learning target type of the learning sample and a classification result of the classifier with respect to the type of the target object and the type of the neighbor object, and

wherein the identified type of the neighbor object is used to verify the identified type of the target object.

12. The learning method of claim 11 , wherein the generating the learning sample comprises:

generating a sample image based on an event signal of an event-based sensor;

segmenting the sample image into a plurality of areas; and

configuring target pixels included in the target area and neighbor pixels included in the neighbor area from among the plurality of areas as the learning sample.

13. The learning method of claim 11 , wherein the training the classifier comprises training the classifier based on a deep belief network (DBN).

14. An object detection apparatus comprising:

a processor configured to generate an event image based on an event signal output by an event-based sensor, the event image comprising a plurality of target pixels and a plurality of neighbor pixels; and

a classifier configured to determine a feature vector based on the plurality of target pixels and the plurality of neighbor pixels, and to determine a target object corresponding to the plurality of target pixels based on the feature vector; and

a verifier configured to verify a type of the target object based on a valid range around a position of the target object, wherein the valid range is determined based on a previous position of the target object and a possible movable range of the type of the target object with respect to a neighbor object corresponding to the plurality of neighbor pixels.

15. The object detection apparatus of claim 14 , wherein the classifier is trained by a learning sample comprising a target area and a neighbor area adjacent to the target area.

16. The object detection apparatus of claim 14 , wherein the possible movable range of the type of the target object is determined based on a connection relationship between the target object and the neighbor object.