Locating a vehicle based on labeling point cloud data of a scene
A point cloud management system provides labels for each point within a point cloud map. The point cloud management system also provides a method to localize a vehicle using the labeled point cloud. The point cloud management system identifies objects within a scene using an obtained image. The point cloud management system labels the identified objects to register the identified objects against the point cloud. The registration of the objects is then used to localize the vehicle.
1. A point cloud management device comprising:
a database that stores labeled point cloud data; and
processing circuitry configured to
retrieve the labeled point cloud data, the labeled point cloud data identifying objects of a scene,
obtain an image of the scene captured by an image capturing device of a mobile device in a vehicle that is not part of the vehicle,
identify at least one object in the image of the scene,
label the identified at least one object in the image of the scene,
register the labeled at least one object in the image of the scene against the labeled point cloud data, the registration including a projection of the labeled at least one object in the image of the scene to the labeled point cloud data such that a label of each point in the point cloud that coincides with the identified at least one object is registered with the label of the identified at least one object, and
determine a location of the vehicle from the registered at least one object in the image of the scene and the registered point cloud data.
2. The point cloud management device according to claim 1 , wherein the processing circuitry is further configured to:
collect unlabeled point cloud data from one of (i) a sensor of the vehicle accessed by the mobile device in the vehicle that is not part of the vehicle and (ii) a sensor of the mobile device that is different from the image capturing device of the mobile device,
determine a semantic label for at least one point within the unlabeled point cloud data, and
label the at least one point of the point cloud data with the semantic label.
3. The point cloud management device according to claim 1 , wherein the processing circuitry is further configured to:
perform a semantic querying to identify the at least one object in the image of the scene, and
label the at least one object in the image of the scene with a semantic label associated with the identification.
4. The point cloud management device according to claim 3 , wherein the identification of the at least one object in the image of the scene is based on a time and location of when the image was captured by the image capturing device of the mobile device.
5. The point cloud management device according to claim 3 , wherein the identification of the at least one object in the image of the scene is based on dynamic image element removal.
6. A localization method for a vehicle using point cloud data comprising:
retrieving labeled point cloud data, the labeled point cloud data identifying objects of a scene;
obtaining an image of the scene captured by an image capturing device of a mobile device in a vehicle that is not part of the vehicle;
identifying at least one object in the image of the scene;
labelling, by a processor, the identified at least one object in the image of the scene;
registering, by the processor, the labeled at least one object in the image of the scene against the labeled point cloud data, the registering including a projection of the labeled at least one object in the image of the scene to the labeled point cloud data such that a label of each point in the point cloud that coincides with the identified at least one object is registered with the label of the identified at least one object; and
determining, by the processor, a location of the vehicle from the registered at least one object in the image of the scene and the labeled point cloud data.
7. The localization method according to claim 6 , further comprising:
collecting unlabeled point cloud data from one of (i) a sensor of the vehicle accessed by the mobile device in the vehicle that is not part of the vehicle and (ii) a sensor of the mobile device that is different from the image capturing device of the mobile device;
determining, by the processor, a semantic label for at least one point within the unlabeled point cloud data; and
labeling, by the processor, the at least one point of the point cloud data with the semantic label.
8. The localization method according to claim 6 , further comprising:
performing a semantic query, by the processor, to identify the at least one object in the image of the scene; and
labeling, by the processor, the at least one object in the image of the scene with a semantic label associated with said identifying.
9. The localization method according to claim 8 , wherein said identifying of the at least one object in the image of the scene is based on a time and location of when the image was captured by the image capturing device of the mobile device.
10. The localization method according to claim 8 , wherein said identifying of the at least one object in the image of the scene is based on dynamic image element removal.
11. A non-transitory computer-readable storage medium storing computer-readable instructions that, when executed by one or more computers, cause the one or more computers to perform a method comprising:
retrieving labeled point cloud data, the labeled point cloud data identifying objects of a scene;
obtaining the image of the scene captured by an image capturing device of a mobile device in a vehicle that is not part of the vehicle;
identifying at least one object in the image of the scene;
labelling the identified at least one object in the image of the scene;
registering the labeled at least one object in the image of the scene against the labeled point cloud data, the registering including a projection of the labeled at least one object in the image of the scene to the labeled point cloud data such that a label of each point in the point cloud that coincides with the identified at least one object is registered with the label of the identified at least one object; and
determining a location of the vehicle from the registered at least one object in the image of the scene and the labeled point cloud data.
12. The non-transitory computer-readable storage medium according to claim 11 , the method further comprising:
collecting unlabeled point cloud data from one of (i) a sensor of the vehicle accessed by the mobile device in the vehicle that is not part of the vehicle and (ii) a sensor of the mobile device that is different from the image capturing device of the mobile device;
determining a semantic label for at least one point within the unlabeled point cloud data; and
labeling the at least one point within the point cloud data with the semantic label.
13. The non-transitory computer-readable storage medium according to claim 11 , the method further comprising:
performing a semantic query to identify the at least one object in the image of the scene; and
labeling the at least one object in the image of the scene with a semantic label associated with said identifying.
14. The non-transitory computer-readable storage medium according to claim 13 , wherein said identifying of the at least one object in the image of the scene is based on a time and location of when the image was captured by the image capturing device of the mobile device.