Lidar debris detection based on annotated image data

Patent metadata

Field	Value
Publication number	US-2025298135-A1
Application number	US-202519232050-A
Country	US
Kind code	A1
Filing date	Jun 9, 2025
Priority date	Oct 31, 2022
Publication date	Sep 25, 2025
Grant date	—

Abstract

Techniques for identifying lidar points associated with static objects, and using such lidar points to annotate objects within two-dimensional images are discussed herein. In some examples, an object manager may receive accumulations of lidar data captured from lidar devices of a vehicle while traversing within a driving environment. In some examples, the object manager may receive a plurality of annotated images. Such annotations may identify static objects within the driving environment. In some instances, the object manager may project a lidar point into an annotated image and determine that the lidar point is associated with an annotated pixel. Based on the pixel being associated with the annotated object, the object manager may determine that the lidar point is associated with object. In some examples, the object manager may determine a subset of lidar points that are associated with the object.

First claim

What is claimed is:

1 . A system comprising:
one or more processors; and one or more non-transitory computer-readable media storing computer-executable instructions that, when executed, cause the system to perform operations comprising:
receiving lidar data;
receiving image data including a set of pixels;
receiving an annotation associated with an object within the image data;
projecting a lidar point of the lidar data into the image data;
determining that the lidar point is associated with a pixel of the set of pixels; and
determining, based at least in part on determining that the pixel is associated with the annotation, that the lidar point is associated with the object.

2 . The system of claim 1 , wherein the image data is first image data and the set of pixels is a first set of pixels, wherein the first image data is captured at a first time, wherein determining that the lidar point is associated with the object further comprises:
receiving second image data that includes a second set of pixels, wherein the second image data is associated with a second time that is different than the first time; receiving second annotation associated with the object within the second image data; projecting the lidar point into the second image data; determining that the lidar point is associated with a second pixel of the second set of pixels; and determining, based at least in part on the pixel being associated with the annotation and the second pixel being associated with the second annotation, that the lidar point is associated with the object.

3 . The system of claim 1 , wherein determining that the lidar point is associated with the object comprises:
projecting the lidar point into a plurality of annotated images; determining, based at least in part on projecting the lidar point into the plurality of annotated images, that the lidar point is associated with an annotated pixel in a number of the plurality of annotated images; and determining, based at least in part on determining that the number of the plurality of annotated images meets or exceeds a threshold number of images, that lidar point is associated with the object.

4 . The system of claim 1 , wherein projecting the lidar point into the image data is based at least in part on:
determining a first transformation of the lidar point from a lidar device frame of reference to a global reference frame; determining a second transformation of the lidar point from the global reference frame to a vehicle reference frame; and determining a third transformation of the lidar point from the vehicle reference frame to a reference frame of an image capturing device that captured the image data.

5 . The system of claim 1 , wherein receiving the lidar data comprises:
determining an accumulation of lidar points received from a first time to a second time that is different than the first time; causing, as the lidar data, the accumulation of lidar points to be represented in a common reference frame.

6 . The system of claim 1 , the operations further comprising:
generating training data comprising the image data and the lidar point; training a machine-learned model based on the training data; and controlling a vehicle based at least in part on the machine-learned model.

7 . One or more non-transitory computer-readable media storing instructions executable by one or more processors, wherein the instructions, when executed, cause a system to perform operations comprising:
receiving image data; receiving segmented lidar data; projecting the segmented lidar data into the image data; determining, based at least in part on projecting the segmented lidar data into the image data, that the segmented lidar data corresponds to a set of pixels; and determining, based at least in part on the segmented lidar data corresponding to the set of pixels, a contour associated with the set of pixels.

8 . The one or more non-transitory computer-readable media of claim 7 , wherein determining the contour is based at least in part on:
determining, based at least in part on the segmented lidar data corresponding to the set of pixels, a set of dilated pixels; determining segment identifiers associated with the set of dilated pixels; and determining, based at least in part on the segment identifiers, that the set of dilated pixels are associated with an object.

9 . The one or more non-transitory computer-readable media of claim 7 , the operations further comprising:
generating training data based at least in part on the contour; and training a machine-learned model to detect static objects within environments, wherein input to the machine-learned model includes the training data.

10 . The one or more non-transitory computer-readable media of claim 7 , wherein the contour is a first contour, the operations further comprising:
projecting the segmented lidar data into the image data; determining that the segmented lidar data are associated with a second set of pixels; determining that the second set of pixels is associated with a second object; and determining a second contour representing the second object in an environment.

11 . The one or more non-transitory computer-readable media of claim 7 , the operations further comprising,
determining that a portion of the set of pixels are located within the contour; and determining, based at least in part on the portion of the set of pixels being within the contour, that the portion of the set of pixels are associated with an object.

12 . The one or more non-transitory computer-readable media of claim 7 , wherein receiving the segmented lidar data is based at least in part on:
receiving lidar data; receiving second image data and an annotation associated with an object in the second image data; projecting the lidar data into the second image data; and determining, based at least in part on determining that a pixel of the second image data associated with the lidar data is associated with the annotation, that the lidar data is associated with the object.

13 . The one or more non-transitory computer-readable media of claim 7 , the operations further comprising:
projecting lidar data into annotated images; determining, based at least in part on projecting the lidar data into the annotated images, that the lidar data is associated with an annotated pixel in a number of the annotated images; and determining, as the segmented lidar data and based at least in part on determining that the number of the annotated images meets or exceeds a threshold number of images, that lidar data is associated with an object.

14 . The one or more non-transitory computer-readable media of claim 7 , wherein the set of pixels correspond to an object.

15 . A method comprising:
receiving a lidar point; receiving a set of pixels; receiving an object segmentation associated with the set of pixels; projecting the lidar point into the set of pixels; determining that the lidar point is associated with a pixel of the set of pixels; and determining, based at least in part on the lidar point being associated with the pixel, that the lidar point is associated with the object segmentation.

16 . The method of claim 15 , wherein the set of pixels is a first set of pixels, wherein the first set of pixels is captured at a first time, further comprising:
determining that the lidar point is associated with an object, wherein determining that the lidar point is associated with the object further comprises:
receiving second set of pixels associated with a second time that is d

Assignees

• Zoox Inc

Inventors

• Mahieu Robert Evan
• Pfeiffer David
• Wang Zeng

Classifications

• G06T2207/20081

  Training; Learning · CPC title

• G06T7/11

  Region-based segmentation · CPC title

• G06T7/174

  involving the use of two or more images · CPC title

• B60W2420/408

  Radar; Laser, e.g. lidar · CPC title

• G01S17/89

  for mapping or imaging · CPC title