Systems and methods for static detection based amodalization placement

1 . A method for controlling a mobile platform, comprising:

obtaining, by a computing device, loose-fit cuboids overlaid on 3D graphs so as to each encompass lidar data points associated with an object;

checking, by the computing device, whether or not the object is static through a time period by performing a cuboid dimension check of the loose-fit cuboids;

identifying, by the computing device, the lidar data points associated with the object inside the loose-fit cuboids and remove ground points;

defining, by the computing device, a tight-fit cuboids by transferring the loose-fit cuboids from a first coordinate system into a second coordinate system;

defining, by the computing device, an amodal cuboid based on the loose-fit cuboids and the tight-fit cuboids;

identifying, by the computing device, a center for the amodal cuboid in response to the object being static; and

causing, by the computing device, driving operations of the mobile platform to be controlled based on the amodal cuboid having the center which was identified.

2 . The method according to claim 1 , wherein the identifying comprises considering a center of one of said loose-fit cuboids as a center for the amodal cuboid when the object is static.

3 . The method according to claim 1 , wherein the checking further comprises performing at least one of a cuboid centroid check and a cuboid rotation check on the loose-fit cuboids.

4 . The method according to claim 3 , wherein the checking further comprises classifying the object as static in a time frame when the cuboid dimension check, the cuboid centroid check and the cuboid rotation check indicate that the object is associated with a same cuboid size, orientation and centroid at two consecutive times in the time period.

5 . The method according to claim 3 , wherein the cuboid dimension check comprises:

obtaining dimensions for a first loose-fit cuboid in a first one of the 3D graphs plotting lidar data points of a first time frame;

obtaining dimensions for a second loose-fit cuboid in a second one of the 3D graphs plotting lidar data points of a second time frame;

computing differences between dimensions of the first loose-fit cuboid and dimensions of the second loose-fit cuboid; and

respectively comparing the differences to threshold dimension values.

6 . The method according to claim 5 , further comprising classifying the object as non-static in the second time frame when one or more of the differences are greater than the threshold dimension values.

7 . The method according to claim 5 , wherein the cuboid centroid check is performed when the differences are all less than the threshold dimension values.

8 . The method according to claim 3 , wherein the cuboid centroid check comprises:

obtaining a distance between a centroid of a first loose-fit cuboid of the loose-fit cuboids and a centroid of a second loose-fit cuboid of the loose-fit cuboids; and

comparing the distance to a threshold distance value.

9 . The method according to claim 8 , further comprising classifying the object as non-static in a second time frame associated with the second loose-fit cuboid, when the distance is greater than the threshold distance value.

10 . The method according to claim 8 , wherein the cuboid rotation check is performed when the distance is less than the threshold distance value.

11 . The method according to claim 3 , wherein the cuboid rotation check comprises:

obtaining a difference between yaw values associated with first and second loose-fit cuboids of the loose-fit cuboids; and

comparing the difference to a threshold yaw difference value.

12 . The method according to claim 11 , further comprising classifying the object as non-static in a second time frame associated with the second loose-fit cuboid, when the difference is greater than the threshold yaw difference value.

13 . The method according to claim 1 , wherein the defining the amodal cuboid includes:

aligning a corner of the amodal cuboid with a closest visible corner of the tight-fit cuboid; and

aligning a corner of the amodal cuboid such that a boundary precision between the amodal cuboid and a loose-fit cuboid is maximized.

14 . A system, comprising:

a processor;

a non-transitory computer-readable storage medium comprising programming instructions that are configured to cause the processor to implement a method for controlling a mobile platform, wherein the programming instructions comprise instructions to:

obtain loose-fit cuboids overlaid on 3D graphs so as to each encompass lidar data points associated with an object;

check whether or not the object is static through a time period by performing a cuboid centroid check on the loose-fit cuboids;

identify the lidar data points associated with the object inside the loose-fit cuboids and remove ground points;

defining a tight-fit cuboids by transferring the loose-fit cuboids from a first coordinate system into a second coordinate system;

defining an amodal cuboid based on the tight-fit cuboids;

identify a center for the amodal cuboid in response to the object being static; and

cause driving operations of the mobile platform to be controlled based on the amodal cuboid having the center which was identified.

15 . The system according to claim 14 , wherein a center of one of the loose-fit cuboids is considered a center for the amodal cuboid when the object is static.

16 . The system according to claim 14 , wherein the check as to whether or not the object is static through the time period is achieved by further performing at least one of a cuboid dimension check, and a cuboid rotation check on the loose-fit cuboids.

17 . The system according to claim 16 , wherein the cuboid dimension check comprises:

obtaining dimensions for a first loose-fit cuboid in a first one of the 3D graphs plotting lidar data points of a first time frame;

obtaining dimensions for a second loose-fit cuboid in a second one of the 3D graphs plotting lidar data points of a second time frame;

computing differences between dimensions of the first loose-fit cuboid and dimensions of the second loose-fit cuboid; and

respectively comparing the differences to threshold dimension values.

18 . The system according to claim 16 , wherein the cuboid centroid check comprises:

obtaining a distance between a centroid of a first loose-fit cuboid of the loose-fit cuboids and a centroid of a second loose-fit cuboid of the loose-fit cuboids; and

comparing the distance to a threshold distance value.

19 . The system according to claim 16 , wherein the cuboid rotation check comprises:

obtaining a difference between yaw values associated with first and second loose-fit cuboids of the loose-fit cuboids; and

comparing the difference to a threshold yaw difference value.

20 . A non-transitory computer-readable medium that stores instructions that are configured to, when executed by at least one computing device, cause the at least one computing device to perform operations comprising:

obtaining loose-fit cuboids overlaid on 3D graphs so as to each encompass lidar data points associated with an object;

defining an amodal cuboid based on the loose-fit cuboids;

checking whether or not the object is static through a time period by performing a cuboid rotation check on the loose-fit cuboids;

identifying a center for the amodal cuboid based on said checking;

identifying the lidar data points associated with the object inside the loose-fit cuboids and remove ground points;

defining a tight-fit cuboids by transferring the loose-fit cuboids from a first coordinate system into a second coordinate system; and

causing driving operations of a mobile platform to be controlled based on the amodal cuboid having the center which was identified.