Point cloud annotation for a warehouse environment
A system is provided for automatic identification and annotation of objects in a point cloud in real time. The system can automatically annotate a point cloud that identifies coordinates of objects in three-dimensional space while data is being collected for the point cloud. The system can train models of physical objects based on training data, and apply the models to point clouds that are generated by various point cloud generating devices to annotate the points in the point clouds with object identifiers. The solution of automatically annotated point cloud can be used for various applications, such as blueprints, map navigation, and determination of robotic movement in a warehouse.
1. A computer-implemented method using an object identification model, the method comprising:
receiving, from a vehicle, optical scan data and image data in real time as the vehicle moves in a warehouse;
recognizing objects that are represented in the image data;
determining identifiers that represent the objects;
annotating the optical scan data with the identifiers to generate annotated point cloud data; and
transmitting the annotated point cloud data to a display device configured to present a point cloud based on the annotated point cloud data in real time.
2. The computer-implemented method of claim 1 , wherein recognizing objects includes:
training the object identification model; and
identifying the objects using the object identification model.
3. The computer-implemented method of claim 2 , wherein training an object identification model includes:
training the object identification model based on the annotated point cloud data.
4. The computer-implemented method of claim 1 , further comprising:
identifying first annotated objects that are represented in the annotated point cloud data; and
filtering data indicative of the first annotated objects from the annotated point cloud data, wherein the point cloud is presented without the first annotated objects.
5. The computer-implemented method of claim 4 , wherein the first annotated objects include forklifts and humans in the warehouse.
6. The computer-implemented method of claim 1 , further comprising:
retrieving a rule defining undesired objects; and
removing data representative of the undesired objects from the annotated point cloud data.
7. The computer-implemented method of claim 1 , further comprising:
receiving a user selection of an object; and
removing data representative of the object from the annotated point cloud data.
8. The computer-implemented method of claim 1 , wherein the vehicle includes a camera system configured to capture images of the objects and generate the image data, and an optical scan system configured to measure distance to the objects and generate the optical scan data.
9. The computer-implemented method of claim 8 , wherein the optical scan system includes a light detection and ranging (LIDAR) system.
10. The computer-implemented method of claim 1 , wherein the vehicle includes a forklift.
11. The computer-implemented method of claim 1 , further comprising:
generating the annotated point cloud data into a blueprint of the warehouse.
12. A system comprising:
a vehicle;
a camera affixed to the vehicle and configured to capture images and generate image data;
an optical scanner affixed to the vehicle and configured to measure distances to objects and generate spatial scan data; and
a computing device communicatively coupled to the camera and the optical scanner, the computing device configured to perform operations comprising:
receiving, from the vehicle, the image data and the optical scan data in real time as the vehicle moves in a warehouse;
recognizing objects that are represented in the image data;
determining identifiers that represent the objects;
annotating the optical scan data with the identifiers to generate annotated point cloud data; and
transmitting the annotated point cloud data to a display device configured to present a point cloud based on the annotated point cloud data in real time.
13. The system of claim 12 , wherein recognizing objects includes:
training an object identification model; and
identifying the objects using the object identification model.
14. The system of claim 13 , wherein training an object identification model includes:
training the object identification model based on the annotated point cloud data.
15. The system of claim 12 , wherein the operations further comprises:
identifying first annotated objects that are represented in the annotated point cloud data; and
filtering data indicative of the first annotated objects from the annotated point cloud data, wherein the point cloud is presented without the first annotated objects.
16. The system of claim 15 , wherein the first annotated objects include forklifts and humans in the warehouse.
17. The system of claim 12 , wherein the operations further comprises:
retrieving a rule defining undesired objects; and
removing data representative of the undesired objects from the annotated point cloud data.
18. The system of claim 12 , wherein the operations further comprises:
receiving a user selection of an object; and
removing data representative of the object from the annotated point cloud data.
19. The system of claim 12 , wherein the optical scanner includes a light detection and ranging (LIDAR) system.
20. A non-transitory computer-readable storage medium having instructions stored thereon which, when executed by one or more processors, cause the one or more processors to perform operations comprising:
receiving optical scan data and image data as the vehicle moves in a warehouse;
recognizing objects that are represented in the image data;
determining identifiers that represent the objects;
annotating the optical scan data with the identifiers to generate annotated point cloud data;
identifying first annotated objects that are represented in the annotated point cloud data;
removing data indicative of the first annotated objects from the annotated point cloud data; and
transmitting the annotated point cloud data to a display device configured to present a point cloud based on the annotated point cloud data, the point cloud being presented without the first annotated objects.