Granted Patent B2

US 12,552,375 · App. 18/241,576 · Granted Feb 17, 2026

Inventors: Hongkai Yu (San Diego, CA); Zhipeng Yan (San Diego, CA); Panqu Wang (San Diego, CA); Pengfei Chen (San Diego, CA)

Assignee: CreateAI, Inc.

B60W30/0956G06F18/2155G06F18/24G06F18/2413G06V10/25G06V10/44G06V10/7753G06V10/87G06V20/56G06V20/58G05D1/0221

Abstract

Claims (37)

1 . A system comprising:

a data processor; and

an autonomous vehicle occlusion detection system, executable by the data processor, the autonomous vehicle occlusion detection system being configured to:

extract an object from a frame of image data;

use a trained image classifier to identify the extracted object;

obtain predefined features and contours of the identified object;

detect an occlusion status of the identified object based on the predefined features and contours; and

perform object-level contour detection on the identified object.

2 . The system of claim 1 wherein the trained image classifier is trained on image data received from a training image data collection system including an array of perception information gathering devices or sensors.

3 . The system of claim 1 wherein the object-level contour detection includes information indicative of a status for the identified object, the status comprising a state from the group consisting of: 1) occluding another object; (2) occluded by another object; (3) there is no overlap with another object, and (4) in between two objects.

4 . The system of claim 1 being configured to associate a plurality of dimensions with the identified object.

5 . The system of claim 1 being configured to apply a trained static image classifier to the extracted object.

6 . The system of claim 1 being configured to apply a bounding box to the extracted object, the bounding box being partitioned into a plurality of portions.

7 . The system of claim 1 being configured to apply a pixel-level object label and bounding box to the extracted object.

8 . A computer-implemented vehicle occlusion detection method comprising:

extracting an object from a frame of image data;

using a trained image classifier to identify the extracted object;

obtaining predefined features and contours of the identified object;

detecting an occlusion status of the identified object based on the predefined features and contours; and

performing object-level contour detection on the identified object.

9 . The method of claim 8 wherein the trained image classifier is trained on image data received from a training image data collection system including an array of perception information gathering devices or sensors.

10 . The method of claim 8 wherein the trained image classifier is trained on image data including ground truth data corresponding to manually generated object labeling.

11 . The method of claim 8 wherein the object-level contour detection includes information indicative of a status for the identified object, the status comprising a state from the group consisting of: 1) occluding another object; (2) occluded by another object; (3) there is no overlap with another object, and (4) in between two objects.

12 . The method of claim 8 wherein the extracted feature is a vehicle object.

13 . The method of claim 8 including associating a plurality of dimensions with the identified object.

14 . The method of claim 8 including tracking a plurality of identified objects across a plurality of image frames.

15 . A non-transitory machine-useable storage medium embodying instructions which, when executed by a machine, cause the machine to:

extract an object from a frame of image data;

use a trained image classifier to identify the extracted object;

obtain predefined features and contours of the identified object;

detect an occlusion status of the identified object based on the predefined features and contours; and

perform object-level contour detection on the identified object.

16 . The non-transitory machine-useable storage medium of claim 15 wherein the trained image classifier is trained on image data received from a training image data collection system including an array of perception information gathering devices or sensors.

17 . The non-transitory machine-useable storage medium of claim 15 wherein the object-level contour detection includes information indicative of a status for the identified object, the status comprising a state from the group consisting of: 1) occluding another object; (2) occluded by another object; (3) there is no overlap with another object, and (4) in between two objects.

18 . The non-transitory machine-useable storage medium of claim 15 wherein the object-level contour detection includes an edge map.

19 . The non-transitory machine-useable storage medium of claim 15 being further configured to apply a pixel-level object label to the extracted object.

20 . The non-transitory machine-useable storage medium of claim 15 wherein the trained image classifier is a trained neural network.

Assignments (3)

Continuity (5)

References Cited (186)