Technologies for fusing data from multiple sensors to improve object detection, identification, and localization
Technologies for performing sensor fusion include a compute device. The compute device includes circuitry configured to obtain detection data indicative of objects detected by each of multiple sensors of a host system. The detection data includes camera detection data indicative of a two or three dimensional image of detected objects and lidar detection data indicative of depths of detected objects. The circuitry is also configured to merge the detection data from the multiple sensors to define final bounding shapes for the objects.
1. At least one non-transitory machine-readable storage medium comprising instructions, which when executed by processor circuitry of a computing device, cause the processor circuitry to perform operations to:
obtain sensor data representative of an environment of a vehicle, the sensor data including:
camera data providing at least one image of an object in the environment, and lidar data providing depth information of the object in the environment;
synchronize the sensor data based on timestamps indicating time of capture for the camera data and the lidar data;
define, from the camera data obtained at the time of capture, a first bounding box for the object;
define, from the lidar data obtained at the time of capture, a second bounding box for the object; and
fuse the first bounding box and second bounding box to define a fused bounding shape of the object, wherein fusion of the first bounding box and second bounding box is based on a confidence value that the first bounding box and the second bounding box refer to the object as the same object, wherein a distance of the object from the vehicle is used as a parameter for the fusion of the first bounding box and the second bounding box.
2. The non-transitory machine-readable storage medium of claim 1 , the instructions to further cause operations to:
identify the object from the camera data, wherein the camera data comprises at least one two-dimensional image that indicates the identified object in the environment.
3. The non-transitory machine-readable storage medium of claim 2 , wherein the first bounding box is created from the camera data using a plurality of features corresponding to the identified object.
4. The non-transitory machine-readable storage medium of claim 1 , the instructions to further cause operations to:
identify the object from the lidar data, wherein the lidar data comprises at least one three-dimensional point cloud that indicates the depth information of the identified object in the environment.
5. The non-transitory machine-readable storage medium of claim 4 , wherein the second bounding box is created from the lidar data using the depth information of the identified object.
6. The non-transitory machine-readable storage medium of claim 1 , the instructions to further cause operations to:
determine a class of the object;
wherein the fused bounding shape of the object is based on the class of the object.
7. The non-transitory machine-readable storage medium of claim 1 , wherein the object is another vehicle or a pedestrian.
8. The non-transitory machine-readable storage medium of claim 1 , wherein the fusion of the first bounding box and the second bounding box is further based on intersection-over-union (IOU) between the first bounding box and the second bounding box.
9. The non-transitory machine-readable storage medium of claim 1 , wherein the sensor data further includes radar data that is captured from a radar system of the vehicle, and wherein the fused bounding shape of the object is further based on corresponding radar objects detected from the radar data.
10. A computing device, comprising:
at least one interface to receive sensor data representative of an environment of a vehicle, the sensor data including: camera data providing at least one image of an object in the environment, and lidar data providing depth information of the object in the environment; and
at least one processor configured to perform operations to:
synchronize the sensor data based on timestamps indicating time of capture for the camera data and the lidar data;
define, from the camera data obtained at the time of capture, a first bounding box for the object;
define, from the lidar data obtained at the time of capture, a second bounding box for the object; and
fuse the first bounding box and second bounding box to define a fused bounding shape of the object, wherein fusion of the first bounding box and second bounding box is based on a confidence value that the first bounding box and the second bounding box refer to the object as the same object, wherein a distance of the object from the vehicle is used as a parameter for the fusion of the first bounding box and the second bounding box.
11. The computing device of claim 10 , the at least one processor further configured to perform operations to:
identify the object from the camera data, wherein the camera data comprises at least one two-dimensional image that indicates the identified object in the environment.
12. The computing device of claim 11 , wherein the first bounding box is created from the camera data using a plurality of features corresponding to the identified object.
13. The computing device of claim 10 , the at least one processor further configured to perform operations to:
identify the object from the lidar data, wherein the lidar data comprises at least one three-dimensional point cloud that indicates the depth information of the identified object in the environment.
14. The computing device of claim 13 , wherein the second bounding box is created from the lidar data using the depth information of the identified object.
15. The computing device of claim 10 , the at least one processor further configured to perform operations to:
determine a class of the object;
wherein the fused bounding shape of the object is based on the class of the object.
16. The computing device of claim 10 , wherein the object is another vehicle or a pedestrian.
17. The computing device of claim 10 , wherein the fusion of the first bounding box and the second bounding box is further based on intersection-over-union (IOU) between the first bounding box and the second bounding box.
18. The computing device of claim 10 , wherein the sensor data further includes radar data that is captured from a radar system of the vehicle, and wherein the fused bounding shape of the object is further based on corresponding radar objects detected from the radar data.